January 26, 2025 • 51 mins
This episode focuses on the evolving landscape of data optimization, emphasizing the intersections of standards, sustainability, and community collaboration. From tackling large-scale implementations to enhancing environmental efficiency, we investigate how SNIA's approach to managing data across various infrastructures has broadened beyond traditional storage paradigms. Our experts cover key trends, challenges, and solutions that organizations need to address data management and sustainably, while optimizing their infrastructure for future growth, discussing:
• Exploration of the concept of optimization in data storage
• The blurred lines between data characteristics and storage infrastructure
• Importance of sustainability in storage practices and life cycle assessments
• The role of SNIA Swordfish and Redfish standards in enhancing interoperability
• Community engagement and collaboration as key drivers for innovation in data management
SNIA is an industry organization that develops global standards and delivers vendor-neutral education on technologies related to data. In these interviews, SNIA experts on data cover a wide range of topics on both established and emerging technologies.
About SNIA:
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:01):
All right, everybody.
Welcome to the SNEA Experts onData podcast.
My name is Eric Wright, I'm thehost here of the SNEA EOD and
I'm also the co-founder of GTMDelta Super excited to have an
amazing crew here today with us.
We're going to talk aboutoptimization.
This is a place that I've beenliving in a long time, both in
(00:23):
the work that I've done and heckon a personal day-to-day basis.
So if I could tell you how tooptimize things, I'd love to.
But even better, I've got threeof the most amazing people who
are going to tell us about whatthe whole SNEA community is
doing around our data focusareas.
Today we're going to talk aboutOptimize.
With that, I'm going to do aquick roundtable, do a quick
(00:45):
introduction.
So we'll start off with ChrisDo you want to lead us out?
And then just a quick littlebio on you, and then we'll jump
to Rochelle and then JM.
Speaker 2 (00:54):
Sure, my name is Chris Leonetti.
I've been a reference architectfor various companies over the
years for probably what the past?
Oh my gosh 30 years now.
So I've implemented very largesolutions all the way down to
very small solutions and largeI'm talking $3 billion kind of
solutions.
(01:14):
But then I've also architecteda lot of solutions that you know
are in the $50,000 range tosolve business needs.
Speaker 3 (01:25):
So Chris has been doing that since he was a
toddler, clearly, but moreimportantly, he is the secretary
of these things.
Correct Secretary of thesethings, yeah, and wears many
hats around our organization.
So I'm Rochelle Alvers.
I am the vice chair of SNA aswell as the chair of the
(01:50):
Swordfish TWIG.
You guys are hearing it herefirst, we just renamed the SSM
TWIG which nobody ever knew whatthat acronym stood for to the
Swordfish TWIG, so that you knowwhat we actually do.
And I'm also the chair of thestorage management community,
which helps promote theeducation, promotion of
(02:14):
everything related to storagemanagement within SNEA, and so
that's actually one of thethings we'll be talking about
here in a little bit.
Oh, sorry, I forgot my day job.
Day job I lead the technologyinitiatives and ecosystem
enabling for Intel, and I'vealso been working on things
related to everything forstorage management and
(02:39):
technology initiatives since Iwas a toddler.
All right, john, I'm JohnMichael.
Speaker 4 (02:43):
Hans, I'm Senior Director of Product Planning at
B a toddler.
All right, john, I'm JohnMichael Hands.
I'm Senior Director of ProductPlanning at FADU, so right now
we're working on enterprise SSDcontrollers for hyperscale.
That's the day job.
I also co-chair the SNEA SSDSIG, so we wrote stuff, like you
know the form factors page onEVSF, the SNEA TCO model, which
is also used by lots ofhyperscalers, and other fun
(03:05):
stuff.
I also am the secretary of a501C nonprofit called Circular
Drive Initiative, where we arefocused on sustainability of
storage, and then very activelyengaged in the Open Compute
Project, sustainabilityInitiative.
So I co-chair one of the workgroups there.
Speaker 1 (03:22):
Everybody's side.
Jobs are like full-time jobs.
It's amazing that you can keepall this on track, but I've seen
you all in action so I know howyou pull it off.
You're a bunch of fantasticfolks.
So, Chris, if you want to,let's just sort of do an initial
intro to people about what itis that we talk about when we
(03:43):
mean the optimize data focusarea and what this is under the
SNEA sort of new description aswe talk about these data focus
areas.
Speaker 2 (03:53):
Yep.
So when people say the wordstorage, they're kind of like
asking how long is a piece ofstring.
There are so many answers towhat storage really means.
A good example is storagemanagement is something that's
desperately needed out there.
That's highly valuable, but itgoes down different paths than
(04:21):
storage environmentals, so like,for instance, the green twig.
So we have different focusareas for storage that deal with
the different aspects of whatstorage does and what storage is
.
So, for instance, we have onegroup that focuses on form
factor.
We have another group thatfocuses on manageability.
We have a third group thatfocuses on efficiency when it
comes to power, that kind ofthing.
(04:42):
The hives of SNEA are basicallyset up to explore all those
different avenues that storagecan manifest itself in.
In fact, rochelle can probablygo into more detail about how
these hives are laid out.
Speaker 3 (05:00):
Yeah, we have six different data focus areas that
we've defined.
Yeah, we have six differentdata focus areas that we've
defined as we've been expandingSNEA over the years.
Snea started as the StorageNetworking Industry Association,
very focused on SANs.
Over time, we expanded to allthings storage, and what we've
been expanding to over the lastfew years is really a
(05:21):
recognition, and the reason wecame up with these data focus
areas is really recognition thatwe're beyond just storage.
We're really looking at datafrom the perspective of storage.
So when we're talking about thisparticular data focus area,
which is the optimizinginfrastructure for data, it's,
it really is.
How are we looking at, you know, all aspects of data and
(05:46):
optimizing the infrastructurefor data?
It is from the perspective ofstorage, but it's really broader
than just the storage elementsin the system.
Everything we're doing withinSNEA is really beyond just the
storage of them.
(06:06):
It's the entire data, theentire ecosystem managing data,
including everything frommanaging data accelerators,
managing this storage fabrics,managing the storage elements.
But it's definitely a broaderpicture than it was, you know,
five or 10 years ago.
Speaker 2 (06:28):
And an example of that would be someplace that the
line blurs between storage anddata.
In fact, you can't reallyoptimize the storage
infrastructure without knowingthe data that's going to be on
it.
And what I mean by that is,let's say, you've got a dataset
that's highly compressible orhighly dedupeable.
You might optimize differentstorage for that dataset than
(06:49):
you would for a dataset that hasno compressibility and no
duplication ability.
You also have to be able tochoose between object files and
block.
So there's a lot ofoptimizations that actually
matter more about what the datais than simply a place to store
it.
You can't simply look atstorage as a box you put on a
wall, that you put things in.
You have to really look atstorage as the optimal way of
(07:14):
putting these boxes together insuch a way to solve the needs or
the requirements of the dataengineer.
The data engineer is going tosay I need this kind of data
stored in this way, with thisIOPS, with this envelope, with
this reliability.
There's a lot of features youhave to build into that design,
into that solution that reallylets you optimize.
So the lines between pure dataand pure infrastructure are so
(07:38):
blurry that they kind of don'texist anymore.
Speaker 1 (07:40):
Well, and even optimize for purpose is even.
You know we can talk aboutoptimize for performance,
optimize for capacity and JM.
You know.
Let's maybe talk for a secondabout optimize for
sustainability, because I knowthat's one area that you're
bringing a lot of focus to anddoing a lot of work on.
How does this play into thesustainability side of of you
(08:03):
know the world and obviouslysnea?
Speaker 4 (08:06):
yeah, uh, we had a really good presentation at snea
, stc this, uh, I guess thisjust a couple months ago on, uh,
what sustainability means tostorage and really the impact
you know for the SNEA projectsreally, which I focused around
circularity, and you know, mediasanitization, which are two
really key aspects.
But when I think, when peoplehear sustainability in the
(08:29):
storage world, they usuallythink about energy efficiency
and that is one really importantaspect of sustainability, right
, it it's like you know, this ishow much energy you're using
during the use phase, when thedevice is powered on.
But I think people forgetreally the other important part,
which is what we call itembodied carbon, which is how
much did it take to manufacturethat device and get it from
(08:50):
point a to point b before iteven gets to the service?
How much does it take todecommission and end of life,
this full life cycle of thedevice?
So Flash, specifically like SSDsand memory, you know, can take
a tremendous amount ofelectricity and resources to
manufacture.
So one of the what ourpresentation was about at STC
was the idea of circularity,which you could basically
(09:12):
sanitize the device, remove allthe user data and then have it
get a second use in another usecase and that might be an
emerging use case somewhere thathas different performance TCO
requirements.
You know SSDs.
You can look at eBay right now.
You can see drives that Iworked on at Intel in 2011.
They're still up there, they'restill working.
So everybody kind ofintuitively knows this that the
(09:33):
devices can last a lot longerthan a five-year warranty, and
so the main barrier is that hasbeen actually being able to
trust that, when you remove thedata from the drive, that you
can reuse it.
So that was the main topic ofour presentation at CISDC.
Speaker 2 (09:46):
And that's a good example of where the hives cross
over too, because you've got agroup like the energy efficiency
side of the house that wants tobe able to reuse these drives.
You've also got the securityside of the house that wants to
make sure that no data leaksfrom your infrastructure.
So you've got to play thoseagainst each other, and at-rest
encryption is one way to solvethat problem, because you can do
a digital shred, but that maynot be good enough for some
(10:07):
businesses.
It may be for others.
So I mean the different hives.
Although we've got theseseparate hives that handle all
the technology, sne as a wholeis all about collaboration
between the hives as well.
Speaker 3 (10:20):
And the place we bring that all together is in
the standardization of themanageability and the
infrastructure where you canexpose in a standardized way all
of those use cases to say, Ihave the instrumentation where I
can report and control andreally prove that all of those
(10:40):
use cases are true, so I canensure that the security is in
place and comply to any you know, any government regulations you
know, and ensure that, yes, Ihave the security in place and I
am enforcing it.
But I'm also reporting back andproviding the metrics to show
(11:04):
that reporting on and theinstrumentation are in place to
say, yes, that all of thosesustainability metrics are in
place, to show that, yes, I amactually getting the green
aspects, the powerinstrumentation and that all of
those things are actuallytrending to the exact reporting
(11:28):
that we are anticipating.
Speaker 1 (11:31):
When we have this interesting thing too of what's
really important about the SNEAas a community and as a
standards body, what we look at,why these things are important,
because the artifacts willcontinuously change, but the way
in which we access and managethose artifacts is where we
(11:51):
standardize.
This is the reason why we knowHTTP has been a standard for so
long, and CRUD and simple thingsin regular development.
But that goes all the way downto physical hardware and people
often lose sight of that.
And I remember seeing the earlywork with, watching what was
going on when Swordfish was, youknow in its earlier days, and I
thought, well, this ismagnificent.
(12:12):
The trick was always gettingeverybody to come on board.
But once that happens, like thepace of innovation was so much
more rapid because we knew everysoftware abstraction layer
could use the same common youknow ways, semantic architecture
, in order to access thesehardware layers.
And then now that meant theinnovation that happens down
(12:34):
there becomes you know, same wayas we look at at Tesla's with
you know full, you know FSD andstuff, where it's now a software
update, a firmware update, toaccess the hardware.
That hasn't changed but we canbetter use the hardware and this
is happening in in storage andwe.
It seems wondrous andmagnificent in the future, but
(12:55):
it's.
It's actually.
It's.
It's been the future for a longtime.
We've been living the futurefor quite a while.
Speaker 3 (13:00):
Yes, exactly, and so so I want to go back to you know
, talking about accelerating howthis works to something that
John Michael said when he wasintroducing himself, about you
know all the different hats he'swearing about working with
circular drive initiative, aboutworking in OCP.
This work is not exclusivelyhappening in the Swordfish twig.
Swordfish is built on top ofRedfish.
(13:21):
A lot of this instrumentationis actually common to Redfish.
Redfish works heavily with OCPon the base instrumentation.
So a lot of these metrics thatwe're developing are actually
common metrics to Redfish thatare actually common
instrumentation to profiles thatare being developed in OCP for
servers.
So we have a lot of leveragebetween servers, between OCP,
(13:45):
between CNA, so the profile andthe instrumentation are actually
common between about four orfive different organizations.
So, as the Circular DriveInitiative says, hey, this is
how we want the profiles to bedeveloped and the metrics,
there's actually very littlelift we need to do to say this
(14:07):
is what we want for thisspecific storage profile.
It's really okay.
I have a couple of these samemetrics that we have on that
server instrumentation that wejust need to apply over here and
we need a new profile and it'sincredibly common
instrumentation.
It's not a big lift.
It's a very small lift thatthen needs to be developed for
(14:30):
this representation.
So it's really that, how to?
What is that specific use casethat we can then apply?
It's not a huge lift and sothat's how we can actually do
rapid instrumentation, rapiddevelopment on that specific use
(14:55):
case, and then focus ondeveloping the alliances between
the organizations to really getthat rapid time, you know, and
not have the silos between theorganizations.
Speaker 1 (15:10):
Yeah, the ability to get consensus within a community
and across communities andthat's the one thing too is like
this is no longer exclusive toone organization and one
standards org.
We've got internationalstandards that are aligning and
we've got new patterns ofworkloads right.
It's not just the artifacts butactually the use cases, and
(15:33):
everything old is new again.
Like sands through the arugaswe store the days of our lives.
So one of the things that Iloved is that I've been talking
about sustainability for aboutas long as I've remembered how
to spell sustainability.
It's been something that about.
I think about six years ago itbecame a real hot topic.
(15:53):
It was dripping off the tonguesof every CIO and network.
You know, a week was publishingall these new things about
seeing what the potential wasfor sustainability.
The problem was we didn't havean understanding of what it
really meant and we didn'tnecessarily have the workloads
that really required it.
The initiative was there but wehadn't figured out how to truly
(16:19):
implement it and see thebenefits of it.
But it's a long game and thatlong game also is surviving a
lot of workload changes, so eventhe traditional SAN style
workloads.
You throw new workloads at aSAN and it fundamentally changes
the life cycle of it, the wearpatterns of it.
And so you know, john Michael,if you want to talk about I know
(16:43):
you've done a lot around whatcompanies are trying to achieve
around, you know, net zero goalsand looking at sustainability
as a practice, but you know, Ithink we're actually seeing it
happen.
We're actually seeing realresults.
This is no longer a pie in thesky.
It would be nice if we couldget there.
We are actively seeing thingsand we have to, because that
(17:04):
little thing called AI isprobably ripping up drives as we
speak on a regular basis.
Speaker 4 (17:12):
Yeah, I mean.
The one thing is that a lot ofthese companies put their 2030
net zero goals in place beforethe AI CapEx spend really ramped
up.
Now these hyperscalers arespending over $30 billion a
quarter and basically tellingthe world that, hey, if these
models keep scaling, we're goingto 10x that and obviously this
(17:32):
directly impacts sustainability.
But thankfully, a lot of thetechnologies for sustainable
data centers actually help getbetter power efficiency and
power density in these AI datacenters.
Actually, two of Rochelle'scolleagues who are part of the
OCP sustainability project SammyI guess Mohan was there and
(17:56):
Eric Dahlin they wrote thispaper from OCP on power usage
effectiveness and some of thenew trends in IT around liquid
cooling, cold plate immersioncooling and these technologies
are.
First, they were oh great,they're so good for
sustainability because theyreduce the energy.
But now it's like, oh, we haveto do it because you can't cool
the AI servers without them.
(18:17):
Oh, but by the way, now youreduce the fan power and the
heat and all the IT load andmake the actual data center rack
more efficient.
So it kind of comes full circleand this whole thing.
They're all connected.
Speaker 2 (18:34):
I mean, it's not ironic that all of the top
supercomputers in the world havebeen liquid-cooled forever?
They kind of knew what theywere doing from the start and
they optimized for it.
Speaker 1 (18:46):
Yeah, and there's an interesting thing.
I guess it's almost even a bitof a Jevons paradox, or Jevons
paradox.
You know, this idea thateventually we will move towards.
You know the thing becomes wemake it so expensive to try to
not use it.
That actually becomesefficiency.
It is the way that we've, likeyou said, we just set these
(19:06):
goals for net zero, for, youknow, for 2030 goals, and then,
somewhere between the time ofsetting the goals and 2030, the
entire face of computing changed.
I mean the fact that Sora waslaunched very recently on OpenAI
, as an example, and for threedays of course, all sorts of
(19:29):
exciting errors happening acrossthe whole world of OpenAI's
infrastructure, completelychanging their consumption and
usage patterns.
So even a company that knowswhat's ahead in the very short
term probably can't even preparefor what's about to happen as
we see it in practice.
So you know what is that impactat like cost of operations and
(20:02):
that cost of a life cycle of apiece of hardware, what is the
real cost and what are theactual metrics that drive that
tco over time?
Maybe jm, if you want to, thisis probably your, your backyard
so I'm gonna pin you want to.
this is probably your backyard,so I'm going to pin you down to
try and throw some numbersaround.
Speaker 4 (20:20):
Yeah, so this is a shameless plug.
We have a SNEA TCO model forstorage.
That's very good and, by theway, when I originally created
this at Intel, it was we werethinking about okay, how do we
replace all the hard drives inthe world with SSDs and some of
these workloads where you need acertain amount of IOPS and you
can change your replication andcompression factor?
(20:41):
These are actually really bigTCO drivers.
So, on the first part of like,if you're looking at just
storing data, this TCO metric of, like TCO dollar per terabyte
effective, which is like, afterall these, like you know,
replication, compression and allthis technology is very
(21:01):
important Per rack, per month.
This is kind of how thehyperscalers think about this,
but there's all sorts of otherTCO models for it.
Like you have to look at yourdifferent KPIs and look at the
different types of workloads,what you're optimizing for, but
for actually storing data,that's that's fairly well
understood.
And now it's just about how doyou optimize the infrastructure
to get to a certain you knowwith, with all the certain
(21:23):
constraints, and leave.
You know that whenever I sharethat TCO model, somebody,
somebody says, oh, this thing'snot right, how did you come up
with this price.
I said it's a model like theyou have to put your own inputs
in.
Like you have to put your owninputs in.
It doesn't just do everythingautomatically for you.
You actually have to have someassumptions and there is no one
size fits all TCO model.
Like, every single customerneeds to go through what matters
(21:45):
for us.
But it's like a really goodstarting point.
But there's all kinds of other,obviously, things like now AI.
They're trying to moving fromthis dollar per gigabyte, you
know, just as being one of theseonly metrics to, okay, maybe in
some of these AI workloadsthey're actually IAPs per
terabyte and IAPs per dollar isnow a really important thing,
(22:06):
you know, for how do you createa new storage technology that
might fill the gap between Flashand DRAM or something like that
?
Speaker 2 (22:13):
And that kind of goes back to the original concept
that you really have to buildthe infrastructure because you
have to decide what theimportant metrics are you going
to be measuring.
There's dollars per gigabyte,there's IOPS per watt?
I mean, that's a weird one.
There's a lot of differentmetrics that you want to take a
look at, and the originalthought with the SSD guys was
(22:36):
let's destroy the hard drivebusiness, let's get everything
onto SSDs.
But I say that with a grain ofsalt, because the tape industry
is still hanging on strong.
In fact, a lot of cloudproviders are buying more and
more tape every year because itmakes a good tier level and the
sustainability for tape isinsane, cause once you write to
(22:58):
a tape, it can sit on a shelffor 10 years unpowered, which I
really wouldn't trust NVMe orSSD to do, or even hard drives,
for that matter.
But the point is that all threeof them have tape block or tape
hard drive, and NVMe and storeand SSDs.
(23:18):
They all have their own niches.
Now SSD niche is very largeright now.
It's eating up a lot of thedata center.
But you really have to go forwhat you're measuring.
Speaker 3 (23:29):
And we really have.
You know not to go completelyoff on a tangent, but I'm going
to do it anyway we have the DNAdata storage.
That is, you know, a potentialfuture disruptor.
That would take this wholemodel into a complete other
direction.
Right, you mentioned tape andarchive.
That would take the model into,you know, yet another complete
(23:53):
dimension as that technologymoves forward.
So there's a lot of differentdimensions you could take in
terms of what's valuable andwhat you know.
What.
What, as John John Michaelmentions, this is a model.
Speaker 1 (24:12):
What attributes are important to you.
Went in on that one, Rochelle,you know, since I got you on mic
already.
Let's talk about what thatmeans as far as the sort of
accessibility, programmability,and where Swordfish comes into
play, when we look at new formfactors, new types of storage,
new patterns of consumption.
But where do you see theSwordfish wins happening today?
(24:37):
Because of, you know, the earlywork that was done.
Speaker 3 (24:44):
So, yeah, there's actually quite a few swordfish
implementations out there.
We don't actually have a lotthat are working with our
conformance test program at thispoint, but we do know there are
a lot of actual swordfishimplementations out there.
We've had multipleimplementations over the time.
(25:04):
We're developing a new adopterspage for SNEA in general.
You can look for that coming inthe next few months where
people will start to registertheir overall implementations.
We are actually starting somework.
We expect to see in 2025 someinitial implementations with our
DNA data storage.
We have some early implementersthere that we actually expect
(25:28):
to see some initial modelingfrom some manageability with the
DNA data storage.
We have some expansion comingwith file storage with some file
system implementers.
We are talking with folks inthe object store space as well.
(25:49):
So there's a lot of additionalexpanding on where we have our
file, our base of block storagefile or you know, base of block
storage.
We are also working on expandingwhere we have our I mentioned
(26:11):
earlier the redfish folks workextensively with OCP.
We have some initial baseprofiles, which is how folks
like OCP basically call what Igenerically call them recipes.
Basically, the work that OCPdoes is largely like defining
how to use other people'sstandards and so think of that
(26:35):
in terms of saying here's therecipe, how we want to use, you
know how to use the standard, sowe define those in terms of
profiles.
We expect to expand on that inthe storage space more so,
working with OCP, working withother standards orgs as well, or
other consortiums, to sayhere's the specific recipes.
(26:58):
Moving forward, we've beendoing that quite a bit with the
NVME consortium.
Moving forward, we've beendoing that quite a bit with the
NVME consortium and we expect tosee that as defining more and
more profiles in conjunctionwith other groups, as to kind of
clarify the expected andtargeted usage for the redfish
(27:19):
and swordfish in the storagespace.
Speaker 1 (27:29):
And you know, this is the very important piece where
we talked about, you know,standards as a common
abstraction, because, at thelowest level, when we start to
do measuring the actual cost ofa life cycle of a piece of
hardware, from manufacturingthrough to, you know,
destruction and it's not justoperation, it's you know, you
mentioned this before too jm wasthe idea that, like, what's the
(27:50):
cost of creating these?
I mean, you know, the penny isthe classic that we always look
back to.
It costs more than a penny tomake a penny, and that's why
many countries have decided toabandon the practice, just
because it just didn'tfinancially make sense to
produce this or environmentallymake sense to produce this
anymore.
And so, when you come tomeasuring around sustainability
(28:14):
and lifecycle and cost of energyto drive something, where does
this?
How are you seeing thesemeasurements be?
Like, how do we find them?
Like, who chooses the rightmeasurement of?
Like, yes, this is the actualcost from, you know birth to
completion for this particulardevice.
Speaker 4 (28:34):
Yeah, so you know there is a lot of
standardization in somethingcalled the life cycle assessment
and it's not perfect, but youknow it's supposed to be under
specific.
You know it's under specificconditions and this is one of
the key places where OCP hasbeen working with the iMasons
Climate Accord on trying tostandardize some of these
categories of like okay, ifyou're an SSD or a hard drive,
(28:56):
like you're kind of how youshould report it, but at the end
of the day, like there's, youknow the memory manufacturers
know how much energy is goinginto the fabs.
They have to run thesemulti-billion dollar fabs.
They can do a life cycleassessment of the materials and
the F gases and stuff going intothe air.
And like one of the metricswe're trying to get to is like
(29:16):
this carbon per gigabyte orcarbon you know, you know
embodied carbon per terabyte.
Well, the numbers that we'vecome up with right now for flash
are like astronomically high.
When I talk to traditionalsustainability, folks they're
like, wow, we really have to fixthis.
So and that's where I mentionedthis like the whole focus of
this circularity is to is to.
(29:37):
It makes sense from a likelogical standpoint that if
you're going to use a thing for10 years instead of five years,
that the per year embodiedcarbon or use case of use of
that device goes down if youamortize it over a longer period
of time.
The equivalent would be like Idrove my car for five years and
then we just threw it in theshredder because it was done.
(29:59):
No, nobody does that right,they sell it because there's
economic value.
You know it still has use leftin it.
So you know we're trying to dothe same with with ssds.
Um, the big unlock this yearhas finally been that like the
hyperscalers are really engaged.
Uh, rochelle might be able to.
You know, she's probably wellaware of all the things that are
(30:22):
going on in OCB with security.
The big two things this year areCalyptra, this hardware root of
trust that can enable likeattestation.
This is like open source IP thatcan go into new devices like
SSDs and hard drive controllersto basically establish this
cryptographic authenticationattestation.
To basically establish thiscryptographic authentication
(30:42):
attestation.
But the really exciting newcapability that they put in
there this year is this thingcalled LOCK, this Layered Open
Source Cryptographic KeyManagement Block.
This is an open source IP blockfor key management backed by
Google, microsoft and then threeof the other SSD vendors to
basically ensure that this IP isreally good around key
(31:03):
management so that they cantrust that when they do a
cryptographic erase.
Or if somebody you know, in thecase of Opal we have a
self-encrypted drive Somebodywalks into a data center, puts
the drive in their pocket, walksout.
They want to know that thatdata cannot be decrypted, even
by a nation state actor withhundreds of millions of dollars,
and the only way they can knowfor a fact that that can't be
(31:28):
decrypted is to know how thekeys are managed on that device
with an open source keymanagement block.
So these are just two examplesof like that you know the
hyperscalers are taking thisvery seriously and you know,
devoting new technology and IPinto this space.
Speaker 1 (31:39):
And I think this takes the abstraction up one
step further and maybe, chris,I'll tap you is the idea of
there's local optimization,there's sort of server level,
rack level, data center leveland then energy sustainable data
center level.
So there's all these differentplaces this is happening.
(32:00):
So what are we seeing aroundactual data center level
optimizations happening?
Speaker 2 (32:06):
Well see, that's the neat news is that if you know
how to talk Redfish, you knowhow to talk Swordfish.
It's the same protocol.
It's an extension that givesyou all the neat things that
enterprise storage gives you.
However, what that means is Ican gather metrics from all of
my servers, all my storage soonto be switches.
I should be able to gather thatall from the same API calls the
(32:28):
same kind of mentality.
But, more importantly, I can doa cross vendor.
So I can talk to an HP, I cantalk to a Dell, I can talk to a
Cisco, I can talk to a Lenovo.
Supermicro, it doesn't matter,I can make the same call and
pull the same metrics.
Super micro, it doesn't matter,I can make the same call and
pull the same metrics.
So, just like we were talkingabout embodied costs with carbon
(32:48):
, that kind of thing, we canalso look at things like wattage
and fan speed and incoming airand outgoing air.
We can look at all that stuffcross-vendor with the same code
base.
And that's really good forcustomers, because customers
don't want to rewrite the codeto an SDK for eight different
vendors if they have eightdifferent vendors.
The average customer out thereuses two different storage
vendors or two different servervendors.
So they want a common code base.
And in fact we also talkedabout metrics.
(33:12):
How do we know what metrics torecord?
That's a neat thing aboutRedfish and Swordfish as well,
is that all of the differentpartners that help write
Swordfish, which make up theindustry, we all agree to it.
No one company comes in andsays I want this metric.
There's an OEM section for that.
Redfish has got an OEM section.
Swordfish has got an OEMsection.
They can put a custom counterin there.
(33:34):
But we encourage, if you've gota counter that's going to be
good for the industry, that youbring it forth and let your
competitors see the same counterand decide okay, we want to
subscribe to the same thing.
Once you have multiple peopledoing the same thing, it becomes
a standard counter.
A good example of this was thatwe were going to bring again.
I work for Hewlett PackardEnterprise.
(33:55):
We purchased a company calledCray Cray Supercomputers.
You probably have heard of themand we said we have a mandate
across our company.
Every server we ship is Redfishcompatible period, end of story
.
And we brought Cray in and Craywas not.
So we said well, we got to putRedfish on you guys and they say
, okay, that's great, but welooked through the protocol and
(34:16):
doesn't support water cooling.
So we're like, no problem,let's get water cooling in the
standard.
So we brought it forward and wehad Dell and super micro and
IBM, lenovo check the work tomake sure it's acceptable.
But now water cooling is in thestandard, so it doesn't take
much to expand the standard tocover everything in the industry
.
Um, but it is a cooperativeeffort, uh, and it's definitely
(34:40):
customer focused and and uh,definitely customer-focused and
definitely customer-friendly.
One code base to rule them all,as it were.
Speaker 1 (34:48):
Yeah, when it sure beats the, you know it's
deprecating and moving forwardin an efficient way and an
optimal way.
Just for people operations isso much better versus before way
.
Just for like people operationsis so much better versus before
.
Like I say this is I'd lookover in a drawer probably still
filled with nine pin to ethernetcisco cables, but just because
(35:08):
you never know when you're goingto need to grab that old switch
.
But imagine that you know I hadthe same thing for my storage
gear and my fiber switches.
It was like there was alwayssome proprietary thing on every
single thing and i've've let'sget away from that it's so
fantastic.
Speaker 2 (35:27):
How many times has one of us old-timers gone into a
lab with a fluke analyzer totry and figure out how much
power a server's pulling, butthen we don't really know what
it's going to do in production?
Or we buy a specialized PDUthat's got all the power draw
information so we can monitorthat thing.
Well, every server that's soldin the market today by every
vendor has a sea of sensors.
(35:48):
They all have a lot of sensorsthat give us a valuable wealth
of information.
If you want to know where ahotspot is in the data center,
that's easy to find.
If you have a situation where acertain model of a certain
vendor's hard drive got recalledbecause there was a bad batch,
you could write one command thatqueries everything in your data
(36:10):
center from different vendorsand finds where that drive
happens to show up.
So you know what you've got toreplace.
There's a lot of cases wherethis standardized approach
really really pays out for thecustomer.
Speaker 1 (36:26):
When it goes up to the software, even to the point
of placement right, like if wethink, uh, I saw some of the
stuff and even a couple of yearsago with like OCP was talking
about, and like Facebook doesthis with their data centers and
they actually move workloadsaccording to heat maps, because
they'll see, like a hot pocketbuilding up and you know that
(36:48):
they just literally can redeployworkloads and distribute them
throughout because of thephysical or whatever it's going
to be, can literally affect theefficiency of your entire data
center, because if it's put intoa spot where there's too much
(37:09):
other than required, it'sincredible.
And that same instrumentationthat tells you what the
temperature is also tells youwhere to place, because you can
choose.
That's the beauty ofstandardization.
Yep, exactly.
Speaker 2 (37:24):
and when you're talking about cooling.
Cooling you turn cooling up ordown like a sledgehammer it's.
You're going to cool 50 racksat a time, either by turning the
cooling up or turning thecooling down.
If you can make sure there areno hot spots across an entire
set of racks let's say 150, 200racks and you can make sure to
eliminate the hot spots, you canrun the entire set of racks
(37:46):
hotter than you normally would.
Generally people run an entireset of racks colder because they
have to make sure that thehottest server in the hottest
rack doesn't go over themandated number.
But if you can prevent thatserver from existing or prevent
that server from ever gettingthat hot, that means you can run
everything hotter.
You can run less airconditioning, you can run
(38:07):
everything closer to the edgeand that's a massive power
savings, because every watt ofCPU or computer IO you put in,
you're also putting a lot ofheat in.
So I mean that's half theequation right there.
Speaker 4 (38:22):
I'll certainly echo the excitement for Redfish being
just standardized.
You know, one of the things thatwe're looking at in the
sustainability space I mentionedis this classification of IT
and non-IT power.
Right, the fans are notcontributing to the useful work
and being able to basicallyeasily enumerate that in Redfish
to say this is the fan power,this is the rest of the system
(38:43):
power.
And you know for, like you know,we work on SSDs having the
ability to, within that server,say, okay, this is how much
power is going to all thesedifferent areas, and then I can
optimize things.
That is extremely helpful,right, because a lot of one of
the things we talked aboutearlier this year at FMS was,
you know, one of the reallybasic thing of just using NVMe
(39:06):
power states to take a drivefrom a 20 or 25 watt TDP down to
a 14 or 16 watt TDP, you canjust say that now you will not
consume more than that power.
Well, that tiny change can havea massive ripple effect on the
fan speed and the overall TCOand if you don't have that type
of Redfish instrumentation, it'svery hard to like actually say
I made this power change in mydrive.
(39:27):
I can do some basic math.
You know, 20 watts to 16 wattsshould save four watts per drive
.
Turns out it's actually a lotmore than that, because you get
the fan power too, and so,knowing that kind of stuff, to
be able to instrument withRedfish.
Speaker 2 (39:43):
And if you can tie the effect in with the value
you're saving, maybe it makessense.
But if you have a bad effectfrom lowering the power but you
can't show the value of thesavings you're getting, you may
never want to do that yeah, andso this is a great example of an
area where we've got a lot ofthe base instrumentation.
Speaker 3 (39:58):
We've got all that base instrumentation in from the
power.
We put the base instrumentationin but we have not and we we've
.
This is an area where we can goback in and put and look and
say when and where should we putthe controls in for those NVMe
power states?
Right, we did the basics in theSwordfish model and this is
where we differentiate betweenRedfish and Swordfish.
(40:18):
Right, redfish is where we putthe base power controls for the
chassis.
Swordfish is where we'll go putthat additional instrumentation
for those NVMe SSDs and wherewe'll look at both the controls
for the NVMe SSDs for thosepower states but also the
(40:40):
security around who gets tocontrol them, because you don't
want to expose those up too much, and so that's an area where we
will look at partnering inagain, putting those profiles,
those recipes in on where do weexpose those up, who should have
control over them?
And working with folks likeJohn Michael and Circularity and
(41:03):
NVME and getting those alliancepartners in to say who are the
folks with the expertise thereand where should we put those
partnerships in place so that wecan accelerate.
Maybe look at that in 2025 here, as we're two weeks away from
(41:23):
going into 2025 as we're doingthis recording today.
But get that as we go into 2025, can we make this a focus?
Get that functionality andthose controls into the
Swordfish specification and getthat base enablement in so we
can get those intoimplementations here in the next
(41:49):
year or two.
Speaker 1 (41:51):
And I think you hit one of the top things and you
know, I wish we could do thisall day long.
I would love to.
I only wish we had another hourto commit people to.
This is so fantastic, but it'sfar beyond just the specifics of
optimization and even when wetalk about sustainability it
from manufacturing to placement,to life cycle, to everything
(42:13):
security we often get mentioned.
People always say you know, whydid we have to?
You know, create this thingcalled dev sec ops?
You know, security is impliedlike oh no, the.
The presence of it was assumed,not implied.
You know, but in this community, what you've got now is the
same practitioner.
You know practitioners areactually then crossing the
(42:35):
working groups and so you cantalk about this and then you've
got security.
People can come in and seewhat's happening and they can
participate in this conversationand they can understand what
the impact is as they look at.
You know, protecting the apAPIs, protecting all the
endpoints, and it's this is thespot where it's people that have
(42:57):
a common goal in their ownindividual section, but the
really it's it's how do weoptimize the community and the
community of people that we meet, community and the community of
people that we meet and maybeI'll just ask each really
quickly say like, what's the oneaccidental thing you found
yourself working in.
It may be crossing workinggroups or just meeting somebody
(43:18):
in the community that yourealize, oh wow, this just saved
me a ton of time, or this was adiscovery that actually helped
you in your primary area,because you were beside somebody
that you wouldn't haveotherwise met had they not been
participating in this in thiscommunity.
Speaker 2 (43:35):
Maybe, chris, I'll start with you- um, it's funny
because I I end up meetingsomebody that does something
drastically different.
That is very interesting.
At every STC I go to I mean themost recent STC, I got to spend
some time with DNA storage guysand got some real deep skinny
(43:57):
on that stuff.
And the time before that it wasanother topic.
The time before that it wasanother.
I find that, unfortunately,there are too many technologies
to get into and I'm interestedin all of them, um, but I
wouldn't have access to thatunless I had gone to an event
(44:18):
where this cross collaborationhappens, and that's really the
magic of it is is meeting peoplewhere you get to cross
collaborate.
I'm looking forward to um.
I know they're working on aautomotive task force right now
to figure out if they can startworking with the people from
that.
Do the the networks for cars?
(44:38):
Um, I'm starting to researchthat.
That looks like really funstuff, um, but there's so much
to choose from there really isyeah, I mean, yeah, he's right.
Speaker 3 (44:50):
I mean, with there's, there's something that pops up
all the time.
It's like I, you know I'll comeback from a show and go.
Yeah, we just went on thisrandom.
Yeah, this one was a random.
Yeah, uh, uh, we dug into this.
Speaker 1 (45:07):
It's like a new journey every time you go there.
I I love it.
Speaker 3 (45:11):
A couple of years ago we were all off on a.
Hey, we're doing storage inspace.
Speaker 1 (45:17):
Yeah, yeah.
Speaker 3 (45:20):
Then there was the.
Here we're doing all of the.
We had the global mapping of.
Here's the story we're puttingstorage in planes, they're going
at high altitudes and doingreally GPS mapping of in really,
(45:46):
really granular detail, of ofof the earth, basically, um, and
so it was like all of this, thestorage aspects of it were were
doing really really high umgranular, and you had to make
sure that all the storage um wasum.
(46:09):
It was like really um.
I'm sorry I'm not describing itvery well.
It was just incrediblyfascinating.
But it was all like sub-zerotemperatures, that it was all
like immediately purged onlanding and different
(46:30):
requirements.
Yeah, really really really HA,and it was incredibly
fascinating.
And I forget almost all of thedetails of it because by the
time we got through, oh my gosh,this is fascinating.
And then I completely forgetabout it and go on to the next
topic.
Speaker 1 (46:42):
So we just solved the New Jersey drone problem right
now.
Is that what this is?
Speaker 2 (46:49):
And sometimes you get inspiration from the weirdest
places, like DreamWorks whenthey came in.
Speaker 3 (46:55):
How they're dead people usually fascinating while
we were working on it, and thenI completely forget all about
it and go to the next one.
Speaker 2 (47:04):
Yeah, which is why you should definitely take trip
notes when you go to theseevents.
Pull out that paper and writenotes through the event, because
you will forget day one by daythree and eventually videos are
posted online.
But you want to have your notesfrom those videos as well, and
if you share that with people atyour own company, not only will
(47:25):
they be more likely to send youthe next year, but you may find
that other people in yourcompany find the information
very useful for their jobs.
Speaker 3 (47:34):
I can't.
I can't remember the details ofall of them, because there's
just been so many of them.
Speaker 1 (47:38):
Yeah.
Speaker 3 (47:40):
Yeah.
Speaker 1 (47:41):
It is truly a wealth of humans and a wealth of
knowledge and such a sharingculture, which this is the
breakdown of.
There is no corporate wall.
When we're there, we make apoint of kind of avoiding our
associations, but we also areproud to say where we're from
and no one questions it.
(48:01):
It's just because we are thenew artifacts, the human
artifact, that can go throughand move from place to place and
achieve these amazing thingsand then again seeing it
actually play out.
Now we're actually seeing thereal results with some of the
stuff in sustainability, and itmay have been a marketing
buzzword six years ago, but whenthat went away, the work didn't
(48:24):
and we kept going and we keptachieving innovation.
So maybe just as a last close,jam M, what's got you excited
and what else in your SNEAconnections has been a wonder
that you probably didn't expect.
Speaker 4 (48:39):
Well, it's funny, I presented at SNEA STC this year
on sustainability, so that wasmy idea of like let's try to get
people that aren't usuallyexposed to the sustainability
world to start caring about itmore.
You know, it's slowly peopleare.
We're making them learn aboutcircularity and whatnot.
But, of course, like I get togo attend these presentations on
storage and AI, which is whatI'm spending most of my day job
(49:01):
learning about right now onmanageability and other stuff
that helps me with some of theside projects for GPUs and AI
workloads and all kinds of otherstuff I'm in.
So, yeah, it's.
It's just, I guess, what wejust said some of the uh, I
brought one of our like juniorsales guys to stc this year just
to like stop by and say hi andhe was like wow, are there
usually this many like seniorpeople from all these companies?
(49:21):
They're like.
He's like every single person Imeet is like director, senior
director, you know some kind oflike fellow or something like
that.
I was like, yes, that's welcometo stand.
Speaker 1 (49:32):
It's a, it's a beautiful thing and, uh well, I
can say I've been lucky as a.
My greatest gift is meeting allof you and being able to share
time and learn on a constantbasis.
And you know, it's, it's.
This is the.
The very reason why I love youknow, because we talked about
SaaS and space, we talked aboutsustainability.
I literally get to sit in onall these conversations and,
(49:54):
having come, you know, I'm inthe space myself as well.
So it's funny, not space, butthe space in the industry, I
guess, and it really isenriching, you know, for me,
because then, as I go to thenext company and I'm advising
companies and startups andeverybody, I say, like you want
to meet people who are drivingchange and who are your future.
(50:17):
You know, collaborators, thisis the group to be in.
So there's lots of virtualevents.
There's lots of in-personevents.
We've had a really, reallygreat change with the membership
structure for 2025.
So there's a podcast where weshare about some of the updates.
I talked with Jay Metz aboutthat.
So definitely recommend peoplego check out the podcast, Go
check out the SNEA YouTubechannel.
(50:38):
There is no shortage offantastic content.
But, most importantly, getsigned up, come to an event.
There's virtual events, thereare in-person events and, oh,
thank goodness, we're back inperson.
It's so great to be able to bein the room, because the hallway
track is one of my favoriteparts of every conference
there's a lot of hallwayconversations.
Speaker 2 (50:59):
That is so true.
Speaker 1 (51:01):
And I can tell you that I've literally met who
would become future clients andfriends and employers through
these hallways.
And so thank you all forletting me in the hallway and
thank you all for the folks thatare listening and watching.
Do continue, so keep your eyesout.
2025 is going to bring a lotmore great content like this.
(51:22):
So, with that, chris RochelleJM, thank you so much.
And again, folks, don't forgetto sign up to the SNEA Experts
on Data podcast.
Hit, subscribe, smash that likebutton, do all those things
that the Zoomers tell us we'resupposed to do, but, most
importantly, get to a SNEA event, find out how this community
can help you and what you cancontribute back.
(51:44):
And thank you all for watchingand listening.
Speaker 4 (51:49):
Thanks all, Bye-bye.
All right, everybody.
Welcome to the SNEA Experts onData podcast.
My name is Eric Wright, I'm thehost here of the SNEA EOD and
I'm also the co-founder of GTMDelta Super excited to have an
amazing crew here today with us.
We're going to talk aboutoptimization.
This is a place that I've beenliving in a long time, both in
(00:23):
the work that I've done and heckon a personal day-to-day basis.
So if I could tell you how tooptimize things, I'd love to.
But even better, I've got threeof the most amazing people who
are going to tell us about whatthe whole SNEA community is
doing around our data focusareas.
Today we're going to talk aboutOptimize.
With that, I'm going to do aquick roundtable, do a quick
(00:45):
introduction.
So we'll start off with ChrisDo you want to lead us out?
And then just a quick littlebio on you, and then we'll jump
to Rochelle and then JM.
Speaker 2 (00:54):
Sure, my name is Chris Leonetti.
I've been a reference architectfor various companies over the
years for probably what the past?
Oh my gosh 30 years now.
So I've implemented very largesolutions all the way down to
very small solutions and largeI'm talking $3 billion kind of
solutions.
(01:14):
But then I've also architecteda lot of solutions that you know
are in the $50,000 range tosolve business needs.
Speaker 3 (01:25):
So Chris has been doing that since he was a
toddler, clearly, but moreimportantly, he is the secretary
of these things.
Correct Secretary of thesethings, yeah, and wears many
hats around our organization.
So I'm Rochelle Alvers.
I am the vice chair of SNA aswell as the chair of the
(01:50):
Swordfish TWIG.
You guys are hearing it herefirst, we just renamed the SSM
TWIG which nobody ever knew whatthat acronym stood for to the
Swordfish TWIG, so that you knowwhat we actually do.
And I'm also the chair of thestorage management community,
which helps promote theeducation, promotion of
(02:14):
everything related to storagemanagement within SNEA, and so
that's actually one of thethings we'll be talking about
here in a little bit.
Oh, sorry, I forgot my day job.
Day job I lead the technologyinitiatives and ecosystem
enabling for Intel, and I'vealso been working on things
related to everything forstorage management and
(02:39):
technology initiatives since Iwas a toddler.
All right, john, I'm JohnMichael.
Speaker 4 (02:43):
Hans, I'm Senior Director of Product Planning at
B a toddler.
All right, john, I'm JohnMichael Hands.
I'm Senior Director of ProductPlanning at FADU, so right now
we're working on enterprise SSDcontrollers for hyperscale.
That's the day job.
I also co-chair the SNEA SSDSIG, so we wrote stuff, like you
know the form factors page onEVSF, the SNEA TCO model, which
is also used by lots ofhyperscalers, and other fun
(03:05):
stuff.
I also am the secretary of a501C nonprofit called Circular
Drive Initiative, where we arefocused on sustainability of
storage, and then very activelyengaged in the Open Compute
Project, sustainabilityInitiative.
So I co-chair one of the workgroups there.
Speaker 1 (03:22):
Everybody's side.
Jobs are like full-time jobs.
It's amazing that you can keepall this on track, but I've seen
you all in action so I know howyou pull it off.
You're a bunch of fantasticfolks.
So, Chris, if you want to,let's just sort of do an initial
intro to people about what itis that we talk about when we
(03:43):
mean the optimize data focusarea and what this is under the
SNEA sort of new description aswe talk about these data focus
areas.
Speaker 2 (03:53):
Yep.
So when people say the wordstorage, they're kind of like
asking how long is a piece ofstring.
There are so many answers towhat storage really means.
A good example is storagemanagement is something that's
desperately needed out there.
That's highly valuable, but itgoes down different paths than
(04:21):
storage environmentals, so like,for instance, the green twig.
So we have different focusareas for storage that deal with
the different aspects of whatstorage does and what storage is
.
So, for instance, we have onegroup that focuses on form
factor.
We have another group thatfocuses on manageability.
We have a third group thatfocuses on efficiency when it
comes to power, that kind ofthing.
(04:42):
The hives of SNEA are basicallyset up to explore all those
different avenues that storagecan manifest itself in.
In fact, rochelle can probablygo into more detail about how
these hives are laid out.
Speaker 3 (05:00):
Yeah, we have six different data focus areas that
we've defined.
Yeah, we have six differentdata focus areas that we've
defined as we've been expandingSNEA over the years.
Snea started as the StorageNetworking Industry Association,
very focused on SANs.
Over time, we expanded to allthings storage, and what we've
been expanding to over the lastfew years is really a
(05:21):
recognition, and the reason wecame up with these data focus
areas is really recognition thatwe're beyond just storage.
We're really looking at datafrom the perspective of storage.
So when we're talking about thisparticular data focus area,
which is the optimizinginfrastructure for data, it's,
it really is.
How are we looking at, you know, all aspects of data and
(05:46):
optimizing the infrastructurefor data?
It is from the perspective ofstorage, but it's really broader
than just the storage elementsin the system.
Everything we're doing withinSNEA is really beyond just the
storage of them.
(06:06):
It's the entire data, theentire ecosystem managing data,
including everything frommanaging data accelerators,
managing this storage fabrics,managing the storage elements.
But it's definitely a broaderpicture than it was, you know,
five or 10 years ago.
Speaker 2 (06:28):
And an example of that would be someplace that the
line blurs between storage anddata.
In fact, you can't reallyoptimize the storage
infrastructure without knowingthe data that's going to be on
it.
And what I mean by that is,let's say, you've got a dataset
that's highly compressible orhighly dedupeable.
You might optimize differentstorage for that dataset than
(06:49):
you would for a dataset that hasno compressibility and no
duplication ability.
You also have to be able tochoose between object files and
block.
So there's a lot ofoptimizations that actually
matter more about what the datais than simply a place to store
it.
You can't simply look atstorage as a box you put on a
wall, that you put things in.
You have to really look atstorage as the optimal way of
(07:14):
putting these boxes together insuch a way to solve the needs or
the requirements of the dataengineer.
The data engineer is going tosay I need this kind of data
stored in this way, with thisIOPS, with this envelope, with
this reliability.
There's a lot of features youhave to build into that design,
into that solution that reallylets you optimize.
So the lines between pure dataand pure infrastructure are so
(07:38):
blurry that they kind of don'texist anymore.
Speaker 1 (07:40):
Well, and even optimize for purpose is even.
You know we can talk aboutoptimize for performance,
optimize for capacity and JM.
You know.
Let's maybe talk for a secondabout optimize for
sustainability, because I knowthat's one area that you're
bringing a lot of focus to anddoing a lot of work on.
How does this play into thesustainability side of of you
(08:03):
know the world and obviouslysnea?
Speaker 4 (08:06):
yeah, uh, we had a really good presentation at snea
, stc this, uh, I guess thisjust a couple months ago on, uh,
what sustainability means tostorage and really the impact
you know for the SNEA projectsreally, which I focused around
circularity, and you know, mediasanitization, which are two
really key aspects.
But when I think, when peoplehear sustainability in the
(08:29):
storage world, they usuallythink about energy efficiency
and that is one really importantaspect of sustainability, right
, it it's like you know, this ishow much energy you're using
during the use phase, when thedevice is powered on.
But I think people forgetreally the other important part,
which is what we call itembodied carbon, which is how
much did it take to manufacturethat device and get it from
(08:50):
point a to point b before iteven gets to the service?
How much does it take todecommission and end of life,
this full life cycle of thedevice?
So Flash, specifically like SSDsand memory, you know, can take
a tremendous amount ofelectricity and resources to
manufacture.
So one of the what ourpresentation was about at STC
was the idea of circularity,which you could basically
(09:12):
sanitize the device, remove allthe user data and then have it
get a second use in another usecase and that might be an
emerging use case somewhere thathas different performance TCO
requirements.
You know SSDs.
You can look at eBay right now.
You can see drives that Iworked on at Intel in 2011.
They're still up there, they'restill working.
So everybody kind ofintuitively knows this that the
(09:33):
devices can last a lot longerthan a five-year warranty, and
so the main barrier is that hasbeen actually being able to
trust that, when you remove thedata from the drive, that you
can reuse it.
So that was the main topic ofour presentation at CISDC.
Speaker 2 (09:46):
And that's a good example of where the hives cross
over too, because you've got agroup like the energy efficiency
side of the house that wants tobe able to reuse these drives.
You've also got the securityside of the house that wants to
make sure that no data leaksfrom your infrastructure.
So you've got to play thoseagainst each other, and at-rest
encryption is one way to solvethat problem, because you can do
a digital shred, but that maynot be good enough for some
(10:07):
businesses.
It may be for others.
So I mean the different hives.
Although we've got theseseparate hives that handle all
the technology, sne as a wholeis all about collaboration
between the hives as well.
Speaker 3 (10:20):
And the place we bring that all together is in
the standardization of themanageability and the
infrastructure where you canexpose in a standardized way all
of those use cases to say, Ihave the instrumentation where I
can report and control andreally prove that all of those
(10:40):
use cases are true, so I canensure that the security is in
place and comply to any you know, any government regulations you
know, and ensure that, yes, Ihave the security in place and I
am enforcing it.
But I'm also reporting back andproviding the metrics to show
(11:04):
that reporting on and theinstrumentation are in place to
say, yes, that all of thosesustainability metrics are in
place, to show that, yes, I amactually getting the green
aspects, the powerinstrumentation and that all of
those things are actuallytrending to the exact reporting
(11:28):
that we are anticipating.
Speaker 1 (11:31):
When we have this interesting thing too of what's
really important about the SNEAas a community and as a
standards body, what we look at,why these things are important,
because the artifacts willcontinuously change, but the way
in which we access and managethose artifacts is where we
(11:51):
standardize.
This is the reason why we knowHTTP has been a standard for so
long, and CRUD and simple thingsin regular development.
But that goes all the way downto physical hardware and people
often lose sight of that.
And I remember seeing the earlywork with, watching what was
going on when Swordfish was, youknow in its earlier days, and I
thought, well, this ismagnificent.
(12:12):
The trick was always gettingeverybody to come on board.
But once that happens, like thepace of innovation was so much
more rapid because we knew everysoftware abstraction layer
could use the same common youknow ways, semantic architecture
, in order to access thesehardware layers.
And then now that meant theinnovation that happens down
(12:34):
there becomes you know, same wayas we look at at Tesla's with
you know full, you know FSD andstuff, where it's now a software
update, a firmware update, toaccess the hardware.
That hasn't changed but we canbetter use the hardware and this
is happening in in storage andwe.
It seems wondrous andmagnificent in the future, but
(12:55):
it's.
It's actually.
It's.
It's been the future for a longtime.
We've been living the futurefor quite a while.
Speaker 3 (13:00):
Yes, exactly, and so so I want to go back to you know
, talking about accelerating howthis works to something that
John Michael said when he wasintroducing himself, about you
know all the different hats he'swearing about working with
circular drive initiative, aboutworking in OCP.
This work is not exclusivelyhappening in the Swordfish twig.
Swordfish is built on top ofRedfish.
(13:21):
A lot of this instrumentationis actually common to Redfish.
Redfish works heavily with OCPon the base instrumentation.
So a lot of these metrics thatwe're developing are actually
common metrics to Redfish thatare actually common
instrumentation to profiles thatare being developed in OCP for
servers.
So we have a lot of leveragebetween servers, between OCP,
(13:45):
between CNA, so the profile andthe instrumentation are actually
common between about four orfive different organizations.
So, as the Circular DriveInitiative says, hey, this is
how we want the profiles to bedeveloped and the metrics,
there's actually very littlelift we need to do to say this
(14:07):
is what we want for thisspecific storage profile.
It's really okay.
I have a couple of these samemetrics that we have on that
server instrumentation that wejust need to apply over here and
we need a new profile and it'sincredibly common
instrumentation.
It's not a big lift.
It's a very small lift thatthen needs to be developed for
(14:30):
this representation.
So it's really that, how to?
What is that specific use casethat we can then apply?
It's not a huge lift and sothat's how we can actually do
rapid instrumentation, rapiddevelopment on that specific use
(14:55):
case, and then focus ondeveloping the alliances between
the organizations to really getthat rapid time, you know, and
not have the silos between theorganizations.
Speaker 1 (15:10):
Yeah, the ability to get consensus within a community
and across communities andthat's the one thing too is like
this is no longer exclusive toone organization and one
standards org.
We've got internationalstandards that are aligning and
we've got new patterns ofworkloads right.
It's not just the artifacts butactually the use cases, and
(15:33):
everything old is new again.
Like sands through the arugaswe store the days of our lives.
So one of the things that Iloved is that I've been talking
about sustainability for aboutas long as I've remembered how
to spell sustainability.
It's been something that about.
I think about six years ago itbecame a real hot topic.
(15:53):
It was dripping off the tonguesof every CIO and network.
You know, a week was publishingall these new things about
seeing what the potential wasfor sustainability.
The problem was we didn't havean understanding of what it
really meant and we didn'tnecessarily have the workloads
that really required it.
The initiative was there but wehadn't figured out how to truly
(16:19):
implement it and see thebenefits of it.
But it's a long game and thatlong game also is surviving a
lot of workload changes, so eventhe traditional SAN style
workloads.
You throw new workloads at aSAN and it fundamentally changes
the life cycle of it, the wearpatterns of it.
And so you know, john Michael,if you want to talk about I know
(16:43):
you've done a lot around whatcompanies are trying to achieve
around, you know, net zero goalsand looking at sustainability
as a practice, but you know, Ithink we're actually seeing it
happen.
We're actually seeing realresults.
This is no longer a pie in thesky.
It would be nice if we couldget there.
We are actively seeing thingsand we have to, because that
(17:04):
little thing called AI isprobably ripping up drives as we
speak on a regular basis.
Speaker 4 (17:12):
Yeah, I mean.
The one thing is that a lot ofthese companies put their 2030
net zero goals in place beforethe AI CapEx spend really ramped
up.
Now these hyperscalers arespending over $30 billion a
quarter and basically tellingthe world that, hey, if these
models keep scaling, we're goingto 10x that and obviously this
(17:32):
directly impacts sustainability.
But thankfully, a lot of thetechnologies for sustainable
data centers actually help getbetter power efficiency and
power density in these AI datacenters.
Actually, two of Rochelle'scolleagues who are part of the
OCP sustainability project SammyI guess Mohan was there and
(17:56):
Eric Dahlin they wrote thispaper from OCP on power usage
effectiveness and some of thenew trends in IT around liquid
cooling, cold plate immersioncooling and these technologies
are.
First, they were oh great,they're so good for
sustainability because theyreduce the energy.
But now it's like, oh, we haveto do it because you can't cool
the AI servers without them.
(18:17):
Oh, but by the way, now youreduce the fan power and the
heat and all the IT load andmake the actual data center rack
more efficient.
So it kind of comes full circleand this whole thing.
They're all connected.
Speaker 2 (18:34):
I mean, it's not ironic that all of the top
supercomputers in the world havebeen liquid-cooled forever?
They kind of knew what theywere doing from the start and
they optimized for it.
Speaker 1 (18:46):
Yeah, and there's an interesting thing.
I guess it's almost even a bitof a Jevons paradox, or Jevons
paradox.
You know, this idea thateventually we will move towards.
You know the thing becomes wemake it so expensive to try to
not use it.
That actually becomesefficiency.
It is the way that we've, likeyou said, we just set these
(19:06):
goals for net zero, for, youknow, for 2030 goals, and then,
somewhere between the time ofsetting the goals and 2030, the
entire face of computing changed.
I mean the fact that Sora waslaunched very recently on OpenAI
, as an example, and for threedays of course, all sorts of
(19:29):
exciting errors happening acrossthe whole world of OpenAI's
infrastructure, completelychanging their consumption and
usage patterns.
So even a company that knowswhat's ahead in the very short
term probably can't even preparefor what's about to happen as
we see it in practice.
So you know what is that impactat like cost of operations and
(20:02):
that cost of a life cycle of apiece of hardware, what is the
real cost and what are theactual metrics that drive that
tco over time?
Maybe jm, if you want to, thisis probably your, your backyard
so I'm gonna pin you want to.
this is probably your backyard,so I'm going to pin you down to
try and throw some numbersaround.
Speaker 4 (20:20):
Yeah, so this is a shameless plug.
We have a SNEA TCO model forstorage.
That's very good and, by theway, when I originally created
this at Intel, it was we werethinking about okay, how do we
replace all the hard drives inthe world with SSDs and some of
these workloads where you need acertain amount of IOPS and you
can change your replication andcompression factor?
(20:41):
These are actually really bigTCO drivers.
So, on the first part of like,if you're looking at just
storing data, this TCO metric of, like TCO dollar per terabyte
effective, which is like, afterall these, like you know,
replication, compression and allthis technology is very
(21:01):
important Per rack, per month.
This is kind of how thehyperscalers think about this,
but there's all sorts of otherTCO models for it.
Like you have to look at yourdifferent KPIs and look at the
different types of workloads,what you're optimizing for, but
for actually storing data,that's that's fairly well
understood.
And now it's just about how doyou optimize the infrastructure
to get to a certain you knowwith, with all the certain
(21:23):
constraints, and leave.
You know that whenever I sharethat TCO model, somebody,
somebody says, oh, this thing'snot right, how did you come up
with this price.
I said it's a model like theyou have to put your own inputs
in.
Like you have to put your owninputs in.
It doesn't just do everythingautomatically for you.
You actually have to have someassumptions and there is no one
size fits all TCO model.
Like, every single customerneeds to go through what matters
(21:45):
for us.
But it's like a really goodstarting point.
But there's all kinds of other,obviously, things like now AI.
They're trying to moving fromthis dollar per gigabyte, you
know, just as being one of theseonly metrics to, okay, maybe in
some of these AI workloadsthey're actually IAPs per
terabyte and IAPs per dollar isnow a really important thing,
(22:06):
you know, for how do you createa new storage technology that
might fill the gap between Flashand DRAM or something like that
?
Speaker 2 (22:13):
And that kind of goes back to the original concept
that you really have to buildthe infrastructure because you
have to decide what theimportant metrics are you going
to be measuring.
There's dollars per gigabyte,there's IOPS per watt?
I mean, that's a weird one.
There's a lot of differentmetrics that you want to take a
look at, and the originalthought with the SSD guys was
(22:36):
let's destroy the hard drivebusiness, let's get everything
onto SSDs.
But I say that with a grain ofsalt, because the tape industry
is still hanging on strong.
In fact, a lot of cloudproviders are buying more and
more tape every year because itmakes a good tier level and the
sustainability for tape isinsane, cause once you write to
(22:58):
a tape, it can sit on a shelffor 10 years unpowered, which I
really wouldn't trust NVMe orSSD to do, or even hard drives,
for that matter.
But the point is that all threeof them have tape block or tape
hard drive, and NVMe and storeand SSDs.
(23:18):
They all have their own niches.
Now SSD niche is very largeright now.
It's eating up a lot of thedata center.
But you really have to go forwhat you're measuring.
Speaker 3 (23:29):
And we really have.
You know not to go completelyoff on a tangent, but I'm going
to do it anyway we have the DNAdata storage.
That is, you know, a potentialfuture disruptor.
That would take this wholemodel into a complete other
direction.
Right, you mentioned tape andarchive.
That would take the model into,you know, yet another complete
(23:53):
dimension as that technologymoves forward.
So there's a lot of differentdimensions you could take in
terms of what's valuable andwhat you know.
What.
What, as John John Michaelmentions, this is a model.
Speaker 1 (24:12):
What attributes are important to you.
Went in on that one, Rochelle,you know, since I got you on mic
already.
Let's talk about what thatmeans as far as the sort of
accessibility, programmability,and where Swordfish comes into
play, when we look at new formfactors, new types of storage,
new patterns of consumption.
But where do you see theSwordfish wins happening today?
(24:37):
Because of, you know, the earlywork that was done.
Speaker 3 (24:44):
So, yeah, there's actually quite a few swordfish
implementations out there.
We don't actually have a lotthat are working with our
conformance test program at thispoint, but we do know there are
a lot of actual swordfishimplementations out there.
We've had multipleimplementations over the time.
(25:04):
We're developing a new adopterspage for SNEA in general.
You can look for that coming inthe next few months where
people will start to registertheir overall implementations.
We are actually starting somework.
We expect to see in 2025 someinitial implementations with our
DNA data storage.
We have some early implementersthere that we actually expect
(25:28):
to see some initial modelingfrom some manageability with the
DNA data storage.
We have some expansion comingwith file storage with some file
system implementers.
We are talking with folks inthe object store space as well.
(25:49):
So there's a lot of additionalexpanding on where we have our
file, our base of block storagefile or you know, base of block
storage.
We are also working on expandingwhere we have our I mentioned
(26:11):
earlier the redfish folks workextensively with OCP.
We have some initial baseprofiles, which is how folks
like OCP basically call what Igenerically call them recipes.
Basically, the work that OCPdoes is largely like defining
how to use other people'sstandards and so think of that
(26:35):
in terms of saying here's therecipe, how we want to use, you
know how to use the standard, sowe define those in terms of
profiles.
We expect to expand on that inthe storage space more so,
working with OCP, working withother standards orgs as well, or
other consortiums, to sayhere's the specific recipes.
(26:58):
Moving forward, we've beendoing that quite a bit with the
NVME consortium.
Moving forward, we've beendoing that quite a bit with the
NVME consortium and we expect tosee that as defining more and
more profiles in conjunctionwith other groups, as to kind of
clarify the expected andtargeted usage for the redfish
(27:19):
and swordfish in the storagespace.
Speaker 1 (27:29):
And you know, this is the very important piece where
we talked about, you know,standards as a common
abstraction, because, at thelowest level, when we start to
do measuring the actual cost ofa life cycle of a piece of
hardware, from manufacturingthrough to, you know,
destruction and it's not justoperation, it's you know, you
mentioned this before too jm wasthe idea that, like, what's the
(27:50):
cost of creating these?
I mean, you know, the penny isthe classic that we always look
back to.
It costs more than a penny tomake a penny, and that's why
many countries have decided toabandon the practice, just
because it just didn'tfinancially make sense to
produce this or environmentallymake sense to produce this
anymore.
And so, when you come tomeasuring around sustainability
(28:14):
and lifecycle and cost of energyto drive something, where does
this?
How are you seeing thesemeasurements be?
Like, how do we find them?
Like, who chooses the rightmeasurement of?
Like, yes, this is the actualcost from, you know birth to
completion for this particulardevice.
Speaker 4 (28:34):
Yeah, so you know there is a lot of
standardization in somethingcalled the life cycle assessment
and it's not perfect, but youknow it's supposed to be under
specific.
You know it's under specificconditions and this is one of
the key places where OCP hasbeen working with the iMasons
Climate Accord on trying tostandardize some of these
categories of like okay, ifyou're an SSD or a hard drive,
(28:56):
like you're kind of how youshould report it, but at the end
of the day, like there's, youknow the memory manufacturers
know how much energy is goinginto the fabs.
They have to run thesemulti-billion dollar fabs.
They can do a life cycleassessment of the materials and
the F gases and stuff going intothe air.
And like one of the metricswe're trying to get to is like
(29:16):
this carbon per gigabyte orcarbon you know, you know
embodied carbon per terabyte.
Well, the numbers that we'vecome up with right now for flash
are like astronomically high.
When I talk to traditionalsustainability, folks they're
like, wow, we really have to fixthis.
So and that's where I mentionedthis like the whole focus of
this circularity is to is to.
(29:37):
It makes sense from a likelogical standpoint that if
you're going to use a thing for10 years instead of five years,
that the per year embodiedcarbon or use case of use of
that device goes down if youamortize it over a longer period
of time.
The equivalent would be like Idrove my car for five years and
then we just threw it in theshredder because it was done.
(29:59):
No, nobody does that right,they sell it because there's
economic value.
You know it still has use leftin it.
So you know we're trying to dothe same with with ssds.
Um, the big unlock this yearhas finally been that like the
hyperscalers are really engaged.
Uh, rochelle might be able to.
You know, she's probably wellaware of all the things that are
(30:22):
going on in OCB with security.
The big two things this year areCalyptra, this hardware root of
trust that can enable likeattestation.
This is like open source IP thatcan go into new devices like
SSDs and hard drive controllersto basically establish this
cryptographic authenticationattestation.
To basically establish thiscryptographic authentication
(30:42):
attestation.
But the really exciting newcapability that they put in
there this year is this thingcalled LOCK, this Layered Open
Source Cryptographic KeyManagement Block.
This is an open source IP blockfor key management backed by
Google, microsoft and then threeof the other SSD vendors to
basically ensure that this IP isreally good around key
(31:03):
management so that they cantrust that when they do a
cryptographic erase.
Or if somebody you know, in thecase of Opal we have a
self-encrypted drive Somebodywalks into a data center, puts
the drive in their pocket, walksout.
They want to know that thatdata cannot be decrypted, even
by a nation state actor withhundreds of millions of dollars,
and the only way they can knowfor a fact that that can't be
(31:28):
decrypted is to know how thekeys are managed on that device
with an open source keymanagement block.
So these are just two examplesof like that you know the
hyperscalers are taking thisvery seriously and you know,
devoting new technology and IPinto this space.
Speaker 1 (31:39):
And I think this takes the abstraction up one
step further and maybe, chris,I'll tap you is the idea of
there's local optimization,there's sort of server level,
rack level, data center leveland then energy sustainable data
center level.
So there's all these differentplaces this is happening.
(32:00):
So what are we seeing aroundactual data center level
optimizations happening?
Speaker 2 (32:06):
Well see, that's the neat news is that if you know
how to talk Redfish, you knowhow to talk Swordfish.
It's the same protocol.
It's an extension that givesyou all the neat things that
enterprise storage gives you.
However, what that means is Ican gather metrics from all of
my servers, all my storage soonto be switches.
I should be able to gather thatall from the same API calls the
(32:28):
same kind of mentality.
But, more importantly, I can doa cross vendor.
So I can talk to an HP, I cantalk to a Dell, I can talk to a
Cisco, I can talk to a Lenovo.
Supermicro, it doesn't matter,I can make the same call and
pull the same metrics.
Super micro, it doesn't matter,I can make the same call and
pull the same metrics.
So, just like we were talkingabout embodied costs with carbon
(32:48):
, that kind of thing, we canalso look at things like wattage
and fan speed and incoming airand outgoing air.
We can look at all that stuffcross-vendor with the same code
base.
And that's really good forcustomers, because customers
don't want to rewrite the codeto an SDK for eight different
vendors if they have eightdifferent vendors.
The average customer out thereuses two different storage
vendors or two different servervendors.
So they want a common code base.
And in fact we also talkedabout metrics.
(33:12):
How do we know what metrics torecord?
That's a neat thing aboutRedfish and Swordfish as well,
is that all of the differentpartners that help write
Swordfish, which make up theindustry, we all agree to it.
No one company comes in andsays I want this metric.
There's an OEM section for that.
Redfish has got an OEM section.
Swordfish has got an OEMsection.
They can put a custom counterin there.
(33:34):
But we encourage, if you've gota counter that's going to be
good for the industry, that youbring it forth and let your
competitors see the same counterand decide okay, we want to
subscribe to the same thing.
Once you have multiple peopledoing the same thing, it becomes
a standard counter.
A good example of this was thatwe were going to bring again.
I work for Hewlett PackardEnterprise.
(33:55):
We purchased a company calledCray Cray Supercomputers.
You probably have heard of themand we said we have a mandate
across our company.
Every server we ship is Redfishcompatible period, end of story
.
And we brought Cray in and Craywas not.
So we said well, we got to putRedfish on you guys and they say
, okay, that's great, but welooked through the protocol and
(34:16):
doesn't support water cooling.
So we're like, no problem,let's get water cooling in the
standard.
So we brought it forward and wehad Dell and super micro and
IBM, lenovo check the work tomake sure it's acceptable.
But now water cooling is in thestandard, so it doesn't take
much to expand the standard tocover everything in the industry
.
Um, but it is a cooperativeeffort, uh, and it's definitely
(34:40):
customer focused and and uh,definitely customer-focused and
definitely customer-friendly.
One code base to rule them all,as it were.
Speaker 1 (34:48):
Yeah, when it sure beats the, you know it's
deprecating and moving forwardin an efficient way and an
optimal way.
Just for people operations isso much better versus before way
.
Just for like people operationsis so much better versus before
.
Like I say this is I'd lookover in a drawer probably still
filled with nine pin to ethernetcisco cables, but just because
(35:08):
you never know when you're goingto need to grab that old switch
.
But imagine that you know I hadthe same thing for my storage
gear and my fiber switches.
It was like there was alwayssome proprietary thing on every
single thing and i've've let'sget away from that it's so
fantastic.
Speaker 2 (35:27):
How many times has one of us old-timers gone into a
lab with a fluke analyzer totry and figure out how much
power a server's pulling, butthen we don't really know what
it's going to do in production?
Or we buy a specialized PDUthat's got all the power draw
information so we can monitorthat thing.
Well, every server that's soldin the market today by every
vendor has a sea of sensors.
(35:48):
They all have a lot of sensorsthat give us a valuable wealth
of information.
If you want to know where ahotspot is in the data center,
that's easy to find.
If you have a situation where acertain model of a certain
vendor's hard drive got recalledbecause there was a bad batch,
you could write one command thatqueries everything in your data
(36:10):
center from different vendorsand finds where that drive
happens to show up.
So you know what you've got toreplace.
There's a lot of cases wherethis standardized approach
really really pays out for thecustomer.
Speaker 1 (36:26):
When it goes up to the software, even to the point
of placement right, like if wethink, uh, I saw some of the
stuff and even a couple of yearsago with like OCP was talking
about, and like Facebook doesthis with their data centers and
they actually move workloadsaccording to heat maps, because
they'll see, like a hot pocketbuilding up and you know that
(36:48):
they just literally can redeployworkloads and distribute them
throughout because of thephysical or whatever it's going
to be, can literally affect theefficiency of your entire data
center, because if it's put intoa spot where there's too much
(37:09):
other than required, it'sincredible.
And that same instrumentationthat tells you what the
temperature is also tells youwhere to place, because you can
choose.
That's the beauty ofstandardization.
Yep, exactly.
Speaker 2 (37:24):
and when you're talking about cooling.
Cooling you turn cooling up ordown like a sledgehammer it's.
You're going to cool 50 racksat a time, either by turning the
cooling up or turning thecooling down.
If you can make sure there areno hot spots across an entire
set of racks let's say 150, 200racks and you can make sure to
eliminate the hot spots, you canrun the entire set of racks
(37:46):
hotter than you normally would.
Generally people run an entireset of racks colder because they
have to make sure that thehottest server in the hottest
rack doesn't go over themandated number.
But if you can prevent thatserver from existing or prevent
that server from ever gettingthat hot, that means you can run
everything hotter.
You can run less airconditioning, you can run
(38:07):
everything closer to the edgeand that's a massive power
savings, because every watt ofCPU or computer IO you put in,
you're also putting a lot ofheat in.
So I mean that's half theequation right there.
Speaker 4 (38:22):
I'll certainly echo the excitement for Redfish being
just standardized.
You know, one of the things thatwe're looking at in the
sustainability space I mentionedis this classification of IT
and non-IT power.
Right, the fans are notcontributing to the useful work
and being able to basicallyeasily enumerate that in Redfish
to say this is the fan power,this is the rest of the system
(38:43):
power.
And you know for, like you know,we work on SSDs having the
ability to, within that server,say, okay, this is how much
power is going to all thesedifferent areas, and then I can
optimize things.
That is extremely helpful,right, because a lot of one of
the things we talked aboutearlier this year at FMS was,
you know, one of the reallybasic thing of just using NVMe
(39:06):
power states to take a drivefrom a 20 or 25 watt TDP down to
a 14 or 16 watt TDP, you canjust say that now you will not
consume more than that power.
Well, that tiny change can havea massive ripple effect on the
fan speed and the overall TCOand if you don't have that type
of Redfish instrumentation, it'svery hard to like actually say
I made this power change in mydrive.
(39:27):
I can do some basic math.
You know, 20 watts to 16 wattsshould save four watts per drive
.
Turns out it's actually a lotmore than that, because you get
the fan power too, and so,knowing that kind of stuff, to
be able to instrument withRedfish.
Speaker 2 (39:43):
And if you can tie the effect in with the value
you're saving, maybe it makessense.
But if you have a bad effectfrom lowering the power but you
can't show the value of thesavings you're getting, you may
never want to do that yeah, andso this is a great example of an
area where we've got a lot ofthe base instrumentation.
Speaker 3 (39:58):
We've got all that base instrumentation in from the
power.
We put the base instrumentationin but we have not and we we've
.
This is an area where we can goback in and put and look and
say when and where should we putthe controls in for those NVMe
power states?
Right, we did the basics in theSwordfish model and this is
where we differentiate betweenRedfish and Swordfish.
(40:18):
Right, redfish is where we putthe base power controls for the
chassis.
Swordfish is where we'll go putthat additional instrumentation
for those NVMe SSDs and wherewe'll look at both the controls
for the NVMe SSDs for thosepower states but also the
(40:40):
security around who gets tocontrol them, because you don't
want to expose those up too much, and so that's an area where we
will look at partnering inagain, putting those profiles,
those recipes in on where do weexpose those up, who should have
control over them?
And working with folks likeJohn Michael and Circularity and
(41:03):
NVME and getting those alliancepartners in to say who are the
folks with the expertise thereand where should we put those
partnerships in place so that wecan accelerate.
Maybe look at that in 2025 here, as we're two weeks away from
(41:23):
going into 2025 as we're doingthis recording today.
But get that as we go into 2025, can we make this a focus?
Get that functionality andthose controls into the
Swordfish specification and getthat base enablement in so we
can get those intoimplementations here in the next
(41:49):
year or two.
Speaker 1 (41:51):
And I think you hit one of the top things and you
know, I wish we could do thisall day long.
I would love to.
I only wish we had another hourto commit people to.
This is so fantastic, but it'sfar beyond just the specifics of
optimization and even when wetalk about sustainability it
from manufacturing to placement,to life cycle, to everything
(42:13):
security we often get mentioned.
People always say you know, whydid we have to?
You know, create this thingcalled dev sec ops?
You know, security is impliedlike oh no, the.
The presence of it was assumed,not implied.
You know, but in this community, what you've got now is the
same practitioner.
You know practitioners areactually then crossing the
(42:35):
working groups and so you cantalk about this and then you've
got security.
People can come in and seewhat's happening and they can
participate in this conversationand they can understand what
the impact is as they look at.
You know, protecting the apAPIs, protecting all the
endpoints, and it's this is thespot where it's people that have
(42:57):
a common goal in their ownindividual section, but the
really it's it's how do weoptimize the community and the
community of people that we meet, community and the community of
people that we meet and maybeI'll just ask each really
quickly say like, what's the oneaccidental thing you found
yourself working in.
It may be crossing workinggroups or just meeting somebody
(43:18):
in the community that yourealize, oh wow, this just saved
me a ton of time, or this was adiscovery that actually helped
you in your primary area,because you were beside somebody
that you wouldn't haveotherwise met had they not been
participating in this in thiscommunity.
Speaker 2 (43:35):
Maybe, chris, I'll start with you- um, it's funny
because I I end up meetingsomebody that does something
drastically different.
That is very interesting.
At every STC I go to I mean themost recent STC, I got to spend
some time with DNA storage guysand got some real deep skinny
(43:57):
on that stuff.
And the time before that it wasanother topic.
The time before that it wasanother.
I find that, unfortunately,there are too many technologies
to get into and I'm interestedin all of them, um, but I
wouldn't have access to thatunless I had gone to an event
(44:18):
where this cross collaborationhappens, and that's really the
magic of it is is meeting peoplewhere you get to cross
collaborate.
I'm looking forward to um.
I know they're working on aautomotive task force right now
to figure out if they can startworking with the people from
that.
Do the the networks for cars?
(44:38):
Um, I'm starting to researchthat.
That looks like really funstuff, um, but there's so much
to choose from there really isyeah, I mean, yeah, he's right.
Speaker 3 (44:50):
I mean, with there's, there's something that pops up
all the time.
It's like I, you know I'll comeback from a show and go.
Yeah, we just went on thisrandom.
Yeah, this one was a random.
Yeah, uh, uh, we dug into this.
Speaker 1 (45:07):
It's like a new journey every time you go there.
I I love it.
Speaker 3 (45:11):
A couple of years ago we were all off on a.
Hey, we're doing storage inspace.
Speaker 1 (45:17):
Yeah, yeah.
Speaker 3 (45:20):
Then there was the.
Here we're doing all of the.
We had the global mapping of.
Here's the story we're puttingstorage in planes, they're going
at high altitudes and doingreally GPS mapping of in really,
(45:46):
really granular detail, of ofof the earth, basically, um, and
so it was like all of this, thestorage aspects of it were were
doing really really high umgranular, and you had to make
sure that all the storage um wasum.
(46:09):
It was like really um.
I'm sorry I'm not describing itvery well.
It was just incrediblyfascinating.
But it was all like sub-zerotemperatures, that it was all
like immediately purged onlanding and different
(46:30):
requirements.
Yeah, really really really HA,and it was incredibly
fascinating.
And I forget almost all of thedetails of it because by the
time we got through, oh my gosh,this is fascinating.
And then I completely forgetabout it and go on to the next
topic.
Speaker 1 (46:42):
So we just solved the New Jersey drone problem right
now.
Is that what this is?
Speaker 2 (46:49):
And sometimes you get inspiration from the weirdest
places, like DreamWorks whenthey came in.
Speaker 3 (46:55):
How they're dead people usually fascinating while
we were working on it, and thenI completely forget all about
it and go to the next one.
Speaker 2 (47:04):
Yeah, which is why you should definitely take trip
notes when you go to theseevents.
Pull out that paper and writenotes through the event, because
you will forget day one by daythree and eventually videos are
posted online.
But you want to have your notesfrom those videos as well, and
if you share that with people atyour own company, not only will
(47:25):
they be more likely to send youthe next year, but you may find
that other people in yourcompany find the information
very useful for their jobs.
Speaker 3 (47:34):
I can't.
I can't remember the details ofall of them, because there's
just been so many of them.
Speaker 1 (47:38):
Yeah.
Speaker 3 (47:40):
Yeah.
Speaker 1 (47:41):
It is truly a wealth of humans and a wealth of
knowledge and such a sharingculture, which this is the
breakdown of.
There is no corporate wall.
When we're there, we make apoint of kind of avoiding our
associations, but we also areproud to say where we're from
and no one questions it.
(48:01):
It's just because we are thenew artifacts, the human
artifact, that can go throughand move from place to place and
achieve these amazing thingsand then again seeing it
actually play out.
Now we're actually seeing thereal results with some of the
stuff in sustainability, and itmay have been a marketing
buzzword six years ago, but whenthat went away, the work didn't
(48:24):
and we kept going and we keptachieving innovation.
So maybe just as a last close,jam M, what's got you excited
and what else in your SNEAconnections has been a wonder
that you probably didn't expect.
Speaker 4 (48:39):
Well, it's funny, I presented at SNEA STC this year
on sustainability, so that wasmy idea of like let's try to get
people that aren't usuallyexposed to the sustainability
world to start caring about itmore.
You know, it's slowly peopleare.
We're making them learn aboutcircularity and whatnot.
But, of course, like I get togo attend these presentations on
storage and AI, which is whatI'm spending most of my day job
(49:01):
learning about right now onmanageability and other stuff
that helps me with some of theside projects for GPUs and AI
workloads and all kinds of otherstuff I'm in.
So, yeah, it's.
It's just, I guess, what wejust said some of the uh, I
brought one of our like juniorsales guys to stc this year just
to like stop by and say hi andhe was like wow, are there
usually this many like seniorpeople from all these companies?
(49:21):
They're like.
He's like every single person Imeet is like director, senior
director, you know some kind oflike fellow or something like
that.
I was like, yes, that's welcometo stand.
Speaker 1 (49:32):
It's a, it's a beautiful thing and, uh well, I
can say I've been lucky as a.
My greatest gift is meeting allof you and being able to share
time and learn on a constantbasis.
And you know, it's, it's.
This is the.
The very reason why I love youknow, because we talked about
SaaS and space, we talked aboutsustainability.
I literally get to sit in onall these conversations and,
(49:54):
having come, you know, I'm inthe space myself as well.
So it's funny, not space, butthe space in the industry, I
guess, and it really isenriching, you know, for me,
because then, as I go to thenext company and I'm advising
companies and startups andeverybody, I say, like you want
to meet people who are drivingchange and who are your future.
(50:17):
You know, collaborators, thisis the group to be in.
So there's lots of virtualevents.
There's lots of in-personevents.
We've had a really, reallygreat change with the membership
structure for 2025.
So there's a podcast where weshare about some of the updates.
I talked with Jay Metz aboutthat.
So definitely recommend peoplego check out the podcast, Go
check out the SNEA YouTubechannel.
(50:38):
There is no shortage offantastic content.
But, most importantly, getsigned up, come to an event.
There's virtual events, thereare in-person events and, oh,
thank goodness, we're back inperson.
It's so great to be able to bein the room, because the hallway
track is one of my favoriteparts of every conference
there's a lot of hallwayconversations.
Speaker 2 (50:59):
That is so true.
Speaker 1 (51:01):
And I can tell you that I've literally met who
would become future clients andfriends and employers through
these hallways.
And so thank you all forletting me in the hallway and
thank you all for the folks thatare listening and watching.
Do continue, so keep your eyesout.
2025 is going to bring a lotmore great content like this.
(51:22):
So, with that, chris RochelleJM, thank you so much.
And again, folks, don't forgetto sign up to the SNEA Experts
on Data podcast.
Hit, subscribe, smash that likebutton, do all those things
that the Zoomers tell us we'resupposed to do, but, most
importantly, get to a SNEA event, find out how this community
can help you and what you cancontribute back.
(51:44):
And thank you all for watchingand listening.
Speaker 4 (51:49):
Thanks all, Bye-bye.
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
Ridiculous History
History is beautiful, brutal and, often, ridiculous. Join Ben Bowlin and Noel Brown as they dive into some of the weirdest stories from across the span of human civilization in Ridiculous History, a podcast by iHeartRadio.