November 29, 2023 • 39 mins
Philip Alsop, Digitalisation World Editor, interviews Dr. J Metz, Chair of the SNIA Board of Directors on the many ways SNIA is setting the standard for storage innovation. Listen as they discuss several new exciting innovations, including computational storage devices, DNA data storage and SDXI (Smart Data Acceleration Interface).
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 (00:06):
Welcome to the SNEA on Data Podcast.
Each episode highlights keytechnologies related to handling
and optimizing data.
You were listening to SNEA onData Podcast SNEA setting the
standard Restorage innovation.
In this episode, philip Alsopof Digitalization World
interviews Dr Jay Mett, chair ofthe SNEA Board of Directors,
discussing several recentexciting innovations at SNEA.
(00:28):
So SNEA.
Speaker 2 (00:30):
You have just released a new set of standards
around a new type of storageproduct with the title
Computational Storage Devices,or CSXs for short.
Can you just give us a bit ofan insight as to what's going on
there?
Speaker 3 (00:45):
Yeah, absolutely.
The idea behind computationalstorage is a little bit of an
advancement of some of thethings that have been happening
over the last several years.
Specifically, what we have beenrealizing as part of the
industry trends is that storageand memory are starting to
become somewhat similar, and ifyou can have memory and storage
(01:08):
using similar semantics or thesame semantics, there's no
reason why you can't havecompute right next to it and you
can't have networking rightnext to it.
So the ultimate idea here atthe bottom line is that if you
want to be able to do any kindof computing on data, it makes
sense to have the computer asclose to the data as possible.
And there are a lot ofscenarios where having data
(01:29):
right at the computer processorsmakes a great deal of sense,
because you don't necessarilywant to move data from one place
to the other if you don't haveto.
So even over a short distance,like you find inside of a server
, any IO that you don't have tosend saves quite a bit more IO
in retransmissions, retiming andthat kind of thing and, more
(01:55):
importantly, it allows you toput the processing capabilities
where it's needed most.
So sometimes you want to haveprocessing on metadata,
sometimes you want to haveprocessing on data itself,
sometimes you want to haveprocessing in the control plane,
and so on.
There's a lot of differentprocesses that go on, but
there's also a lot of differentstorage that goes on.
Not all storage is created equal.
So what you want to do is youwant to find the right tool to
(02:18):
place the right job.
What computational storage doesis it gives you a series of
options that allows you to dojust that.
I mean, for instance, right nowpeople think of computational
storage as a drive with a nicelittle processor on it, and
while that's true, that iscomputational storage, that's
not all it is.
So storage is more than justdata.
(02:39):
It's more than just storingdata.
It includes all the protectionthat you have to have to make
sure that you get the correctbit back that you asked for.
It means that you have to havethe services like encryption and
compression and that kind ofthing.
It means you have to have thefile systems being able to
operate on All of these things.
That all work together to makestorage have a wide swath of
(03:06):
responsibilities, and placingthem where they're best used is
what computational storage isall about.
So it's the very beginningstages of what's promising.
Speaker 2 (03:18):
And I'm wondering is that?
I mean, everyone's been talkingabout the edge computing for
quite a while.
Is it one of the main drivers,or is that a happy coincidence
that bringing computing storagetogether, which I guess is
making, will make sense of theedge?
Is it just a happy coincidence,or is it that people have been
looking at and saying, hey,isn't this a good idea?
That's one of the majorbenefits, or, as you've
(03:38):
explained in many other places,it makes sense as well.
Speaker 3 (03:42):
Well, it's kind of serendipitous.
I don't think it was acoincidence and I don't think it
was necessarily by design, butI do think that when you get
technologies that happen toevolve along similar timelines,
you're likely to find thosekinds of intersections.
Now, any kind of edge computingsituation, whether it be
automotive, whether it be mobiledevices or edge of a data
(04:06):
center all this IoT stuff that'sbeen happening and just the
sheer amount of informationdoesn't necessarily have to be
sent somewhere to be processed,if it doesn't have to be.
Some of it does.
Sometimes you really do want tohave massive collections of
data to be able to do the kindof number crunching that a
handheld device can't really do.
But if you want to, let me giveyou an example of what some of
(04:27):
the things you can do.
So, for instance, in machinelearning, where you actually
have to transform the data to beable to process it, you can't
just do analytics on objectstorage, not natively.
You have to make the datatransform so that the analytics
engines can use it.
Well, what we normally have todo without any kind of
computational devices is youhave to send it somewhere for
(04:49):
that pre-processing to be doneand then you have to run the
models.
But that first step, thatpre-processing element, could
very well be done at the edge,so that you're not necessarily
sending all that data over intoa centralized location.
So one really good example forthat is facial recognition
(05:11):
software for cameras.
So if you think about the waythat that would work, you have a
centralized server that doesthe processing for facial
recognition.
The camera sends the feeds backto that central location, that
application starts to do thefacial processing and then it
has to communicate with all thedifferent devices.
However, with a computationalstorage solution, you could run
(05:34):
the facial net recognition righton the camera and then only
send back the information thatyou wanted to find, as opposed
to just sucking in the ocean andtrying to find that small
little bit that you're lookingfor.
Being able to do thatseparation of duties is a major
time saver, which of coursesaves processing power, saves
(05:56):
cooling, saves network.
Overall the long term it justdoes an awful lot of saving of
energy and economics.
Speaker 2 (06:07):
Okay, and then I mean you've been very busy so I'm
aware as well.
You've just published version Ithink 1.0, of the I'll have to
read this computational storagearchitecture and programming
models.
So can you give us thebackground on that?
Speaker 3 (06:22):
Well, in order to give vendors the general
concepts that they want to usein order to develop a
computational storage device, ithelps to be able to understand
A what it is, b what you can dowith it and C how you can access
it.
So what this does is it providesa model of computational
(06:43):
storage that allows you tocreate different form factors
that the nature of therelationship between a processor
, the storage that it uses, thememory that it uses and the
corresponding form factors comeas a result of it.
The APIs is simply a way ofhaving a programmer be able to
access a computational storagedevice to operate it from any
(07:06):
kind of host or remote location.
Now, that's different thanwhat's being done inside of the
NVM Express group.
By the way, demo will make itclear that there's a separation.
So the model and thearchitecture and the APIs are
being handled inside a SNEA andthe commands for NVME are being
done inside of NVM Express.
So the NVM Express is aslightly more narrowly focused
(07:30):
version.
It's really focusing on thedrives at the moment, not so
much the arrays or theprocessors or any of the kinds
of broader flexibility.
Like I said, we're at thebeginning stage of this, but
NVME is responsible for theirown command sets, their own
theory of operation, which isoutside the scope of SNEA.
So we're working together withthat organization to make sure
(07:52):
that everything is aligned.
Speaker 2 (07:55):
OK, and just for a little bit of understanding with
both the things we've coveredso far, is it vendors start
producing these new types, thenew ideas, and then you get
involved if you're like referee,to come up with open stands
that we're going to work to, ordo you foresee or anticipate
something coming along and thenstep in early, just for people
(08:16):
to understand where does SNEAget involved at what stage?
Speaker 3 (08:19):
Well, the ideas can come from anywhere.
There really is no one way tostart an idea, but the process
of once an idea gets done, youget a couple of other companies
or other members to think thatyeah, that's something that's
worth working on.
And then, if you get a criticalmass, they come together and
they start to identify what isit that we should do to make
(08:42):
this work for the long term, forthe long haul, I mean?
Ultimately, if you're lookingjust to have a solution that two
vendors can work together, youdon't need a standards body.
But once you get beyond that,you start to realize that I need
to make sure that if I'm goingto get a computational storage
device from vendor A and I'mgoing to get another
(09:03):
computational storage devicefrom vendor B, outside of the
special sauce that they want tohave to make it work, you kind
of want it to be relativelysimilar in the structure, like
the electric plug in your wall.
You don't want severaldifferent plug sizes, you don't
want several different plugsthemselves.
So at the very fundamentallevel you have to agree on what
that's going to look like.
(09:23):
Now that doesn't prevent youfrom being innovative by working
to build on top of that.
But in this particular case, youhave a couple of companies who
are trying to developcomputational storage models
that they're interested in, andso they get together with other
(09:43):
companies who want to do thisand they kind of hash out the
basics so that they can actuallygo onto the market and say I
have this device, and then thecustomers and the administrators
will kind of have a generalidea of what you're talking
about.
For example, in NVMe we startedoff with NVMe Express for PCIe
(10:04):
devices, and the whole purposewas that if you want to have
that kind of really fast storagedevice and you want to stick it
into a PCIe slot, you don'twant to have to rewrite a driver
for every single thing that youwant to put in there.
Well, computational storage isthe same way.
You don't want to have to havea very specific solution for
(10:28):
every single one that comes downthe road Because, quite frankly
, there are a lot of differentpossibilities and options and
you've got to have some place tostart with.
So in this particular case, weactually have 52 different
members who happen to have avery similar idea and they came
together to start working on thespecification.
Speaker 2 (10:50):
OK, and in a slightly similar vein, I see that the
DNA data storage alliance havebecome an associate member.
Yeah, technically OK.
So, and I'm just interested, Ithink there was a track on DNA
storage in your recent StorageDevelopers conference.
But again, for those of usmyself include that maybe I'll
sit too up to speed on that,what is it, and is it as
(11:12):
exciting as I'm imagining itcould be?
Speaker 3 (11:15):
I think it is personally.
I mean, I have to confess I'mno biologist, so I don't really
understand a lot of the veryspecific things or chemist, I
guess, is probably the best wayto, I don't know.
It's.
It covers a lot of technicalareas but, in a nutshell, one of
the things that's beenconcerning a lot of people is
the fact that we have just hadan explosion of information in
(11:37):
the last 10 years.
Right, we've, we've, we'vedeveloped what we've, we have
created more information in thelast three or four years than
all of human existence At themind boggling.
You know, you know, it's justone of those.
It's one of those weird thingswhere you know the numbers stop
(11:57):
me having any meaning but wherewe were doing long term
retention and archiving of thisdata, it was reasonably
straightforward.
I mean, you have tape, you'vegot the hard drives, you've got
other types of records that youcan keep for the long term
no-transcript.
But this type of data, thisamount of data, means you've got
(12:18):
to really start thinkingoutside the box.
You really have to startfiguring out.
What does it mean to have trulylong-term archiving information
that's going to last a verylong time and going to be
retrievable in the future.
What DNA storage is trying todo is say look, the DNA molecule
can actually act as a storagemedium for that long-term.
(12:42):
Now it's still in the veryearly stages.
What that means, and if youthink about it from a very basic
common sense perspective, youhave to be able to access that
DNA strand.
You have to be able to encodeit in such a way that's readable
in 40, 50 years or 100 years orwhatever it is.
You have to be able to make themachinery that can do all that
(13:04):
work.
You have to be able to identifythe standards of access, the
software that's supposed to be,or what's it going to look like?
What do you need?
Because you're not going toread DNA the same way that you
would read data off of a platter, for example.
It's not exactly the blockstorage that we're used to with
an SSD.
All of these things need to bedefined.
That's where the DNA datastorage alliance has been
(13:30):
starting off, and what Sneobrings to this is the umbrella
to be able to do that work inthe atmosphere of a storage
related context.
We can provide the supportnecessary for them so that they
can focus on the work that theywant to do.
Speaker 2 (13:49):
I'm a student of the storage industry for quite a
while and I think it was aholographic storage was out
there for a while.
I think people were storing onglass as well at one point.
I'm not sure whether those arestill going, but they seemed a
bit more pie in the sky.
Do you think this is a realconcrete thing that's going to
happen, or is it still too earlyto actually work out whether
(14:10):
it's going to have the impactthat it might have?
Speaker 3 (14:13):
Well, anytime you start talking about science
fiction-like stuff, you alwaysrun the risk of it either
petering out or hitting aroadblock that you couldn't have
anticipated.
I'm not sure what happened withholographic storage.
I thought it was reallyinteresting as a concept and I
remember they were talking aboutthis back in I think it was
like late 90s, if I remember ifthey started it.
(14:35):
And there's also molecularstorage, which doesn't sound all
that unrelated.
I can't say for sure because Ireally don't know anything about
that, but I do think that onceyou hit a critical mass of
interest and the amount ofinvestment that goes into it's
(14:56):
not just in terms of money buthuman investment and being able
to put this together, eventuallythe need will outweigh the risk
and you'll have to getsomething to work over time.
We just simply don't have thecapacity in our existing storage
media to handle the influx ofdata that's coming in for any
(15:21):
kind of long-term science,especially when you start to
think about the skill andshriveless of privacy and
retention and the need to holdon to all this data for
regulatory purposes.
So you wind up with this kindof external influence that would
perpetuate this and an internalmotivation to keep going.
(15:43):
I think that, from what I'veseen from the people and the
companies that are involved inDNA storage, that there is a
critical mass of people who arelooking to make this work, and
work successfully.
They've organized themselves insuch a way that they're
obviously in this for the longhaul.
Speaker 2 (16:01):
OK, so that was, I say, one of the main themes, I
think, at your StorageDevelopers Conference.
I think the other tracks werearound SRM or, dare I say, old
friends, storage ResourceManagement although I'm sure
there was something new to sayabout that and also cloud
storage.
So can you just give us I thinkyou've just come back from
there, but can you give us justa bit of a highlights of what
those topics brought to light,if you like?
Speaker 3 (16:25):
Well, yeah, sure.
So let's talk about themanagement for a second, because
you're right, it is a majorissue that needs to be resolved.
I think, in general, peopletend to think of management as
an afterthought when they'repromoting storage.
Because, let's face it,management really isn't sexy
unless you're an administrator,and it makes your life easier.
The end user is the one that'sreally going to get excited
(16:49):
about this kind of thing, likethe same way that an accountant
would be excited about aspreadsheet, just to make the
lives easier.
Well, management in the storageworld has been needing to evolve
to keep up with the changesthat are going on, because what
used to be the storagemanagement of choice, which was
something called SMIS, wasdesigned to be around large
(17:12):
arrays, centralized storagedevices.
Well, since that time, we'veactually moved into other forms
of storage, includingdisaggregated storage, ethernet,
attached SSDs, ebofs, thosekinds of Ethernet bunch of flash
, and you've got to be able toprovision those somehow.
So the DMTF has its Redfishspecification for server
(17:40):
management, and so SNEA has arelationship with DMTF to create
the swordfish extensions forRedfish.
So swordfish is really Redfish,but for storage.
So, if you want to manage thesenew types of devices, these
NVMe, ethernet, attached SSDs orthe eBofs or the disaggregated
storage.
(18:01):
Swordfish will allow you to dothis without having to reinvent
the wheel, especially if you'realready using Redfish for your
servers.
So if you have a storage server, which is basically a box full
of sword devices, you can managethe whole thing with one basic
management structure.
And so we just released thelatest version of swordfish as
(18:21):
an ISO specification, and thatjust happened a couple of weeks
ago, I believe.
So we are heavily promoting thecompliance testing programs for
Swordfish.
I think before we were tryingto find the overlapping time
because companies still use SMIS, but the development for us has
(18:44):
shifted over to Swordfish, butit's never a clean transition
from one to the other.
What we've done this year isthat look, this is where we
think we need to go.
We're spending a lot of timepromoting and helping people
understand how to use Swordfishto get the best management
structure for their devices.
Speaker 2 (19:07):
Cloud storage again.
That's been around for a while.
Was there anything particularlynew there, or was it just more
of the same, do I say?
Speaker 3 (19:15):
Well, I have to confess, when it comes to the
cloud storage, it came at aconflict of some other things
that I was doing and I didn'tget a lot of chance to really
spend it.
But the cloud storage is in andof itself is rather vague.
You've got the storage devices,for instance.
You've got the storage access.
You've got the storageprotocols, object stores, those
(19:35):
kinds of things.
I think that one of the thingsthat is striking the cloud
storage folks is that it's morethan just more performance.
The issue with cloud storage ingeneral is that you've got to go
through an awful lot of layersto get to the storage from the
cloud applications.
What's been happening,especially recently, is that
(20:01):
when you wind up with theseapplications that don't quite
work the way a double-arbitwants it, they just build
another layer of abstraction.
I don't know if you've spent alot of time going through all of
the different layers of how toget a single application in the
container to talk to its actualstorage, but it's impressive how
many layers there are.
(20:21):
From an application'sperspective, all they want is
more raw power.
They want more raw processingpower, and so you eliminate all
of the benefits and features.
That is built into the protocolthat's built into the devices,
that's built into the processors.
Excuse me, what we were talkingabout in part was well, how do
(20:44):
you capitalize on those?
You're just ignoring a lot ofthings that could make your life
easier, but most softwaredevelopers really don't know
about it.
We were tying the stringstogether and hopefully aren't
knots but we're trying to getthem to start paying attention
to the idea that there arethings that you can do in
(21:06):
storage that you may notnecessarily be aware of because
it's been hidden from you for solong.
I hope personally that thedevelopers and the software
authors and the people who arein the living in the container
world can realize that we'vebeen doing this for a long time
and can help.
We can actually help you getthe performance you need without
(21:28):
just waiting for the next Uberprocessor from AMD or Intel.
I think that, ultimately, whatyou want to do is you want to
work smarter, not necessarilywork harder, and that's where
the cloud storage stuff reallytends to shine.
Speaker 2 (21:43):
Okay, it sounds as if you're very busy as an
organization, but are there anyother things on the road map we
need to be aware of at thismoment in time, or with what
we've discussed?
Is that keeping you pretty muchfully occupied?
Speaker 3 (21:56):
Well, as a chair, I have a desire to see the
organization expand and grow interms of the strength of its
quality of offerings, notnecessarily become bigger, but
definitely become much moreresilient and reliable for those
(22:17):
who are depending upon us.
So, for example, we're talkingabout the cloud storage stuff.
One of the really, reallyinteresting things that's
happening is something calledSDXI, which stands for the Smart
Data Accelerator Interface, andeffectively what it does at a
very high level is it actuallygoing back to the cloud storage
conversation?
It does that very thing that wewere talking about with the
(22:39):
application being able to usethe smarts of the hardware.
So let's say, for example, ifyou were to, let's say you have
a storage VM and a regular VMand you need to move the data
from one back to the other.
In an environment with a lot oflayers of abstraction, you
effectively have this Pachinkochip that has to go all the way
down to the hardware and all theway back up to the other
(23:02):
virtual machine.
But with SDXI, what you can dois you can tell the hardware,
using a strict API, toeffectively do the data movement
from one memory location toanother privileged memory
location.
Once you set up the privilegedrelationship.
It'll just happen because thehardware can do it for you.
You don't have to go throughall of these layers of APIs to
(23:24):
get that layer to do this andthen that layer to do this and
then pass the copy down all theway down to the hardware and
then back up again.
Just tell the hardware to moveit, hardware moves it and you
can really speed up andaccelerate that data movement.
Let's say you want to zero out alocation, a memory location.
You don't need anywhere.
(23:45):
These containers are ephemeral.
You want to zero it out andmake it nice and secure and
start all over from scratch.
Tell the hardware to do it.
Don't just have to go throughand cycle through the software.
To have it done in that fashionCan save up an awful lot of
time.
There's work that's being doneinside of SNEA to help assist
that in a vendor neutral way, sothat you're not dependent on
(24:06):
any one particular type ofhardware vendor, any type of
software vendor.
It's all basically part of anopen standard for that kind of
data memory movement.
That's one example what we'vealready talked about Swordfish,
we talked about computationalstorage.
We've talked about some cloudsecurity and data privacy are a
(24:28):
big part of it.
I think energy efficiency is aswell.
So one of the things that'shappening all over the world is
the movement of governmentregulations and, as most people
know, either intuitively orthrough experience and if it's
from experience, I share theirpain but government regulations
(24:52):
are not always consistent and sooftentimes you'll come across
very strict rules that conflictwith each other.
So one of the things that we'retrying to do, especially with
the energy efficiency stuff andespecially with the data privacy
and protection, is we're tryingto make that approachable for
(25:12):
people.
So we have something called theGreen Storage Initiative, which
allows companies to runcompliance of their machinery
against the energy starrequirements that are here in
the United States.
There are some EU and UK mostlythe EU, not more than the UK
(25:33):
work that is being done forhandling data location, where
the data actually gets stored,how it gets stored, what the
energy needs to be for those.
We're working with thosedifferent governments as well as
the United States, to make surethat it's easy for vendors to
comply, or easier.
They'll always build a moreinteresting regulation.
But those are the kinds ofthings that we're doing that
(25:55):
aren't necessarily sexy, but thepenalties could be rather
severe.
Speaker 2 (26:01):
Yeah, okay.
And yet mention of energyefficiency.
I mean, the world is headingtowards net zero at varying
paces depending where you are,what industries, what, I suppose
, responsibility do you thinkSNEA has as an organization?
You mentioned the energyefficiency, but that sounds as
if, at least at one level,you're responding to strict sort
(26:22):
of government requirements.
But more generally, theindustry, are you helping them
move towards that sort of target?
And in terms of data, I knowobviously your vendors are there
to provide storage and there'smore and more data as you
discuss being producedridiculous amounts.
Do you think you have any sortof requirement, obligation, to
(26:46):
start some kind of debate orenter into the debate as to
should we just be storing allthis information forever and
ever, or do we actually need tostart thinking about the life
cycle of data a bit moreseriously?
Would you just some thoughts onwhat your association's role is
in the sustainability debate?
Speaker 3 (27:02):
That requires a shift in our focus a little bit, and
it's a shift we've been talkingabout.
So from my perspective, I havealways taken the stance that, as
the quote unquote storageexperts, I think we have a moral
responsibility for being ableto lend that experience into the
world, where these types ofregulations can have deleterious
(27:24):
effects on whether or notyou're storing the data or
whether or not you're storingthe data properly.
And I do think that differentgovernments may have the best
intentions, or they may not younever really know but a lot of
them are making decisions thatdon't make an awful lot of sense
from a technical perspective.
And I think that SNEA is anorganization that has a moral
(27:47):
duty to be a friend of thosecourts, to be able to offer some
recommendations and someguidance.
We're not a lobbyistorganization, so we don't go
onto K Street and we don't meetwith anybody in the governments,
but we do send out statementsand white papers and position
papers for certain things, whenprobably we've done a few of
(28:08):
them this summer, as a matter offact, for some of the security
things that have come out withregards to some of the changes
that have been going on Forstandards.
It's really kind of funnybecause you have this
overarching umbrella ofaccountability, right?
So when you talk to somebodywho's in standards, they get
(28:31):
very mired in the word shall orthe word should and it seems
like minutia to the layperson.
But that becomes the basis fora lot of this regulation, right?
So if the regulation says youshall do something, or you
should do something, or you mustdo something, you're never
(28:52):
gonna see something that's like,well, you may do it.
It doesn't really quite work outvery well when you start using
the word like yeah, you can oryou may find the language every
once in a while slip through,but very rarely.
But our role is to make that benice and clear, right?
Because if the governmentresolves and the government
(29:14):
needs to know what's doable, youcan't just say do this and have
it be an impossibility.
At the same time, when they dopass these kinds of regulation,
we need to tell our members thisis what it means, this is how
you'd actually start applyingthis.
So we act as kind of anintermediary of information.
And no, I have to confess it'snot sexy, it's not the kind of
thing that attracts a lot ofpeople, but it does have a
(29:38):
fundamental purpose that needsto be an ongoing solution.
Speaker 2 (29:46):
And out of the pandemic.
I mean sustainability.
I'm not saying it came out ofthe pandemic, but it's become a
bigger topic.
So, similarly, the supply chainhas come under sort of intense
scrutiny and some people, Iguess, maybe hoping, once the
situation in Eastern Europe etcetera, calms down, things will
go back to normal, whatever thatmight mean.
(30:07):
But do you think supply chainsyour members are already on need
to start re-evaluating wherethey're getting their resources
from?
Or, as you say, do you think ifthey're sitting out for another
six months, everything's backto normal?
Or do you think it's along-term issue that needs
addressing?
Speaker 3 (30:27):
I would caution anybody into thinking that they
can wait and expect things to goback to normal, Because what
constitutes normal nowadays?
It's not the way it was twoyears ago.
We have the geopolitical impacton supply, for data and
communication has radicallychanged in the last three years,
(30:48):
so the Eastern Europe situationis actually a little bit more
dire than people realize.
There are a lot of softwarehouses in the Ukraine.
There's a lot of informationthat are coming out of that area
of the world.
Even Russia itself has offeredconsiderable amounts of positive
(31:09):
technology contributions intothe industry.
There's China, there's Taiwan.
There's the global impact ofthe tensions going on there.
We still have some of thesanctions on Chinese companies.
In the United States we havenew regulations about what can
(31:30):
and can't be sent into and outof the Chinese territories.
So that's not going away in sixmonths.
I think anybody who thinks thatit's gonna take a six month
hiatus and then have everythinggo back to perfection is
probably fooling themselves.
What I think we have to do iswe have to.
(31:50):
We can't just try to accept allthis in one lump sum.
We have to be vigilant, excuseme.
We have to be diligent andvigilant.
Say that three times fast.
We need to be aware of what'shappening in the tensions in
India and Pakistan, and all ofthese things have a much greater
(32:14):
impact all the way intoPoughkeepsie.
I mean, it's one of thosethings or leads.
Either way, it's going to beimpacted somewhere, whether it's
going to be at the servicelevel, it's going to be impacted
at the price level, it's goingto be impacted on the regulation
level.
I think one of the other thingsthat's happening too is the
(32:37):
content of the data is startingto get more scrutiny.
There are things that arehappening with unelected
governmental, non-governmentalorganizations that are starting
to make people question whatkind of data can be sent to whom
and when.
Who should be able to have it,who shouldn't be able to use it?
(32:59):
Those all have really profoundimpacts.
So I honestly think that wedon't have all the answers.
I'm not pretending that we do,but I do think that you need to
have somebody.
You can ask a question tosomebody who is also taking
ongoing attention to what'shappening, Because if you're not
(33:21):
careful, you could wind upbeing on the business end of his
event, and I don't thinkanybody really wants that.
Speaker 2 (33:30):
Okay, and then maybe, just as we finish, the storage
industry more generally.
I mean, I've sort of followedit on and off since the early
days of Fiber Channel, which isa wee while ago, and there's
plenty of names that were like,if you like, the bedrock of the
industry then and no longeraround.
It's a classic, I guess,technology that there were a lot
of people in early.
(33:50):
Then it's consolidated.
Do you worry at all that theconsolidation has perhaps gone
too far and that the gene pool,if you like, is too small, so
there's not enough innovation,because whenever somebody does
something clever, it seems to behoovered up, dare I say, but
one of the large organizers.
Or do you think the industry isvery healthy and, as we
discussed earlier, there aresome innovative ideas going on
(34:12):
and therefore all is good in thegarden?
Speaker 3 (34:16):
Well, I don't know how good the garden is, but I
think that if you look at thehistory of technology, you'll
see that there's a bit of anaccordion effect.
Expansion and contraction hasbeen normal and it's been going
on ever since the days of Univac.
So if you follow through thehistory of any technology all
(34:40):
the way back, if you want to goall the way back to the
telegraph, it's exactly the samething.
The history has been consistent.
Same thing with hardware andsoftware.
There was a period of time nottoo long ago, when people said
that hardware was completelyirrelevant.
All you needed to do issoftware.
And now, where has really theinnovation come?
Through Hardware, we now havethe universal chiplet consortium
(35:05):
that started out.
We've got SNEA, we've gotEnviom Express, we've got CXL,
which has really taken theserver world by storm.
You have all of thesetechnologies that are
hardware-oriented,hardware-related.
That will make the softwarebetter and then the software
will improve as well.
And then the software will getbetter and we'll start to
(35:26):
eliminate some of those layersof abstraction, because that got
too big.
The software component is.
I mean, if you would just lookat a stack diagram of all the
latency, software is huge andpretty soon.
People want to get betterperformance, so they reduced
that software and they startfocusing on the hardware.
The accordion starts to comeback.
(35:46):
As long as the accordion ismoving, we're healthy.
It's when people start toartificially try to stop it,
then things get into trouble.
And from SNEA's perspective, weare perfectly happy to play the
accordion.
We are happy to go ahead andencourage the development where
(36:07):
it needs to be, like the DNAdata storage lines.
Perfect example.
That is real, true, innovativethinking is truly thinking
outside the box and literally asit turns out.
So I think that there's alwaysgoing to be somebody who wants
(36:27):
something more or somethingbetter, something more efficient
, something less costly, andthey're going to have an idea to
do it.
Some of them will win, some ofthem won't, and by winning I
mean just be used.
It doesn't have to be the best,it doesn't have to be the
greatest, it just has to besomething that people are
working on.
From our perspective, we'reperfectly happy to be home for
(36:48):
those kinds of things, andthat's, I think, where we're
going to be focusing in thefuture.
Speaker 2 (36:55):
Okay, and then maybe just finally.
Finally, in terms of the enduser, I always feel slightly
sorry for them because they'regetting bombarded with all these
great new ideas and thingsgoing on.
They're one day being told tobe on the premise and then the
next go in the cloud and all therest.
I know it's very difficult tolimit yourself, maybe, to one or
two, but if you had a couple ofpieces of advice, when users,
(37:16):
faced with this enormousportfolio of storage
technologies and stuff, where doyou think they should
concentrate to get the quickestresults?
Or what do they need to belooking at, do you think, to
improve their storageperformance?
Speaker 3 (37:31):
Well for all.
Number one don't panic andbring your towel.
Number two SNEA has been doingan awful lot of work over the
last several years to try tobreak a lot of this down into
plain English.
So we have probably a libraryof hundreds of hours of plain
English content that they can goto by going to the website or,
because we have a YouTube video,you can go into YouTube and go
(37:53):
to our YouTube channel too.
It's vendor neutral.
It's technology neutral, whichmeans that there's no promotion
of any one vendor over anotheror anyone technology over
another.
So if you want to find out aboutthe pros and cons of Fiber
Channel, which is still verystrong storage technology, you
can do that.
If you want to find out aboutXPUs, which are brand new and
very confusing to a lot ofpeople SmartNix, computational
(38:15):
storage these are just some ofthe more recent ones, but we
also go back to the basics.
If you don't know what RAID isor what RAID means, if you're
not sure what an initiator or atarget is, you just want to get
started.
We have those kinds of thingsas well.
We have a series calledEverything you Wanted to Know
About Storage, what we're tooproud to ask.
So, everything from theabsolute basics to the very new
(38:37):
complex topics, we do an awfullot of education that is just
freely available for people.
All they have to do is look,and so we're going to continue
to do that.
It's a major pillar of ourmission.
Speaker 2 (38:51):
Okay, well, I've taken up a lot of your time and
I really appreciate it.
Give us some fascinatinginsights, so thank you very much
indeed for being with us.
Thank you.
Speaker 3 (38:59):
Thank you for the invite.
Speaker 1 (39:01):
Thank you for listening.
For additional information onthe material presented in this
podcast, be sure and check outour educational library at
sniaorg.
Welcome to the SNEA on Data Podcast.
Each episode highlights keytechnologies related to handling
and optimizing data.
You were listening to SNEA onData Podcast SNEA setting the
standard Restorage innovation.
In this episode, philip Alsopof Digitalization World
interviews Dr Jay Mett, chair ofthe SNEA Board of Directors,
discussing several recentexciting innovations at SNEA.
(00:28):
So SNEA.
Speaker 2 (00:30):
You have just released a new set of standards
around a new type of storageproduct with the title
Computational Storage Devices,or CSXs for short.
Can you just give us a bit ofan insight as to what's going on
there?
Speaker 3 (00:45):
Yeah, absolutely.
The idea behind computationalstorage is a little bit of an
advancement of some of thethings that have been happening
over the last several years.
Specifically, what we have beenrealizing as part of the
industry trends is that storageand memory are starting to
become somewhat similar, and ifyou can have memory and storage
(01:08):
using similar semantics or thesame semantics, there's no
reason why you can't havecompute right next to it and you
can't have networking rightnext to it.
So the ultimate idea here atthe bottom line is that if you
want to be able to do any kindof computing on data, it makes
sense to have the computer asclose to the data as possible.
And there are a lot ofscenarios where having data
(01:29):
right at the computer processorsmakes a great deal of sense,
because you don't necessarilywant to move data from one place
to the other if you don't haveto.
So even over a short distance,like you find inside of a server
, any IO that you don't have tosend saves quite a bit more IO
in retransmissions, retiming andthat kind of thing and, more
(01:55):
importantly, it allows you toput the processing capabilities
where it's needed most.
So sometimes you want to haveprocessing on metadata,
sometimes you want to haveprocessing on data itself,
sometimes you want to haveprocessing in the control plane,
and so on.
There's a lot of differentprocesses that go on, but
there's also a lot of differentstorage that goes on.
Not all storage is created equal.
So what you want to do is youwant to find the right tool to
(02:18):
place the right job.
What computational storage doesis it gives you a series of
options that allows you to dojust that.
I mean, for instance, right nowpeople think of computational
storage as a drive with a nicelittle processor on it, and
while that's true, that iscomputational storage, that's
not all it is.
So storage is more than justdata.
(02:39):
It's more than just storingdata.
It includes all the protectionthat you have to have to make
sure that you get the correctbit back that you asked for.
It means that you have to havethe services like encryption and
compression and that kind ofthing.
It means you have to have thefile systems being able to
operate on All of these things.
That all work together to makestorage have a wide swath of
(03:06):
responsibilities, and placingthem where they're best used is
what computational storage isall about.
So it's the very beginningstages of what's promising.
Speaker 2 (03:18):
And I'm wondering is that?
I mean, everyone's been talkingabout the edge computing for
quite a while.
Is it one of the main drivers,or is that a happy coincidence
that bringing computing storagetogether, which I guess is
making, will make sense of theedge?
Is it just a happy coincidence,or is it that people have been
looking at and saying, hey,isn't this a good idea?
That's one of the majorbenefits, or, as you've
(03:38):
explained in many other places,it makes sense as well.
Speaker 3 (03:42):
Well, it's kind of serendipitous.
I don't think it was acoincidence and I don't think it
was necessarily by design, butI do think that when you get
technologies that happen toevolve along similar timelines,
you're likely to find thosekinds of intersections.
Now, any kind of edge computingsituation, whether it be
automotive, whether it be mobiledevices or edge of a data
(04:06):
center all this IoT stuff that'sbeen happening and just the
sheer amount of informationdoesn't necessarily have to be
sent somewhere to be processed,if it doesn't have to be.
Some of it does.
Sometimes you really do want tohave massive collections of
data to be able to do the kindof number crunching that a
handheld device can't really do.
But if you want to, let me giveyou an example of what some of
(04:27):
the things you can do.
So, for instance, in machinelearning, where you actually
have to transform the data to beable to process it, you can't
just do analytics on objectstorage, not natively.
You have to make the datatransform so that the analytics
engines can use it.
Well, what we normally have todo without any kind of
computational devices is youhave to send it somewhere for
(04:49):
that pre-processing to be doneand then you have to run the
models.
But that first step, thatpre-processing element, could
very well be done at the edge,so that you're not necessarily
sending all that data over intoa centralized location.
So one really good example forthat is facial recognition
(05:11):
software for cameras.
So if you think about the waythat that would work, you have a
centralized server that doesthe processing for facial
recognition.
The camera sends the feeds backto that central location, that
application starts to do thefacial processing and then it
has to communicate with all thedifferent devices.
However, with a computationalstorage solution, you could run
(05:34):
the facial net recognition righton the camera and then only
send back the information thatyou wanted to find, as opposed
to just sucking in the ocean andtrying to find that small
little bit that you're lookingfor.
Being able to do thatseparation of duties is a major
time saver, which of coursesaves processing power, saves
(05:56):
cooling, saves network.
Overall the long term it justdoes an awful lot of saving of
energy and economics.
Speaker 2 (06:07):
Okay, and then I mean you've been very busy so I'm
aware as well.
You've just published version Ithink 1.0, of the I'll have to
read this computational storagearchitecture and programming
models.
So can you give us thebackground on that?
Speaker 3 (06:22):
Well, in order to give vendors the general
concepts that they want to usein order to develop a
computational storage device, ithelps to be able to understand
A what it is, b what you can dowith it and C how you can access
it.
So what this does is it providesa model of computational
(06:43):
storage that allows you tocreate different form factors
that the nature of therelationship between a processor
, the storage that it uses, thememory that it uses and the
corresponding form factors comeas a result of it.
The APIs is simply a way ofhaving a programmer be able to
access a computational storagedevice to operate it from any
(07:06):
kind of host or remote location.
Now, that's different thanwhat's being done inside of the
NVM Express group.
By the way, demo will make itclear that there's a separation.
So the model and thearchitecture and the APIs are
being handled inside a SNEA andthe commands for NVME are being
done inside of NVM Express.
So the NVM Express is aslightly more narrowly focused
(07:30):
version.
It's really focusing on thedrives at the moment, not so
much the arrays or theprocessors or any of the kinds
of broader flexibility.
Like I said, we're at thebeginning stage of this, but
NVME is responsible for theirown command sets, their own
theory of operation, which isoutside the scope of SNEA.
So we're working together withthat organization to make sure
(07:52):
that everything is aligned.
Speaker 2 (07:55):
OK, and just for a little bit of understanding with
both the things we've coveredso far, is it vendors start
producing these new types, thenew ideas, and then you get
involved if you're like referee,to come up with open stands
that we're going to work to, ordo you foresee or anticipate
something coming along and thenstep in early, just for people
(08:16):
to understand where does SNEAget involved at what stage?
Speaker 3 (08:19):
Well, the ideas can come from anywhere.
There really is no one way tostart an idea, but the process
of once an idea gets done, youget a couple of other companies
or other members to think thatyeah, that's something that's
worth working on.
And then, if you get a criticalmass, they come together and
they start to identify what isit that we should do to make
(08:42):
this work for the long term, forthe long haul, I mean?
Ultimately, if you're lookingjust to have a solution that two
vendors can work together, youdon't need a standards body.
But once you get beyond that,you start to realize that I need
to make sure that if I'm goingto get a computational storage
device from vendor A and I'mgoing to get another
(09:03):
computational storage devicefrom vendor B, outside of the
special sauce that they want tohave to make it work, you kind
of want it to be relativelysimilar in the structure, like
the electric plug in your wall.
You don't want severaldifferent plug sizes, you don't
want several different plugsthemselves.
So at the very fundamentallevel you have to agree on what
that's going to look like.
(09:23):
Now that doesn't prevent youfrom being innovative by working
to build on top of that.
But in this particular case, youhave a couple of companies who
are trying to developcomputational storage models
that they're interested in, andso they get together with other
(09:43):
companies who want to do thisand they kind of hash out the
basics so that they can actuallygo onto the market and say I
have this device, and then thecustomers and the administrators
will kind of have a generalidea of what you're talking
about.
For example, in NVMe we startedoff with NVMe Express for PCIe
(10:04):
devices, and the whole purposewas that if you want to have
that kind of really fast storagedevice and you want to stick it
into a PCIe slot, you don'twant to have to rewrite a driver
for every single thing that youwant to put in there.
Well, computational storage isthe same way.
You don't want to have to havea very specific solution for
(10:28):
every single one that comes downthe road Because, quite frankly
, there are a lot of differentpossibilities and options and
you've got to have some place tostart with.
So in this particular case, weactually have 52 different
members who happen to have avery similar idea and they came
together to start working on thespecification.
Speaker 2 (10:50):
OK, and in a slightly similar vein, I see that the
DNA data storage alliance havebecome an associate member.
Yeah, technically OK.
So, and I'm just interested, Ithink there was a track on DNA
storage in your recent StorageDevelopers conference.
But again, for those of usmyself include that maybe I'll
sit too up to speed on that,what is it, and is it as
(11:12):
exciting as I'm imagining itcould be?
Speaker 3 (11:15):
I think it is personally.
I mean, I have to confess I'mno biologist, so I don't really
understand a lot of the veryspecific things or chemist, I
guess, is probably the best wayto, I don't know.
It's.
It covers a lot of technicalareas but, in a nutshell, one of
the things that's beenconcerning a lot of people is
the fact that we have just hadan explosion of information in
(11:37):
the last 10 years.
Right, we've, we've, we'vedeveloped what we've, we have
created more information in thelast three or four years than
all of human existence At themind boggling.
You know, you know, it's justone of those.
It's one of those weird thingswhere you know the numbers stop
(11:57):
me having any meaning but wherewe were doing long term
retention and archiving of thisdata, it was reasonably
straightforward.
I mean, you have tape, you'vegot the hard drives, you've got
other types of records that youcan keep for the long term
no-transcript.
But this type of data, thisamount of data, means you've got
(12:18):
to really start thinkingoutside the box.
You really have to startfiguring out.
What does it mean to have trulylong-term archiving information
that's going to last a verylong time and going to be
retrievable in the future.
What DNA storage is trying todo is say look, the DNA molecule
can actually act as a storagemedium for that long-term.
(12:42):
Now it's still in the veryearly stages.
What that means, and if youthink about it from a very basic
common sense perspective, youhave to be able to access that
DNA strand.
You have to be able to encodeit in such a way that's readable
in 40, 50 years or 100 years orwhatever it is.
You have to be able to make themachinery that can do all that
(13:04):
work.
You have to be able to identifythe standards of access, the
software that's supposed to be,or what's it going to look like?
What do you need?
Because you're not going toread DNA the same way that you
would read data off of a platter, for example.
It's not exactly the blockstorage that we're used to with
an SSD.
All of these things need to bedefined.
That's where the DNA datastorage alliance has been
(13:30):
starting off, and what Sneobrings to this is the umbrella
to be able to do that work inthe atmosphere of a storage
related context.
We can provide the supportnecessary for them so that they
can focus on the work that theywant to do.
Speaker 2 (13:49):
I'm a student of the storage industry for quite a
while and I think it was aholographic storage was out
there for a while.
I think people were storing onglass as well at one point.
I'm not sure whether those arestill going, but they seemed a
bit more pie in the sky.
Do you think this is a realconcrete thing that's going to
happen, or is it still too earlyto actually work out whether
(14:10):
it's going to have the impactthat it might have?
Speaker 3 (14:13):
Well, anytime you start talking about science
fiction-like stuff, you alwaysrun the risk of it either
petering out or hitting aroadblock that you couldn't have
anticipated.
I'm not sure what happened withholographic storage.
I thought it was reallyinteresting as a concept and I
remember they were talking aboutthis back in I think it was
like late 90s, if I remember ifthey started it.
(14:35):
And there's also molecularstorage, which doesn't sound all
that unrelated.
I can't say for sure because Ireally don't know anything about
that, but I do think that onceyou hit a critical mass of
interest and the amount ofinvestment that goes into it's
(14:56):
not just in terms of money buthuman investment and being able
to put this together, eventuallythe need will outweigh the risk
and you'll have to getsomething to work over time.
We just simply don't have thecapacity in our existing storage
media to handle the influx ofdata that's coming in for any
(15:21):
kind of long-term science,especially when you start to
think about the skill andshriveless of privacy and
retention and the need to holdon to all this data for
regulatory purposes.
So you wind up with this kindof external influence that would
perpetuate this and an internalmotivation to keep going.
(15:43):
I think that, from what I'veseen from the people and the
companies that are involved inDNA storage, that there is a
critical mass of people who arelooking to make this work, and
work successfully.
They've organized themselves insuch a way that they're
obviously in this for the longhaul.
Speaker 2 (16:01):
OK, so that was, I say, one of the main themes, I
think, at your StorageDevelopers Conference.
I think the other tracks werearound SRM or, dare I say, old
friends, storage ResourceManagement although I'm sure
there was something new to sayabout that and also cloud
storage.
So can you just give us I thinkyou've just come back from
there, but can you give us justa bit of a highlights of what
those topics brought to light,if you like?
Speaker 3 (16:25):
Well, yeah, sure.
So let's talk about themanagement for a second, because
you're right, it is a majorissue that needs to be resolved.
I think, in general, peopletend to think of management as
an afterthought when they'repromoting storage.
Because, let's face it,management really isn't sexy
unless you're an administrator,and it makes your life easier.
The end user is the one that'sreally going to get excited
(16:49):
about this kind of thing, likethe same way that an accountant
would be excited about aspreadsheet, just to make the
lives easier.
Well, management in the storageworld has been needing to evolve
to keep up with the changesthat are going on, because what
used to be the storagemanagement of choice, which was
something called SMIS, wasdesigned to be around large
(17:12):
arrays, centralized storagedevices.
Well, since that time, we'veactually moved into other forms
of storage, includingdisaggregated storage, ethernet,
attached SSDs, ebofs, thosekinds of Ethernet bunch of flash
, and you've got to be able toprovision those somehow.
So the DMTF has its Redfishspecification for server
(17:40):
management, and so SNEA has arelationship with DMTF to create
the swordfish extensions forRedfish.
So swordfish is really Redfish,but for storage.
So, if you want to manage thesenew types of devices, these
NVMe, ethernet, attached SSDs orthe eBofs or the disaggregated
storage.
(18:01):
Swordfish will allow you to dothis without having to reinvent
the wheel, especially if you'realready using Redfish for your
servers.
So if you have a storage server, which is basically a box full
of sword devices, you can managethe whole thing with one basic
management structure.
And so we just released thelatest version of swordfish as
(18:21):
an ISO specification, and thatjust happened a couple of weeks
ago, I believe.
So we are heavily promoting thecompliance testing programs for
Swordfish.
I think before we were tryingto find the overlapping time
because companies still use SMIS, but the development for us has
(18:44):
shifted over to Swordfish, butit's never a clean transition
from one to the other.
What we've done this year isthat look, this is where we
think we need to go.
We're spending a lot of timepromoting and helping people
understand how to use Swordfishto get the best management
structure for their devices.
Speaker 2 (19:07):
Cloud storage again.
That's been around for a while.
Was there anything particularlynew there, or was it just more
of the same, do I say?
Speaker 3 (19:15):
Well, I have to confess, when it comes to the
cloud storage, it came at aconflict of some other things
that I was doing and I didn'tget a lot of chance to really
spend it.
But the cloud storage is in andof itself is rather vague.
You've got the storage devices,for instance.
You've got the storage access.
You've got the storageprotocols, object stores, those
(19:35):
kinds of things.
I think that one of the thingsthat is striking the cloud
storage folks is that it's morethan just more performance.
The issue with cloud storage ingeneral is that you've got to go
through an awful lot of layersto get to the storage from the
cloud applications.
What's been happening,especially recently, is that
(20:01):
when you wind up with theseapplications that don't quite
work the way a double-arbitwants it, they just build
another layer of abstraction.
I don't know if you've spent alot of time going through all of
the different layers of how toget a single application in the
container to talk to its actualstorage, but it's impressive how
many layers there are.
(20:21):
From an application'sperspective, all they want is
more raw power.
They want more raw processingpower, and so you eliminate all
of the benefits and features.
That is built into the protocolthat's built into the devices,
that's built into the processors.
Excuse me, what we were talkingabout in part was well, how do
(20:44):
you capitalize on those?
You're just ignoring a lot ofthings that could make your life
easier, but most softwaredevelopers really don't know
about it.
We were tying the stringstogether and hopefully aren't
knots but we're trying to getthem to start paying attention
to the idea that there arethings that you can do in
(21:06):
storage that you may notnecessarily be aware of because
it's been hidden from you for solong.
I hope personally that thedevelopers and the software
authors and the people who arein the living in the container
world can realize that we'vebeen doing this for a long time
and can help.
We can actually help you getthe performance you need without
(21:28):
just waiting for the next Uberprocessor from AMD or Intel.
I think that, ultimately, whatyou want to do is you want to
work smarter, not necessarilywork harder, and that's where
the cloud storage stuff reallytends to shine.
Speaker 2 (21:43):
Okay, it sounds as if you're very busy as an
organization, but are there anyother things on the road map we
need to be aware of at thismoment in time, or with what
we've discussed?
Is that keeping you pretty muchfully occupied?
Speaker 3 (21:56):
Well, as a chair, I have a desire to see the
organization expand and grow interms of the strength of its
quality of offerings, notnecessarily become bigger, but
definitely become much moreresilient and reliable for those
(22:17):
who are depending upon us.
So, for example, we're talkingabout the cloud storage stuff.
One of the really, reallyinteresting things that's
happening is something calledSDXI, which stands for the Smart
Data Accelerator Interface, andeffectively what it does at a
very high level is it actuallygoing back to the cloud storage
conversation?
It does that very thing that wewere talking about with the
(22:39):
application being able to usethe smarts of the hardware.
So let's say, for example, ifyou were to, let's say you have
a storage VM and a regular VMand you need to move the data
from one back to the other.
In an environment with a lot oflayers of abstraction, you
effectively have this Pachinkochip that has to go all the way
down to the hardware and all theway back up to the other
(23:02):
virtual machine.
But with SDXI, what you can dois you can tell the hardware,
using a strict API, toeffectively do the data movement
from one memory location toanother privileged memory
location.
Once you set up the privilegedrelationship.
It'll just happen because thehardware can do it for you.
You don't have to go throughall of these layers of APIs to
(23:24):
get that layer to do this andthen that layer to do this and
then pass the copy down all theway down to the hardware and
then back up again.
Just tell the hardware to moveit, hardware moves it and you
can really speed up andaccelerate that data movement.
Let's say you want to zero out alocation, a memory location.
You don't need anywhere.
(23:45):
These containers are ephemeral.
You want to zero it out andmake it nice and secure and
start all over from scratch.
Tell the hardware to do it.
Don't just have to go throughand cycle through the software.
To have it done in that fashionCan save up an awful lot of
time.
There's work that's being doneinside of SNEA to help assist
that in a vendor neutral way, sothat you're not dependent on
(24:06):
any one particular type ofhardware vendor, any type of
software vendor.
It's all basically part of anopen standard for that kind of
data memory movement.
That's one example what we'vealready talked about Swordfish,
we talked about computationalstorage.
We've talked about some cloudsecurity and data privacy are a
(24:28):
big part of it.
I think energy efficiency is aswell.
So one of the things that'shappening all over the world is
the movement of governmentregulations and, as most people
know, either intuitively orthrough experience and if it's
from experience, I share theirpain but government regulations
(24:52):
are not always consistent and sooftentimes you'll come across
very strict rules that conflictwith each other.
So one of the things that we'retrying to do, especially with
the energy efficiency stuff andespecially with the data privacy
and protection, is we're tryingto make that approachable for
(25:12):
people.
So we have something called theGreen Storage Initiative, which
allows companies to runcompliance of their machinery
against the energy starrequirements that are here in
the United States.
There are some EU and UK mostlythe EU, not more than the UK
(25:33):
work that is being done forhandling data location, where
the data actually gets stored,how it gets stored, what the
energy needs to be for those.
We're working with thosedifferent governments as well as
the United States, to make surethat it's easy for vendors to
comply, or easier.
They'll always build a moreinteresting regulation.
But those are the kinds ofthings that we're doing that
(25:55):
aren't necessarily sexy, but thepenalties could be rather
severe.
Speaker 2 (26:01):
Yeah, okay.
And yet mention of energyefficiency.
I mean, the world is headingtowards net zero at varying
paces depending where you are,what industries, what, I suppose
, responsibility do you thinkSNEA has as an organization?
You mentioned the energyefficiency, but that sounds as
if, at least at one level,you're responding to strict sort
(26:22):
of government requirements.
But more generally, theindustry, are you helping them
move towards that sort of target?
And in terms of data, I knowobviously your vendors are there
to provide storage and there'smore and more data as you
discuss being producedridiculous amounts.
Do you think you have any sortof requirement, obligation, to
(26:46):
start some kind of debate orenter into the debate as to
should we just be storing allthis information forever and
ever, or do we actually need tostart thinking about the life
cycle of data a bit moreseriously?
Would you just some thoughts onwhat your association's role is
in the sustainability debate?
Speaker 3 (27:02):
That requires a shift in our focus a little bit, and
it's a shift we've been talkingabout.
So from my perspective, I havealways taken the stance that, as
the quote unquote storageexperts, I think we have a moral
responsibility for being ableto lend that experience into the
world, where these types ofregulations can have deleterious
(27:24):
effects on whether or notyou're storing the data or
whether or not you're storingthe data properly.
And I do think that differentgovernments may have the best
intentions, or they may not younever really know but a lot of
them are making decisions thatdon't make an awful lot of sense
from a technical perspective.
And I think that SNEA is anorganization that has a moral
(27:47):
duty to be a friend of thosecourts, to be able to offer some
recommendations and someguidance.
We're not a lobbyistorganization, so we don't go
onto K Street and we don't meetwith anybody in the governments,
but we do send out statementsand white papers and position
papers for certain things, whenprobably we've done a few of
(28:08):
them this summer, as a matter offact, for some of the security
things that have come out withregards to some of the changes
that have been going on Forstandards.
It's really kind of funnybecause you have this
overarching umbrella ofaccountability, right?
So when you talk to somebodywho's in standards, they get
(28:31):
very mired in the word shall orthe word should and it seems
like minutia to the layperson.
But that becomes the basis fora lot of this regulation, right?
So if the regulation says youshall do something, or you
should do something, or you mustdo something, you're never
(28:52):
gonna see something that's like,well, you may do it.
It doesn't really quite work outvery well when you start using
the word like yeah, you can oryou may find the language every
once in a while slip through,but very rarely.
But our role is to make that benice and clear, right?
Because if the governmentresolves and the government
(29:14):
needs to know what's doable, youcan't just say do this and have
it be an impossibility.
At the same time, when they dopass these kinds of regulation,
we need to tell our members thisis what it means, this is how
you'd actually start applyingthis.
So we act as kind of anintermediary of information.
And no, I have to confess it'snot sexy, it's not the kind of
thing that attracts a lot ofpeople, but it does have a
(29:38):
fundamental purpose that needsto be an ongoing solution.
Speaker 2 (29:46):
And out of the pandemic.
I mean sustainability.
I'm not saying it came out ofthe pandemic, but it's become a
bigger topic.
So, similarly, the supply chainhas come under sort of intense
scrutiny and some people, Iguess, maybe hoping, once the
situation in Eastern Europe etcetera, calms down, things will
go back to normal, whatever thatmight mean.
(30:07):
But do you think supply chainsyour members are already on need
to start re-evaluating wherethey're getting their resources
from?
Or, as you say, do you think ifthey're sitting out for another
six months, everything's backto normal?
Or do you think it's along-term issue that needs
addressing?
Speaker 3 (30:27):
I would caution anybody into thinking that they
can wait and expect things to goback to normal, Because what
constitutes normal nowadays?
It's not the way it was twoyears ago.
We have the geopolitical impacton supply, for data and
communication has radicallychanged in the last three years,
(30:48):
so the Eastern Europe situationis actually a little bit more
dire than people realize.
There are a lot of softwarehouses in the Ukraine.
There's a lot of informationthat are coming out of that area
of the world.
Even Russia itself has offeredconsiderable amounts of positive
(31:09):
technology contributions intothe industry.
There's China, there's Taiwan.
There's the global impact ofthe tensions going on there.
We still have some of thesanctions on Chinese companies.
In the United States we havenew regulations about what can
(31:30):
and can't be sent into and outof the Chinese territories.
So that's not going away in sixmonths.
I think anybody who thinks thatit's gonna take a six month
hiatus and then have everythinggo back to perfection is
probably fooling themselves.
What I think we have to do iswe have to.
(31:50):
We can't just try to accept allthis in one lump sum.
We have to be vigilant, excuseme.
We have to be diligent andvigilant.
Say that three times fast.
We need to be aware of what'shappening in the tensions in
India and Pakistan, and all ofthese things have a much greater
(32:14):
impact all the way intoPoughkeepsie.
I mean, it's one of thosethings or leads.
Either way, it's going to beimpacted somewhere, whether it's
going to be at the servicelevel, it's going to be impacted
at the price level, it's goingto be impacted on the regulation
level.
I think one of the other thingsthat's happening too is the
(32:37):
content of the data is startingto get more scrutiny.
There are things that arehappening with unelected
governmental, non-governmentalorganizations that are starting
to make people question whatkind of data can be sent to whom
and when.
Who should be able to have it,who shouldn't be able to use it?
(32:59):
Those all have really profoundimpacts.
So I honestly think that wedon't have all the answers.
I'm not pretending that we do,but I do think that you need to
have somebody.
You can ask a question tosomebody who is also taking
ongoing attention to what'shappening, Because if you're not
(33:21):
careful, you could wind upbeing on the business end of his
event, and I don't thinkanybody really wants that.
Speaker 2 (33:30):
Okay, and then maybe, just as we finish, the storage
industry more generally.
I mean, I've sort of followedit on and off since the early
days of Fiber Channel, which isa wee while ago, and there's
plenty of names that were like,if you like, the bedrock of the
industry then and no longeraround.
It's a classic, I guess,technology that there were a lot
of people in early.
(33:50):
Then it's consolidated.
Do you worry at all that theconsolidation has perhaps gone
too far and that the gene pool,if you like, is too small, so
there's not enough innovation,because whenever somebody does
something clever, it seems to behoovered up, dare I say, but
one of the large organizers.
Or do you think the industry isvery healthy and, as we
discussed earlier, there aresome innovative ideas going on
(34:12):
and therefore all is good in thegarden?
Speaker 3 (34:16):
Well, I don't know how good the garden is, but I
think that if you look at thehistory of technology, you'll
see that there's a bit of anaccordion effect.
Expansion and contraction hasbeen normal and it's been going
on ever since the days of Univac.
So if you follow through thehistory of any technology all
(34:40):
the way back, if you want to goall the way back to the
telegraph, it's exactly the samething.
The history has been consistent.
Same thing with hardware andsoftware.
There was a period of time nottoo long ago, when people said
that hardware was completelyirrelevant.
All you needed to do issoftware.
And now, where has really theinnovation come?
Through Hardware, we now havethe universal chiplet consortium
(35:05):
that started out.
We've got SNEA, we've gotEnviom Express, we've got CXL,
which has really taken theserver world by storm.
You have all of thesetechnologies that are
hardware-oriented,hardware-related.
That will make the softwarebetter and then the software
will improve as well.
And then the software will getbetter and we'll start to
(35:26):
eliminate some of those layersof abstraction, because that got
too big.
The software component is.
I mean, if you would just lookat a stack diagram of all the
latency, software is huge andpretty soon.
People want to get betterperformance, so they reduced
that software and they startfocusing on the hardware.
The accordion starts to comeback.
(35:46):
As long as the accordion ismoving, we're healthy.
It's when people start toartificially try to stop it,
then things get into trouble.
And from SNEA's perspective, weare perfectly happy to play the
accordion.
We are happy to go ahead andencourage the development where
(36:07):
it needs to be, like the DNAdata storage lines.
Perfect example.
That is real, true, innovativethinking is truly thinking
outside the box and literally asit turns out.
So I think that there's alwaysgoing to be somebody who wants
(36:27):
something more or somethingbetter, something more efficient
, something less costly, andthey're going to have an idea to
do it.
Some of them will win, some ofthem won't, and by winning I
mean just be used.
It doesn't have to be the best,it doesn't have to be the
greatest, it just has to besomething that people are
working on.
From our perspective, we'reperfectly happy to be home for
(36:48):
those kinds of things, andthat's, I think, where we're
going to be focusing in thefuture.
Speaker 2 (36:55):
Okay, and then maybe just finally.
Finally, in terms of the enduser, I always feel slightly
sorry for them because they'regetting bombarded with all these
great new ideas and thingsgoing on.
They're one day being told tobe on the premise and then the
next go in the cloud and all therest.
I know it's very difficult tolimit yourself, maybe, to one or
two, but if you had a couple ofpieces of advice, when users,
(37:16):
faced with this enormousportfolio of storage
technologies and stuff, where doyou think they should
concentrate to get the quickestresults?
Or what do they need to belooking at, do you think, to
improve their storageperformance?
Speaker 3 (37:31):
Well for all.
Number one don't panic andbring your towel.
Number two SNEA has been doingan awful lot of work over the
last several years to try tobreak a lot of this down into
plain English.
So we have probably a libraryof hundreds of hours of plain
English content that they can goto by going to the website or,
because we have a YouTube video,you can go into YouTube and go
(37:53):
to our YouTube channel too.
It's vendor neutral.
It's technology neutral, whichmeans that there's no promotion
of any one vendor over anotheror anyone technology over
another.
So if you want to find out aboutthe pros and cons of Fiber
Channel, which is still verystrong storage technology, you
can do that.
If you want to find out aboutXPUs, which are brand new and
very confusing to a lot ofpeople SmartNix, computational
(38:15):
storage these are just some ofthe more recent ones, but we
also go back to the basics.
If you don't know what RAID isor what RAID means, if you're
not sure what an initiator or atarget is, you just want to get
started.
We have those kinds of thingsas well.
We have a series calledEverything you Wanted to Know
About Storage, what we're tooproud to ask.
So, everything from theabsolute basics to the very new
(38:37):
complex topics, we do an awfullot of education that is just
freely available for people.
All they have to do is look,and so we're going to continue
to do that.
It's a major pillar of ourmission.
Speaker 2 (38:51):
Okay, well, I've taken up a lot of your time and
I really appreciate it.
Give us some fascinatinginsights, so thank you very much
indeed for being with us.
Thank you.
Speaker 3 (38:59):
Thank you for the invite.
Speaker 1 (39:01):
Thank you for listening.
For additional information onthe material presented in this
podcast, be sure and check outour educational library at
sniaorg.
Popular Podcasts
Cold Case Files: Miami
Joyce Sapp, 76; Bryan Herrera, 16; and Laurance Webb, 32—three Miami residents whose lives were stolen in brutal, unsolved homicides.  Cold Case Files: Miami follows award‑winning radio host and City of Miami Police reserve officer  Enrique Santos as he partners with the department’s Cold Case Homicide Unit, determined family members, and the advocates who spend their lives fighting for justice for the victims who can no longer fight for themselves.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.