Cloud Terminology Part II

In this episode we dig deeper into terminology and concepts around cloud computing. Learn more about not only what things mean, but why you should care.

Episode Transcript

Speaker 1 (00:02):
Welcome to Azure for the win for mostly Azure focus
cloud podcast.
Brought to you by blue silvershift.
You're like to get together on aFriday, talk about cloud
technology.
Usually have a few drinks, justtry to have a few laughs and
have some fun.
Hello everyone.
Welcome to another episode ofAzure for the wind podcast a.

(00:22):
We have your blue silver shiftteam here.
Uh, we have your moderator,myself, Mark Volsan.
We also have the partners, JohnDobson.
Hello everybody.
And quick Slack.
Hello again.
All right, so this is actually apart two, uh, of a, uh, two
partner that we, um, have been,have been working on.
Uh, so we did part one a whileback actually.

(00:43):
So I think it was, it was lateAugust, so we're, we're into
late September now.
Uh, you know, summertime peoplego away on vacation trips like
that.
It's kind of act to school backto school's caught of the way,
but we're, we're back at it andwe're hopping to a part two.
So this is part two of cloudterminology.
And, uh, we caught a bunch ofnew terms.
We're going to be going overtoday.
Uh, so we're going to be, we'vegot a bit of a list of been

(01:04):
debating about the order.
We're going to go through them.
Uh, it's gonna be mostly Craigand John going through these,
these terms and I'm going to bepopping out some new ones.
I actually have some on the listthat I actually am really
curious about cause I don't havea really good definitions
myself.
Um, so I'm looking forward tothis before we jump into the
meat of our podcasts as perusual.
We do have a couple of drinks.

(01:26):
It's a Friday afternoon here forus.
So we're going to be a enjoyingcouple drinks.
I have some scotch and uh, thisis, uh, when we've had before.
[inaudible] you wanna you LISlay it on.
It's only because it's got myname in it.
Craig Aleke.
It's from a Speyside.
Uh, any regular listeners wouldknow that.
That's one of my favoriteregions.

(01:47):
Um, and I'm going to have thesame.
So I will take that.
And John has his, a usual, Ihave my usual and any regular
listeners will know what that isand for those.
And what is that for those nonregular listeners?
What are you, it's um, Ryan Coke.
I think you have just plain ryelast time.

(02:08):
Just Brian ginger, don't youknow I do Ryan coat every short
glass with that regular Coke.
Just like Julian[inaudible]trailer park right there.
Your ship.
He does happy Friday.
Okay.
So before we get into this, I'msurprised that uh, all of us are

(02:29):
here.
You know, none of us are down inNevada for the, uh, storming of
the area 51.
It's happening today.
I believe 75 people turned outis what I heard.
I was supposed to be 2 million.
I'm really surprised, but Ihonestly thought that was going
to be a couple thousand peoplethat we're going to show up.
You know, I didn't think theywere actually going to try to
storm area 51, but I thoughtthere was gonna be a lot more

(02:49):
people in the desert.
Werrington foil hats.
There are still that may peoplein the desert wearing tempo
hats.
They're just not at[inaudible].
Exactly.
The burning man.
They're the real sites.
That's true.
Yes.
I see.
I saw some videos with someNaruto running going on.
Um, there's a term for you andtheir Ruto running.

Speaker 2 (03:08):
Okay.
I don't even know that one.
So you're going to educate us atthe end of the podcast.
You want to be able to know,we'll record it later.
I do the outtake.
What has never rude or runningor do route or running is, um,
let's get into our terms.

Speaker 1 (03:22):
Sure.
So, um, cloud terminology, wewent through a bunch last time
we talked about private versushybrid versus public cloud,
multi cloud IaaS path as drowseas all those cool things.
So, uh, if everything I justsaid to you is jibberish go
check out episode one, um, we'regonna be jumping around a few

(03:42):
different topics today are a fewdifferent terms, but the one I
wanted to start off with wasactually one that I find a lot
of people have a hard timedescribing.
Even Microsoft has a hard timedescribing, I've had people ask
me what this word means manytimes.
And I feel like no one everreally has a very good answer.
So let's, let's hear.
We got some really good answertenant and we're going to John

(04:02):
first on this.
What is a tenant?
So

Speaker 3 (04:05):
tenant tenant means different things to different
people in different situations.
Uh, from a business standpoint,um, you are renting space on, uh
, in a variety of ways fromMicrosoft and Azure.
Um, and you're renting, um, youknow, a server, um, a virtual

(04:27):
server, a PaaS service orotherwise within their space and
their space is shared is, is adata center or a different data
centers.
Um, and a tenant is your areainside of there a space and a
tenant in the active directoryterms is, um, you call it an

(04:50):
active directory tenant becausethat is your security boundary
around your area as it relatesback into a subscription.
So if you think about anapartment building, it's like an
apartment building and you, youget one suite within that
apartment building that is, youare a tenant inside of that part
of the building.
So in reality it's your, thetenant, the business is the

(05:13):
tenant consuming the servers.
But, but Microsoft has laid outthis area for you.
Okay.
And that's actually thesubscription.

Speaker 2 (05:22):
Uh, no because you can have multiple subscriptions.
[inaudible] talk aboutsubscription in a second.
But I like the analogy of anapartment because like an
apartment, you've got walls andyou're, you're sharing a wall
with your neighbor but you can'twalk through that wall.
So there is that, that's a, ina, in the apartment context of
physical boundary.

(05:44):
In the cloud con you all havethe virtual Bauer, same water,
right.
Same, uh, eat grass and in grass.
Yeah.
Elevator building, getting inand out of that.
Yeah, you're right.

Speaker 1 (05:54):
It's actually not far off from a description of gave
off the top of head, but it's alittle bit, a little bit
different.
I think yours probably makes abit more sense.
When I described tenure tosomebody, once I described it as
the tenant is your PO, youpurchased a plot of land.
So your tenant is your plot ofland.
You can build anything you wanton it.
So you can, you can actuallybuild multiple structures.
And I guess to me those would bethe different subscriptions that

(06:16):
your tenant is your plot ofland.
Yeah, that makes sense.
Yeah.

Speaker 3 (06:19):
And, and you want to know what the other thing that's
applicable in your analogy isthe mom and pop shop that sells
milk on the corner can have anAzure tenant and general
electric can have a tenant andthey can be on the same, they
could be a physicalinfrastructure, but you will
never know.
But it could be, right?
Yup.
They're sharing that.
So, so

Speaker 2 (06:38):
the cloud is all about shared services, shared
infrastructures, share like, Hey, you know, you're, you got the
masses that are contributing to

Speaker 3 (06:47):
yeah.
Lower costs because it'soverall, so you say terms like,
um, that is my Azure tenant.
[inaudible],

Speaker 1 (06:53):
what does AWS, what is he equivalent AWS
terminology?
Tenant does it tenants.
They have subscriptions as well.

Speaker 3 (07:00):
Yeah.
So, but a tenant could havemultiple subscriptions as you
said.
Yes.
So that's leads into the nextone,

Speaker 1 (07:08):
which is subscribe.
Well yeah.
So, so I'm sure[inaudible] havewe covered that or does that
mean kind of not really.
No.

Speaker 2 (07:13):
So if we take that boundary concept further, the
subscriptions can be the roomsin an apartment.
So you've got your, yourwashroom where you go and have a
shower, brush your teeth orwhatever you use.
The services that you do in thatroom are contained like there of
one type of service.

(07:34):
Uh, now their subscription mightbe your bedroom, you go to their
, to go to sleep or whatever.
Right.
So subscription is a, anotherboundary of resources.
And you know, you put these likeresources into a subscription.
Um, and uh, you know, we won'tgo further into talking about
resources cause there'sdifferent strategies.
There is, there's kind of ahierarchy of the tenant is

(07:57):
first, then the subscription andthen resource groups.
Then the resources.

Speaker 3 (08:01):
Yeah.
And policy gets applied in thatorder as well.
Yeah.
And uh, the only thing I wouldadd in for from my perspective
is a good thing for thelisteners is, um, a subscription
is a, um, top level area of costaccountability.
So, um, as an example, if youare inside a company, everyone

(08:23):
tries to go to one subscription,you could actually slice up and
say, this department gets thatsubscription, this department
gets to this subscription andthey can all be inside the same
as your region and interact withno extra egress charges.
But the point is, is, um, youcan have many[inaudible]

Speaker 1 (08:40):
would you say that it is typical for the relationship
to be one-to-one for company andtenant?

Speaker 3 (08:46):
No, no, no.
Depends on the size.
You would have production anddevelopment subscriptions if you
want to keep them separate thatway because remember, you could
that, um, the developers canaccess production and uh, they
can only access development asan example.
Or maybe you would alloweveryone into it, but you then
you would give them differentrights.
Some would just have readaccess, others would have read.

(09:07):
Right.

Speaker 1 (09:08):
Okay.
So[inaudible] prescription.
So would you have like I, I,you'd have multiple strip
subscriptions but you could havemultiple tenants as well.
I was trying to, I'm justthinking about multiple tenants.

Speaker 3 (09:17):
Yeah.
But it tenant goes back to thebilling entity.
Really.
Okay.
So all subscriptions are have tobe paid by a company and the
tenant,

Speaker 2 (09:26):
the tenant also in, in the Azure contacts anyway.
Purity bounced.
Security boundary.
Exactly.
So there's the, the Azure activedirectory is tied at the tenant
level.
You can have as your activedirectory multiple as your
active directory is thoughseparate.
You can, but that's a differentconcept.
We won't get into that today.
But, um, the, it's a securityboundary and you have a users,

(09:48):
uh, or or servers, accounts orwhatever at the tenant level
that can be granted access toresources, subscriptions and so
on within that tenant.
So it's, again, you've got thekey to get in.
So you're getting into yourapartment, you opened the door,
everything in there isaccessible, but you may not be

(10:08):
able to open.
Like your girlfriend may have akey to, you know, her jewelry
box that you can't get into.
Well that's another boundary.
I agreed.
Agreed.

Speaker 3 (10:19):
Um, that's a good idea.
And also on that, I just wantedto add that, um, most
enterprises, the goal though the, the Valhalla or Holy grail or
whatever you want to call it, isum, to have one authenticating
authority, one act of directoryor one otherwise ability to
actually get in there.
So you would have one tenantgoverning them all.

(10:43):
Okay.
That sounds like a saying like

Speaker 1 (10:46):
I have just one new friend, one tenant and then them
.
And then there's also

Speaker 3 (10:53):
a M a VPN, a Piering, uh, principle vignette period.
The net peering can happen atthe tenant level as well.
Right.

Speaker 1 (11:01):
I crossed subscriptions.
We should come up with some, I'mgoing to come up with some good
Lord of the rings cloud of memesnow.
Okay.
That's a good challenge.
Yup.
Okay.
So you said he's, John's rollinghis eyes.
I don't like word of the runslike a, excuse me.
Uh, I think you mean star Trekand star Trek.
Yes.
I'm going to do some star Trekand come on some star Trek cloud

(11:22):
memes, although I feel likethat's too easy.
They write themselves, tryharder.
They do.
Okay.
A locator.
So you said something, actuallythat was a little bit down our
list.
We'll jump into into that.
Um, one of John's favorite topictopics.
Ingress versus egress, right.

Speaker 3 (11:39):
So, um, the idea is on-prem on premises or in the
cloud, you have to have aningress solution or plan.
Let's just leave it at that back

Speaker 1 (11:50):
up to define ingress.
Yeah.
Well, pardon me.
So that's the way

Speaker 3 (11:54):
people use the term is to actually, I can't use the
term[inaudible].

Speaker 1 (11:57):
Yeah.

Speaker 3 (12:00):
Um, so you have to have a plan for inbound and
outbound network traffic.
So ingress is in inbound trafficand egress is outbound traffic
and the inbound has manydifferent variations.
So you could have a web serveron the front end of your system.

(12:21):
You could have um, uh, peopleRDP being in bound to your, to
your core of your networkthrough an inbound public
service point.
And you have to actually come upwith ingress style solutions for
that.
And conversely, egress.
And you can add onto this Craig,uh, yeah.
On or do you want to add onto,

Speaker 2 (12:42):
I was going to add on the ingress.
Yeah.
So things that you need toconsider on ingress is, you know
, protection.
So firewall, yeah.
Web application, firewall, thosesorts of things that are going
to forms and network securitygroups can play a role in that
too.
So you behind, so

Speaker 1 (12:59):
yeah, you're literally, you guys are
literally listed over, but we'lldefine those later.

Speaker 2 (13:03):
Find those later.
But, but you've got to protectthe ingress traffic.
So if you open up everything tothe world, you're going to get
hacked and you know, whatever isgoing to happen.
But if you, you know, protectthat ingress traffic, that's
number one.
So no[inaudible]

Speaker 3 (13:19):
he, uh, ingress and uh, that's a good point.
You have to protect thattraffic.
You've got to look in the streamand protected the available
number of ports, as Craig said,cause everyone and their dog has
a scanner.
Now, um, the egress is acontrolling the outbound access
and that's really to control.
Um, if you do get compromised orif you've got a process that is

(13:42):
compromised outside your system,that your system will go and
bring it in to you or do youmean?
Well, um, you, you actually have, uh, processes where there's a
server inside that's going toget an update or something or a
DNS spoofing issue that happensinside and your servers will go
through if they, if they justhave an, um, not tethered but

(14:05):
on, you know, the, if nothing'srestricting it and nothing,
nothing's a reverse web proxythey call it.
Um, what happens is, is thatservers can talk to anything out
there and those things can bedirty or compromised and bring
traffic right back in or badcompromises right back in.
So you have to have an ingressand egress plan is my point.
And a lot of people don't eventhink about it.

Speaker 2 (14:28):
And I don't think we need to go too far into this
because if I recall correctly ona previous podcast, we've talked
at detail about ingress andegress.
Maybe not.
Maybe we did.

Speaker 1 (14:38):
I dunno.
Well, it's all good if we don'tthink we did on the terminology.
Yeah,

Speaker 2 (14:43):
no, not on the terminology.
No, no.
Yeah.
But if we didn't, we'll listento our previous podcast.
If we didn't, we'll do a wholepodcast on John's topic.
Ingress and egress.
Oh, I did a M as your Fridays,an internal thing that we have
on it on ingress.
So that's true.
Maybe I can ever do one onNegress cause it's boring.
Well the other thing with Egrass is if you've got a

(15:05):
multi-cloud strategy, I'll addthis, right?
So it's a something forcompanies you consider when you
in the public cloud realm, youdon't pay for ingress traffic,
the amount of bandwidth you'reconsuming data coming into
Azure, AWS, GCP, whoever.
You're not paying that for that.
That's free.
They want all your data.

(15:25):
That's okay to get data out orany, any traffic that's going
across the wire out to theinternet, whether it be to your
own on-prem or to another cloudvendor because you got a hybrid
or multi-cloud strategy thatcomes at a cost.
So you actually have to pay forthat.

(15:45):
There are things you can do tominimize those costs and we
won't get into those in thistopic.
But a, uh, that point issomething to consider.
ER, ER, ER, egress charges canbe excessive.
It's like the reverse of theLangdale ferry

Speaker 1 (16:02):
[inaudible] to go over, but it's free to come
back.
Gotcha.
Our listeners

Speaker 2 (16:07):
in Australia won't actually understand what that is
.
So it's like taking a ferry fromthe Sydney to a, uh, you don't
take it to Kedzie beach, butwhatever you're, yeah, there's
no barrier that way, but

Speaker 1 (16:18):
all right.
So in other words, the ferrythat you only have to pay one
way because they know that youcan't get back.
They just want to get here.
Yeah.
Okay.
On the way back.
I got ya.
All right.
So, um, I'm just gonna go inorder cause you got, you guys
mentioned a whole bunch ofthings that are on this list
down below.
So I'm just gonna go to the nextone.
So network security groups,verse firewalls.

(16:40):
Okay.
And John is pointing that.

Speaker 2 (16:42):
Yeah, we kind of did touch on this already.
So very briefly, networksecurity groups in the Azure
context.
And uh, you know, this is prettyuniversal, but it is, it's not a
firewall.
Some people think it's afirewall, some people think it
will protect their virtualnetwork within Azure in the
cloud.

(17:03):
Um, the thing is it willactually, it's just, it's
another security boundary, butthere's no, there's no logging.
There's no, there's minimal, uh,controls you're controlling at
the port level or IP addresslevel to say we allow this
traffic from this sourcelocation or source IP or

(17:23):
anywhere in the world.
Yeah.
To this destination point orthis entire sub net or whatever.
[inaudible] it's essentiallyjust a, it's a port forwarding
port, um, um, control rule.
If you mess up those ruleswithout a firewall in place,
you're opening yourselves up andyou may not know because it's

(17:44):
very difficult to look atsomething and sometimes you can
have, you know, you're allowedup to 200 rules, NSG, uh, and
some of those rules are going tohave multiple IP addresses and
ports associated with them.
So it can be become veryunmanageable and due the lack of
logging on it and so on.
So versus firewalls, you saidnetwork and SGS, network

(18:06):
security groups versus firewall.
So firewalls provide thatprotection.
They've got the logging builtin.
You know, if you're using one ofthe, uh, the builtin, like in
Azure or you use the applicationgateway, which has a laugh and
it has other features.
But there there's capabilitieson top of that that are like
machine learning.
Microsoft is designing all thebehind the scenes looking for

(18:28):
common threats across the worldof things that are coming into
their data centers and they'lltrigger things based on, uh,
what they're seeing.
So that's my, my thing.
So John, you're, you're going toadd to that, um,

Speaker 3 (18:45):
on NSG.
So we also use an SGS, uh, anetwork security groups inside
the network to further partitionit.
And so between a front end and abackend network or front end
network would have web serversthat face customers, a backend
network would have a databasehas or I as or other or S or

(19:06):
even SAS.
And that backend network wouldonly talk to the front end vs a
strict port and we would alsolimit it to that web server that
was being able to talk to it.
So we use a network securitygroups that way.
Yeah.
As that control a control point.

Speaker 1 (19:21):
Okay.
I'm going to switch it up alittle bit.
Huh.
I'm going to jump over toanother section here.
Okay.
I'm going to, let's talk alittle bit about, well,
financial impact.
There's, there's actually a fewterms here that relate to each
other, um, that are aboutfinancial impact.
So, uh, and this is somethingthat I look at, you know, every

(19:42):
once in a while when we're doing, um, estimates for, uh, for new
clients around theirconsumption.
So pay as you go versus reservedinstances.
And also I'm going to addelasticity in there cause it
kind of relates, right.
So, um,

Speaker 4 (19:58):
yeah,

Speaker 1 (19:59):
talk to me about pay as you go first reserved
instances and in conjunctionwith elasticity.

Speaker 2 (20:07):
Okay.
Uh, so can you think of pay asyou go as just the retail price
then MSRP manufacturer'ssuggested retail pricing.
So you're, you're, every publiccloud vendor has their retail
price lists, they've got pricingcalculators.
You can go in and plug innumbers and figure out what it's
gonna cost to run a specificresource, um, with certain

(20:31):
inputs.
So that, uh, is the pay as yougo price and

Speaker 1 (20:37):
we're talking about, so just to kinda frame this
further, okay.
Okay.
Sorry.
Actually.
Yeah.
So that there,

Speaker 4 (20:44):
right

Speaker 2 (20:46):
to take it to the next, I was more thinking

Speaker 1 (20:48):
of, um, the concept that you pay for only what you
need.
I think that's where you'regoing, right?
Yeah.
So it will also say like, what,what is you're paying for?
So you're paying, so this, we'retalking about like the annual
costs, like the, or sorry, themonthly, first per minute

Speaker 2 (21:02):
per second costs.
Right?
So that's where our billing getsdown to is they've got these
systems that are actuallymonitoring and tracking your,
your spend, um, right to the,uh, to the second in some cases.
Um, so you, you pay as you go.
So the beauty of pay as you go,as you're, you're not committed,

(21:22):
you're not going out and buyinga$50,000 server and depreciating
that over five years.
You're like, Hey, I want to do atest and I don't know how much,
if I'm going to need thislongterm.
So you, uh, you just try it out.
Okay, it didn't work.
Okay.
I don't need that, you know, 16core system or whatever.
Um, so that's the pay as you gopiece.

Speaker 1 (21:44):
All right everybody, we had a little bit of a battery
malfunction on a, on John's aheadset, but we're, uh, we're
back now.
So John's just trying to, tryingto figure out where he was.

Speaker 3 (21:53):
I know where I was, I was on capex versus OPEX.
So previously people used to buyon premises servers and now the
pay as you go term is reallysynonymous, synonymous with op
ex.
And that means that, um, you canrapidly experiment as Craig
suggested, with turning on aserver, doing a test and then

(22:14):
destroying it.
And the pay as you go means thatyour, your monthly cost is
entirely variable.
It's on what you need.
So it could be linked to demandinbound or on what you consume
instead of paying a set rate.
And everyone loves that becauseit feels like you have control
back.

(22:34):
So since we're talking aboutterms Cappex and op-ex,

Speaker 2 (22:40):
okay, yeah sure.
I'll put all expense.
Yeah.
So capital expense is somethingthat you buy and it has value
over time that depreciates ordiminishes over time that value
say a$100,000 today in fiveyears it's down to zero.
So you, you, you know there'sdifferent types of a pre
depreciation but the easiest oneis a straight line depreciation.

(23:03):
So in a$500,000 investment overfive years goes down to zero.
After year one it's going to be$80,000 after year two at
$60,000, so on and so forth.
So it's a straight line justyeah, that's, that's a
depreciation.
But yeah, the accounting conceptis that you are buying something
upfront that you can't expensethat full a hundred thousand

(23:25):
dollars your as a company, itjust doesn't make sense to do
that.
Um, there's rules around it.
So, cause some companies coulddo that to get their costs down
significantly or their, theirtax liability down
significantly.
So there's TA accounting rulesaround how you recognize these
capital investments.
And so a cap cap backs costs issomething that, um, you, you

(23:50):
spend real dollars today but youactually are not expensing it
all at once.
You're expensing it over aperiod of time, whatever that
useful life is of that asset.
Whereas op ex

Speaker 3 (23:59):
hold on.
So I just want to say one thingabout cap ex.
Um, cap ex forces you to buy allthe capacity in one shot and
you've got to buy it in the sameincrements or bigger or like,
whereas op, uh, the pay as yougo means, Oh, I'll just take it
in incremental chunks.
Yeah.

Speaker 2 (24:16):
Sorry.
And you're, you're making anassumption of what that future
need is going to be.
So you're saying or you're usingit or not.
Yeah.
And it's wasteful.
It is wasteful.
That's the problem.
Yeah.
So OPEX is the, if you want tosay the exact opposite where
it's, um, it's an operatingexpense.
So as soon as you spend it, it'san operating expense on your
books and it goes out.

(24:37):
And the accounting litmus testfor that is, is there anything
useful left at the end of the,what you produced or what you
spent money on.
So if you, so in the cloudconcept, this is what
accountants struggle with allthe time is um, we get it from
our customers is, you know, we,how do we, how do we recognize

(24:59):
this on our books?
So it's now you, you pay for theconsumption that you used in the
last 30 days.
So you get your monthly bill forAzure consumption.
There's no useful life.
Like you've now spent thatmoney, there's nothing left that
you could turn off all thoseresources, delete everything.
I'll, all the storage, you'vegot nothing left.

(25:19):
So that is an operating expensecause it was okay, we had to
spend that money, it spent done,there's nothing lapped.
Whereas capital, you've spentthe money, you could turn off
the server, put it in, it's gotsome useful ass value so you can
put it in the closet or whateverand dust it off.
And then pull it out later.
It has some useful life.
So that's very simplistic viewof capex versus OPEX.

(25:42):
Yup.
Okay.
I didn't think we had that onour list part.

Speaker 1 (25:45):
We didn't get into some accounting terms.
No, I mean I just figured, yeah,just in case.
I mean, you know, it's alwaysfunny when you do a terms or a,
you know, a terminologydiscussion because you
inevitably go down this rabbithole because you're using
[inaudible]

Speaker 2 (26:00):
terms to describe terms to describe terms and so
on and so forth.

Speaker 3 (26:03):
I think also Craig, it would be good to go over what
companies are doing today.
Like they, they have to makethat transition.
I know you touched on it, but togo in, in more depth, they start
with um, capital expenses thatthey've been doing for years on
this.
And then to switch over inparticular, if they're a public
company, it becomes more, itcould be

Speaker 2 (26:24):
Carter.
Yeah, that might be a wholepodcast in itself.
The idea in very quick, 32nd,um, maybe it'll be longer than
30 seconds, but so a companythat's got capital expenses from
three, five years ago, theirexpense or expensing over
amortizing over a three to fiveyears, those expenses are still

(26:44):
hitting their books.
If they decide to switch 100% toan OPEX model or a hundred to a
hundred percent public cloud,they're kind of getting a double
whammy.
So they're getting that expense.
The are the amortization,they're still expensing that and
now they've got the cloudconsumption expense on top of
it.
So it can be a big hitfinancially, you know, uh, or at

(27:06):
least it looks that way on thebooks.
The way that you advertise it, Ican always have a hard time with
that word and motorize the valueof the, about you a recorded
saying that a few times, but Ihave to, they have to blend the
two together kind of.
Yeah, that's the, you have tosee that over time.
So plan out your massivemigration to the cloud over a

(27:27):
period of time where the hitisn't so,

Speaker 3 (27:29):
and that's a factor.
Some people don't take intoconsideration and it is actually
a factor that inhibits them fromgoing to the cloud.
Right.

Speaker 2 (27:37):
But that's, this is, we're moving off of definitions.
We're diving into concept.
No.
Okay.
So we didn't finish talkingabout the, you had pay as you go
and then we reserved andservices.
Next slide.
So let's talk about reserve[inaudible].
So yeah, I'll let you start thistime.
So reserved instances, is thatMicrosoft and AWS[inaudible]

(27:57):
cloud provider all, it'ssomething different though,
right?
No, they call it AWS.
Had it before.

Speaker 3 (28:02):
Yeah.
Then they have to plan outcapacity for their data center.
So because the pay as you gomodel as out there and so
successful when people want it,they'll offer you the consumer,
the ability to reserve a SQLdatabase.
A I as instance, I'm not surethere's a few, two or three

(28:23):
other things that you can dothis on where you can prepay for
one year or three yearsconsumption up front and get
upwards.
I'm not gonna, I'm just gonnathrow a number out there.
63% savings on a three year,that 63% savings.
But the catch is, is that you'vegot to pay it all upfront.
And what Microsoft does isreserve that capacity for you.

(28:45):
So just to, for them having the,the ability to, to know that
you're going to be there inpre-buy the space, reduces their
model costs so much that theycan offer the savings to you.
So

Speaker 2 (28:58):
two weeks ago, Microsoft announced, uh, you,
you can buy reserve instances,lock in for the one year, three
year, but pay monthly.
Yeah.
So it's[inaudible] stream isthat the savings are there
savings aren't as no, they'reexactly the same.

(29:19):
They are exactly the same.
Someone has to be making moneyon that.
[inaudible] so what's happeningnow is Microsoft is doing it
like there's no financing feesor whatever, but there's still a
penalty if you decide to switchSKU's partway through.
There's a 12% penalty point onthe remainder.
So, so you're, you are lockingyourself in, but the business

(29:39):
case speaks for itself, right?
You're, you're saying, okay,we've got, uh, you know, however
many dozens of VMs or whateverwe're paying through the nose,
that pay is, you go pricing thefull retail price, but you can
lock in those savings and Hey,you, you need to scale your
environment and half of theservers are going to be a

(29:59):
totally different skew.
You cancel your RIS or reservedinstance, a subscription pay a
12% penalty.
I think it's on, it's proratedon the remainder of the term so
it's not totally bad.
And then you buy a new RI on thenext skew and it's, the savings
is still significant.

Speaker 1 (30:20):
So I have a few questions related to that.
Oh my God.
Okay.
Actually don't know, I mean, Iknow what a skew is, but I know
it.
I actually don't know.
Go, you're going to Google itright now.
So while he's, we will in that.
So I mean for one, that'sawesome.
Um, I'm, I mean, I guess it'sbecause Microsoft probably
realizes that they'd like therisk of people of them not

(30:42):
getting the money at the end ofthe day is low.
Right?
Like know your contracts andlocked in.
Um, where does it stands out?
Stock keeping unit.
Okay.
Um, wait.
Okay, so stock as it like it'sholding stock of, of inventory.
Stock room.
It's old school.
That's warehouse.
That's like a, I always say whenI think of skew, I think of a

(31:05):
UPC code right on you.
But that's actually just like askid of product.
Um, okay.
So

Speaker 3 (31:13):
he's asked me the other day what the, what the
save icon in office means.
Like they didn't know what Ohthis is.

Speaker 1 (31:19):
Oh.
And I was like, well that's howdo you call it is you must have
a floppy disc in your, in yourplace somewhere or you must have
a couple school.
I've got some, but my kids wouldknow.
Um, all right.
So, okay, so reserved instances,I'm assuming the use case would
heavily impact your ability topredict your monthly spend,

(31:42):
right?
Like, yeah, like if you, so ifyou're, if you're hosting a site
that is going or maybe like a,you know, a SAS application
where your um, you know, ongoingdaily, weekly, monthly traffic
could have a massive up anddown, then your, how would,
would you be able to takeadvantage of reserved instances

(32:03):
if you're not able to accuratelypredict

Speaker 3 (32:06):
has, and that's on a website and I don't believe
that's available for our eyesbecause it's elastic in a, in
nature elasticities on so, so,so, so you don't want that on
there.
It's around certain compute andstorage, uh, features that you
want to provide the capacity.
So if you think the old methodwas you gotta buy X amount of

(32:28):
capacity and now you're saying,well, I know I need X minus one
or X plus one, whatever way youwant to look at at capacity.
I just don't want to buy theextra stuff.
I know for guaranteed for yearone, two and three, I'm going to
need that much capacity.
You'll buy it.
But a website is not in thatbill.

(32:48):
Meaning a website is going tobe, you could have more
campaigns, you could be selling,not selling, almost don't

Speaker 2 (32:54):
you want that?
You don't want to lock yourselfin from a website if it,
especially if it's like aeCommerce site.
Right, right.
Cause the elasticity we want[inaudible]

Speaker 1 (33:04):
so can anybody guess what the next term is?
I don't know where it was first.
You think it might be last isn'tbeing all so next term, uh,
elasticity.
AKA on demand compute.
Okay.
I think the listener canprobably guess what it is based
on when we talk.

Speaker 2 (33:20):
Yeah.
So it will stick with thewebsite concept.
So you've got a website,eCommerce website, um, black
Friday's coming.
Uh, you, you may, in the oldway, you would, you would go and
buy a whole bunch of serversthat would be sitting there
idle, not doing anything.
They'd sit in the server roomand then, but you have the

(33:41):
maximum capacity that you needto achieve black Fridays demand.
And then the day after blackFriday sales are done, you're
sitting idle again.
Elasticy allows you to scale ondemand as required.
So you can do it automatically,ideally, or you can do it
manually still, but elasticyjust allows you to scale out or

(34:06):
scale up.
So suddenly you now have, Oh,I've got, that's different terms
we've got to define nowvertically and horizontally.
Um, yes.
But, uh, so scaling out, being,adding more of the same compute,
uh, systems, um, virtual systemsor scaling up is going to a
larger instance size or as wesaid earlier, a skew of, uh,

(34:29):
that model.
So, uh, getting more computefor, to meet the demand
essentially and memory.
And

Speaker 3 (34:37):
you know why this is so hard and this should actually
be a drinking game is that we'vetalked about terms and each time
you introduce a new term, itshould be a drink.
And that's what makes this socomplicated is because there's
term upon term upon term uponterm.
Yeah.
You've got to understand allthese concepts in succession and

(34:57):
we just throw it out.
Like if you just throw out ascalability on top of that,
which,

Speaker 1 (35:03):
well, so, so yeah, like you lost the city and also
scalability will, the vaginahasn't addressed.
Um, it's no drawn.
You can talk.
So yeah, the scale, I'll giveyou our, you asked the city, Oh,
you had a great, I'll give you agood example of something that's
kind of funny.
So, um, funny and sad.
So, uh, there's a term, um, thatsome people may recognize.

(35:24):
It's called the Reddit talk ofdeath.
You want the Reddit hug of deathis so, um, you know, Reddit,
they call themselves a frontpage of the internet.
It's a, you know, basically a,uh, uh, a site filled sub sites
around certain topics.
Right?
So what happens quite often issomebody somehow, organically

(35:45):
we'll, uh, basically puts aproduct or service that they're
offering into a conversation andit will jump up in, in, people
will notice it and it'll lookout upvoted or whatever.
So there was an example one timeof, you know, this guy who had
been ready to use it for a longtime and him and his mom as a
side business just for fun,built made soap.

(36:05):
They made homemade soap and theywould sell the soap online.
They actually had their ownwebsite.
It was hosted at a data centerand they're selling their soap.
And one day he's talking tosomebody on Reddit and then
going back and forthcommunicating and their their
comments, get up voted orwhatever.
And then he puts a post and itgot to the front page and all of
a sudden the link to this, tohis site is on the front page of
Reddit and Reddit.
And you know, a lot of usersrealize, Hey, this guy's a good

(36:27):
guy.
You know, he's a nice guy, he'sgot a business with his mom
there so and so let's do thisguy solid.
Let's promote his business.
We'll all order some soap offthis guy.
So the Reddit hug of death isthat his site gets so much
traffic and it's such a shortspan of tight.
In short, short span of time,the site crashes, it goes down.
So all those orders, all bypopularity are not able to be

(36:49):
facilitated.
Therefore he doesn't takeadvantage of that massive hit.
24 hours later, the traffic'sgone.
His site backs is not up.
But in the short, the shortattention span on the internet,
nobody cares anymore.
So all that's all thosesupporters don't get fulfilled.
I've

Speaker 3 (37:04):
seen firsthand where a server is sitting under a desk
physically and um, a company.
And in this particular instance,it was a tire company that had a
tire sale every year and tiresales happen like in the winter.
Uh, people buy their a wintertires and everything went

(37:25):
through that.
So the entire company's entireyear is based on that server
being running.
And guess what server getsknocked out of commission done
during the sale, during thesale.
And it causes all kinds ofproblems because they're not
just selling directly toconsumers or selling to
distributors and there's allkinds of things we're talking
hundreds of thousands of dollarsa day in, lost them in the

(37:48):
servers down.
And another company down heremakes these plaques that appear
on Oprah all the time.
And those plaques actually everytime that Oprah, because Oprah
loves them, every time she talksabout them, they're, their site
gets knocked out.
Guess where that server sittingunder someone's desk as well.
So I hear, I've got a tirecompany, I've seen it and that,
and this, the last testy isdirectly counter intuitive to

(38:12):
what those people are doing,except you have to architect the
system and the solution to beable to scale, leverage and to
scale.
Yeah.
You can't just say go into Azureand say, Oh, make this scalable.
It doesn't work that way.
Right?
Correct.
Correct.
Yeah.
Um, what it will do is allow youto do at a, at a base level is
to get into our next topic,which is, um, scalability.

(38:33):
The vertical allows it to, to goup that way without the
programming being great, meaningthe programmings or that it's
fully distributed or it candistribute work tasks one way or
the other horizontally.
So you've got vertically andhorizontally and vertically just
means that it's going to grab,it's just going to add more CPU

(38:54):
or memory to it up or down.
So if your compute, if you're,if you're using four processors
and a 20 gigs of Ram, um, that'spretty beefy.
Uh, that's at the high end.
And when your consumption cominginbound is low, you can actually
set rules that will drop itright down to two gigs of Ram
and, and uh, you know, uh, twoCPS or one CPU.

(39:18):
And then conversely as Craigmentioned, the horizontal
scaling your applications gottabe aware of it a little more.
It might involve a load balanceor as well where the
application's going to sayyou're taking the first request
server one, you're taking thesecond request server to stick
with your people, go back tothem and then you know, we'll
bring on the third and fourthserver when, when required, and

(39:41):
again, so you can go from havingone server to having 50 or back
again, wherever thatscalability, horizontal and
vertical from me.
And in the old, old on premworld scalability was you go buy
another server or you add moreRam or you add more CPU if you
could do that or more desk.
Yup.
But you're constrained by thephysical, you are constrained by

(40:04):
it for sure.
Okay.
There's unconstrainedscalability in the cloud.
Is infinite scalabilityinfinite.
That's the beauty of it.
Yeah.

Speaker 1 (40:13):
Well guys, we've covered off quite a bit.
Um,

Speaker 3 (40:17):
[inaudible]

Speaker 1 (40:17):
we can, we've got a few options here.
We can go a little bit longer.
Do you want it?
John's looking out his, uh, his,his phone.
You're looking at the time.
I can't tell if he thinks we'vegone too long or too low or too
little.
I think we have gone, we need acomment section that tells us if
people like us, you likesubscribe.
Oh yeah, we got it.
You owe it.
You know, you gotta add thatwhole thing in like a Hey guys,

(40:38):
you know, if you like your carcontent please like and
subscribe and uh, you know,share our content with friends
and family.
Um, uh, but, uh, yeah, no, Ithink, I think we covered quite
a bit today.
Um, I mean there's a few othersbut there's no end to it, right?
I mean, I don't, we could do apart three, but I don't think we
will.
I mean we'll end up justcovering more as we do other
topics.

(40:58):
I think we got into our three.
You want to do a part three?
I think we do geo replicatingavailabilities.
There's a lot in here aroundreplicating.
One piece of paper was alsoaround, you know, availability
zones all belong to thatdrinking game that I was
suggesting.
I won't drive the, I won't drivethat day and we can do that.
Um, we also, there's one other,a couple of others here like, uh

(41:19):
, you know, D dos, a avers ADEwife CDN and we'll keep adding
to this list.
So yeah.
Okay.
Part three.
It may not be

Speaker 3 (41:27):
in the near term, but we'll definitely keep it on
there.

Speaker 1 (41:31):
And we thought we've hired some new staff.
We should have probably had atleast one of them in here today.
I didn't think about it untilit's too late, but, um, cause we
had Theresa monitoring, althoughshe heard Mike wasn't working no
on part one, but we'll figurethat out for part two.
Yeah, we got there.
I Leora was keen.
She's been listening on to the,uh, the podcast lately.
And who does she thinks thebiggest character?

(41:53):
Oh no, we haven't asked her.
Oh, I think I'm probably the,probably you there.
But, um, all right, well, uh,yeah, I, I don't really have a
bonus question.
I haven't[inaudible] to shootfor our audience.
Just so you know, Mark May notsound himself today.
He tore his Achilles heel a weekago.

(42:14):
I'm actually on a significantamount of morphine right now.
So the fact that I've been ableto, you know, pull it all
together and even participate inthis podcast.
I've been hanging on, justbarely hanging on.
But uh, no, I survived.
Yes, no I did.
Uh, unfortunately, uh, terabyteKelly's song going to be, um, on

(42:35):
some crutches for a while.
But what can you do?
Well, the crutches I think foranother probably two months.
Um, yeah.
And then I know that fullrecovery, they said nine to 12
months, but you're not playingsoccer, hockey or anything
anymore?
Well, I will never play socceragain because I don't, I didn't

(42:55):
even like soccer to begin with.
And that's what I do with soccer.
I know.
Well, I mean it's how it ha.
It's what happened to be doing.
So how are you going to stay inshape?
Um, well I'm going to continueto go to the gym.
I'm just going to have to doeverything above the waist.
So can you get one of thosecaddies?
Have you seen those caddies?
Uh, I've, I'm not sure if, ifwe're on it.

(43:17):
Yeah.
Yeah.
The, the walking yeah.
Thing or whatever.
I don't know what I'm going toend up with.
I gotta wait till I get thisthing off and then have a chat
with them.
When I say this thing, I havelike a old school plaster cast
wrapped in bandages, but um,download, we'll see, uh, just
gonna you know, go to the gymand 50 years ago would it have
been dead rest for no.

(43:39):
They would have taken her outback and shot him.
Sorry buddy.
You're done.
Yeah, no, it's interesting causenowadays they almost never do
the surgery whereas two yearsago, like only in the last year
or two of, they decided that, uh, um, you don't need surgery for
this.
And I had to basically begged toget the surgery done.
Right.
But, uh, yeah, so I was able toget in and get some surgery.
So anyways, a little off topicbut uh, so thanks everyone for,

(44:03):
for joining in listening to ourpodcast today.
Uh, we will be back in a, Idon't know, probably another
probably two weeks.
We're not going to let it go aslong as we last time and I'm not
sure what the topic is, but wewill come up with a topic.
I will come up with a bonusquestion and uh, we'll see you
all then.
Same bat time, same bat channel.
Cheers.
Thanks.
Everybody's alright.

(44:33):
One is the loneliest number.

Speaker 4 (44:36):
Yeah.
And I'm glad you mentioned thatcredit because that's the
difference.
That's the new way to do it.
Say you're in the cap ex spendand you're getting to be able to
do experiments like this.
Like you just said, okay, let'sstop.
We have to stop.

Speaker 2 (44:52):
You're some battery died.
Is that one still alive?
This is your, your number one,right?
Yeah.
This thing was started flashingand now I'm not getting any
input so I will, I, yeah, I willedit this out, but I know,
sorry.
What can you do?
You're not going to be on thepodcast otherwise.

(45:14):
So let me just go swap that out.
Uh, those battery.
Oh no, this battery.
Sorry.

Speaker 4 (45:20):
Do you know why

Speaker 1 (45:22):
you swap it over?
Maybe for Watson?
That one right there.

Speaker 4 (45:25):
Oh, did I just change?
Yes.
You're right.
No two format format that no, Idid not[inaudible] format
because it's ability tooverwrite.
What do you mean?
Well, why not,

Speaker 2 (45:57):
um, batteries died halfway through the recording.
No, they're not, but I've gotbatteries, but yeah, no, I
bought Costco.

(46:17):
I didn't know.
No, this, my house runs onCostco.
Not aware.
My house runs on Costco.

Speaker 1 (46:26):
Yeah.
Well, honestly we should have, Ididn't think about it until we
have a few minutes.
And then we had, we had Theresain the first one.
I don't know why we didn'tnobody layer in this one.

Speaker 4 (46:36):
I thought about it a few minutes in.

Speaker 1 (46:40):
Oh, well,

Speaker 4 (46:44):
alright, sorry about that.
Do you remember where you weretalking about

Speaker 2 (46:51):
and you and you, you just, um, go from where you
said, I'm glad you brought thatup.

Speaker 1 (46:59):
I can't remember what the next point was.
Well, I think, well, we shouldjust start off with, Hey, we had
a battery malfunction here.
Let's, uh, we're gonna hop rightback in.
It's less awkward to do it thatway, but to pretend, um, pay as
you go.
What were we talking about onthat?
Alright, so let's, we'll holdon.
Okay, I'm going, so we're gonnajump in right now.
Oh, we're recording right offand cap pass.
That's where you're going.

(47:20):
Okay.
X and cap X.