Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
Today's sponsor is Datadog, real time
0:02
monitoring platform that unifies metrics,
0:04
logs, and distributed traces from your cloud
0:06
containers and orchestration software. Datadog's
0:09
container centric monitoring features allow you
0:11
to track the health and performance of your dynamic
0:13
container environment. The container map
0:15
provides a bird's eye view of your container fleet
0:17
And the live container view searches groups
0:20
and filters your containers with any criteria
0:22
like tags, pods, or workspaces.
0:25
To start monitoring container clusters, sign up
0:27
for free trial today, and Datadog will
0:29
send you a free t shirt. Visit datadog
0:31
dot com slash container dash
0:34
cloudcast to get started. That's data
0:36
dot com slash container dash
0:38
cloudcast.
0:38
Cloudcast media presents from the
0:41
massive studios in Raleigh, North Carolina.
0:43
This is the cloud cast with our undeveloped
0:46
and Brian Graceland, bringing user
0:48
of cloud computing from around the
0:50
world.
0:52
Good
0:52
morning, good evening, wave y'all, and welcome back
0:54
to the cloud cast. We're coming to you live from our
0:56
massive podcast studios here in Raleigh,
0:58
North Carolina. And
1:00
as I hit record on this, it is
1:02
the middle of September. And
1:05
probably from a news standpoint, there's
1:08
two big news articles we need
1:10
to talk about, and we're gonna go ahead and jump
1:12
right in. First of all, VMware
1:15
Explorer, formerly known
1:18
as VMware World, happened recently. And
1:20
Alex William, over at the new stack,
1:22
put together a really good piece.
1:25
Kinda highlighting some maybe
1:28
some of the less covered aspects of
1:30
it, but of a particular interest to
1:32
our listeners. And so this is less
1:34
about virtualization and less about
1:37
infrastructure, but more about
1:39
cloud native, about containers,
1:42
tanzoo, and also the
1:44
rise of what they're referring
1:46
to as a platform engineer. So a platform
1:49
engineer, of course, is somebody
1:51
that is building the platform for
1:53
the developers because you see a lot of companies
1:55
these days. How do we talk to developers? I wanna
1:57
talk to developers. At the same
1:59
time, they may not be the ones
2:03
doing the purchasing at times, or
2:05
they just don't care because
2:07
they don't care about the platforms underneath it. It's
2:09
not something they think about on a daily basis.
2:12
So interesting article
2:14
that also goes in to mention just
2:16
the dev experience in general
2:18
and VMware's history there
2:21
and They also mentioned Backstage,
2:24
which is an open source project VMware
2:26
uses to support its developer experience.
2:29
They mention Pivotal and
2:31
the history with Pivotal and the spin
2:33
in and spin out and they mentioned Heptio
2:36
and a good friend of the
2:38
show, chip Childers who
2:40
is the chief open source officer
2:42
at VMware has
2:44
some things to say in the article as
2:46
well. So definitely worth
2:48
checking out if that is of interest to you.
2:51
And for our second article, the other big thing
2:53
going on in the industry right now, if
2:56
you follow crypto, is Ethereum,
2:58
and the Ethereum blockchain merges
3:00
and changes coming up. I'm
3:02
not gonna go into this too
3:05
in-depth, but I did want to mention it because
3:07
it is a pretty big news piece, because it
3:09
has been attempting
3:11
to happen for a long time now. I wanna
3:14
say, gosh, a good six months
3:16
or so now they've been trying to do this merge.
3:18
And if you're not familiar
3:20
with it, what is it? Well, you have really
3:23
two ways to really do blockchain.
3:25
One is called proof of stake. The
3:27
other one is called proof of work.
3:30
Proof of work is the older way. It's what
3:32
the original Bitcoin was built
3:35
around in Ethereum as well. Most
3:37
new ones are what we call proof of stake.
3:39
Why is proof of stake necessarily better?
3:42
Each one has their benefits and downsides,
3:44
but the big benefit for proof of stake
3:47
much, much lower energy
3:50
consumption. Proof of work requires
3:53
a lot of computation. When you think
3:55
about a Bitcoin miner, or the,
3:57
you know, the big crypto miners and the big
3:59
rigs and with big GPUs and lots
4:01
of fans and taking all the energy, they're
4:03
typically proof of work. So
4:05
by moving over to proof of stake,
4:07
hopefully that will alleviate some
4:10
of the concerns around that in Ethereum.
4:12
I mean, the two big downsides
4:15
to Ethereum have always been. It's
4:17
it tends to be a little slower than a lot
4:19
of the others, and it tends to be a power hog.
4:21
So this will at least
4:24
start to fix some of that, and we'll
4:26
lay the foundation for more improvements
4:28
over time as well. With
4:31
that, I'm going to wrap up cloud news of the
4:34
week. Coming up right after the break,
4:36
we are gonna be talking about Kubernetes
4:38
cost optimization.
4:40
offering.
4:42
Oh, annoying. Did
4:44
you know that seventeen out of twenty people stopped
4:46
watching a video because of stalling and rebuffering?
4:48
Don't let your users click away to a competitor
4:50
site. If your business lies in online
4:52
media, rely on CDN seventy seven
4:54
to deliver a seamless online experience to
4:56
your audience. CDN seventy
4:58
seven is one of the leading global providers of
5:00
content delivery network services. They
5:02
power the world's most popular websites
5:04
and apps such as Udemy, ESL
5:06
gaming, live sports TV, and social
5:08
media platforms. Aside from their massive and
5:10
redundant global network, you're gonna love their
5:12
no BS attitude and skilled team of engineers
5:15
ready to help twenty four seven. No
5:17
chatbots, no tickets bouncing around unresolved
5:19
for days, just people who know your use
5:21
case can immediately help you pinpoint and fix
5:23
the problems. Don't wait until your users run
5:25
out of patience. Go to CDN seventy
5:27
seven dot com slash cloudcast
5:29
and ask her a free trial with no duration or
5:31
traffic limits. That's CDN seventy
5:33
seven dot com slash cloudcast.
5:35
in
5:38
And we're back. And folks, as
5:40
we get closer and closer to Quban,
5:43
North America, and Detroit, and if
5:45
October of this year. We've been spending a little more
5:47
time this summer throughout the last
5:49
few weeks and so forth. Really kind of digging
5:51
into the things that are kind
5:53
of adjacent to the core Kubernetes
5:56
project. We've obviously talked about that quite a bit
5:58
and we've been looking at more and
6:00
more areas in which as more people
6:02
deploy Kubernetes, as more people optimize
6:04
their Kubernetes as they look to
6:06
kind of expand where it's being used and so
6:08
forth. We continue to look areas in
6:10
which, you know, there's technologies, there
6:12
are new ways of thinking about how to make those
6:14
environments better, how to make them more efficient, how
6:16
to make them more responsive
6:19
to your business in order for you to be able
6:21
to to go solve business problems and not have to think
6:23
about Kubernetes so much. So You know, we
6:25
continue to kind of dig into that and, you know, one of the
6:27
things that we love to do on the show is not only
6:29
look at the new technologies that are making that
6:31
happen, but also you know, new companies
6:33
that are, you know, at the forefront
6:35
of making that happen. And so today, we're really excited to
6:37
kinda do both of those things. So so really
6:39
excited to have Shahar, Yakhov who is
6:41
product manager at Granite, who is an
6:43
intel company. Shahira, welcome to
6:45
show. Great to have you on. Hey, Brian.
6:47
Happy to be here. Yeah.
6:49
So excited. You know, we we love having
6:52
new companies. You know, we love having
6:54
companies focusing on new areas. Tell
6:56
us you know, before we dive into into granulate
6:58
and all the things you guys are doing, tell us a
7:00
little bit about your background and, you know,
7:02
ultimately, what what, you know, what you
7:04
guys focus on at granulate? Sure.
7:07
Yeah. So first of all, my name
7:09
is Shahriar Khong. It's really hard to
7:11
pronounce it in Hebrew. In the English because
7:13
it's an Hebrew name. And so
7:15
Sharia Cobb. Currently working as a
7:17
product manager at Granite, and
7:20
I started my journey around there.
7:22
decided to go as a software engineer in the
7:25
Israel Defense Force. And since
7:27
then, I had the opportunity to work with
7:29
different companies, and specialized
7:31
mostly on the dev side
7:33
and CSD aspect and
7:35
some things about monitoring and observability.
7:38
And I did a transition from a software
7:40
engineer into project manager somewhere
7:42
in the middle. And right now,
7:44
as I said, I'm working on twilight, really
7:46
excited about that because I have the opportunity
7:48
to both use my technical skill
7:51
that I got from my my technical
7:53
background and my product run with
7:55
skills. Yeah.
7:56
No. That's that's great. We've had a number of
7:58
of alumni of of of Israeli defense
8:01
really really smart people. They trade you
8:03
very well. You you work on very challenging
8:05
problems. So, yeah, when we see
8:07
when we see those folks move into
8:09
into the technology product space, we
8:11
always know that there's really good foundation
8:13
there. Let's let's start by
8:15
talking, you know, I highlighted at the beginning, we're
8:17
gonna talk a little about cost optimization.
8:21
As you think about it as a as product
8:23
manager as to, you know, how you
8:25
help companies that are running Kubernetes
8:28
or any large compute environment,
8:30
like, how do you think about, you know, what are the
8:32
things that we can do to help people around
8:34
cost optimization? What's kinda what's the
8:36
bigger picture the way you think about it?
8:39
Yeah. So I I
8:41
think you highlighted it's pretty good. Right
8:43
now, it's pretty pretty
8:46
important to understand where your
8:48
money goes to? What what is the cost that you are
8:50
paying for your environment? Whether it is,
8:52
like, on premise environment or
8:54
cloud environment? Yep. It doesn't matter. Because
8:56
right now, the burn rate, it's really
8:59
important thing to understand
9:01
where the burn market and the burn market
9:03
there, everything to change, the economics
9:05
So understanding how to improve
9:07
your margins, I think it's pretty pretty
9:09
important right now. Right. And
9:12
and and as an optimization
9:15
company, and we have multiple solution.
9:17
Yeah. We have a good solution on portfolio,
9:19
mostly focusing on cost
9:21
optimization. and performance optimization.
9:24
And also, when we are focusing on
9:26
performance optimization, most of the time we
9:28
translate those kinds of optimization, those
9:30
kinds of performance optimization, into
9:32
cost optimization. There are
9:34
sometimes when we translate them into
9:37
improvement of the SLA and
9:39
competitive SLA and that stuff, but
9:41
mostly what care of the company that
9:43
we are working on, what really care about them and
9:45
what the biggest pain pain point of
9:47
that is the cost. how much they
9:49
pay for the cloud provider, how much they pay
9:51
for their infrastructure, and they want to
9:53
eliminate it. And I think
9:55
that that the few things on their
9:57
side, like, is it boarding to other
9:59
step? And the the the step
10:01
where the company gets to the point
10:03
where they need to understand, like, how
10:05
much they paying for stuff? Sometimes
10:07
it get, like, too far from the from
10:09
the point where it important on the
10:11
standard. And so we think the first and
10:14
foremost things to understand that cost
10:16
optimization is too important, then
10:18
that need to be handled. And
10:21
I think this is the the the most important
10:23
point
10:23
to to understand right now. Yeah.
10:26
Yeah. Well, and and I think and you highlight
10:28
something that's really important and this isn't
10:30
always obvious. You know,
10:32
what I've found having worked around
10:34
Kubernetes for a long time is, you know,
10:36
there's there's a period of time when when people are
10:38
just getting started. They're trying to figure out
10:40
the technology. And and oftentimes,
10:42
you know, they'll they'll look at other things.
10:44
They look at cost optimization
10:46
or or even security or stuff, and they
10:48
and they say, well, we only have one
10:50
cluster right now. I don't I don't really need to
10:52
worry about those things. And
10:54
what inevitably happens, and
10:56
we we see this time and time again
10:58
at different companies. Once they
11:01
figure out the basics of making things work
11:03
and they realize, okay, I can I can
11:05
help deploy applications faster. We can
11:07
scale them automatically. Like,
11:09
their ramp their ramp up is really
11:11
fast. And I and I think what you're really
11:13
highlighting is, while the
11:15
ramp up in terms of getting applications can
11:17
go fast, like, the cost can can
11:19
grow really quickly too. And you you you
11:21
need to be thinking about that early
11:23
in the process rather than waiting till you
11:25
get those first big bills and then not
11:27
knowing what to do. Yeah.
11:28
Yeah. Exactly. And as
11:30
I said, there there are several ways to to look
11:33
at that. there is a like, you you
11:35
can look at that in in many points of
11:37
view. The first one, you can look at that on
11:39
a cost visibility view. Like, where the money
11:41
goes to, which services are cost the
11:43
most. And what the resources that
11:45
derive the most cost and the the
11:47
like, wasting the the most of money
11:49
or investing the the most of money. The
11:51
second way to look at that is, like,
11:53
how do I manage my infrastructure better?
11:56
how do we manage that that the different
11:58
resource that that loads pools from
11:59
Kubernetes or autoscaling groups from
12:02
a a VM based application,
12:04
compute based application. And
12:05
our approach at graphite is looking
12:08
looking of that on a really unique
12:10
and other way from a point of
12:12
view of the application says As
12:14
the application is really, really dynamic.
12:16
And as the company, the the
12:18
massive growth, the rapid growth, as
12:20
the company grows, and
12:22
you it's becoming the
12:24
hardest place, the the the lost
12:26
densify area for the
12:28
company because microservices, age,
12:31
microservices, like, top every every
12:33
deal too in the environment. And
12:36
from a big program company, the trial
12:38
is, like, Right? Two to three deployments, it
12:40
become really very
12:42
fast, like hundreds of deployments and
12:44
multiple clusters and multiple staging.
12:47
and most of them in the production. So
12:49
and our approach is to look in that in the
12:51
application level layer, really, like, from the
12:53
application level perspective. Okay.
12:54
Yeah. And I was I was gonna ask you you sort
12:56
of led me into my next question. It's,
12:59
you know, there's there's always ways you can
13:01
focus. Do you focus on infrastructure? Do you
13:03
focus on on application layer.
13:05
You know, kind of walk me
13:07
through a little bit as your as
13:09
your, you know, as your products
13:11
and platform are looking at the application
13:13
layer. what are some of the common things
13:15
that that you see? And
13:17
and, you know, some of the areas that people often
13:19
make mistakes or or have, you know,
13:21
problems that lead them to higher bills?
13:23
Yeah.
13:23
Sure. So as I said,
13:25
our first solution were to
13:27
to to check there that to to
13:29
get cost reductions through performance
13:31
improvement So we gather first solution
13:33
in the portfolio, the the continuous
13:35
runtime optimization, which trying
13:37
to identify utilization pattern
13:39
in your application, and then by
13:42
identifying this button, we know how to leverage
13:45
all this button and open a button
13:47
next some bottlenecks in
13:49
your application. And that's most of the
13:51
time leads to performance improvement, which
13:53
we translated in the cost optimization.
13:54
But as we look in that
13:57
from the predictive point of view, And I
13:59
want that a little bit to that.
14:01
It's that looking on the competitive point
14:03
of view, it's like become
14:05
pretty different. It's like behaved really
14:07
different. because as
14:09
a as a competitive container's
14:11
work, if you'll improve the
14:13
performance utilization part of each
14:15
application utilization part, did
14:17
did did did not mean necessarily that
14:20
anyhow
14:21
your your computer, your cost
14:23
did improve. because Kubernetes, the the
14:25
way it works, the the scheduler part of
14:27
Kubernetes is just taking
14:29
those those atomic box
14:31
and which we call deployments
14:33
workloads, products, whatever, and
14:35
taking those atomic boxes
14:37
and trying to locate them into
14:39
the different nodes. And by
14:41
locating the those those
14:43
boxes into the the the node is now
14:46
that the orchestration, the command is
14:48
orchestrate the rock. Yeah. Schedule
14:50
the rock. And but yeah.
14:52
Yeah. So what we actually did, you know, for
14:54
our Kubernetes users, we
14:56
did the first phase of the performance improvement.
14:59
reduce the utilization part,
15:01
but has has the
15:03
like, we we show them the the performance
15:05
improvement They didn't achieve any
15:07
significant cost reduction immediately.
15:09
So
15:11
we we put our in investigator
15:13
and the research ahead and trying to
15:15
understand why why the immediate cost reduction reduction
15:18
haven't achieved. So
15:20
after investigating that and beat,
15:22
we identified the hover provisioning
15:24
gap. And by what I mean
15:26
by say over provisioning is that
15:28
there is the requested resource set from
15:30
Kubernetes, the the requested value. that
15:32
each deployment of prod like
15:34
set, and there is the utilization part of
15:36
that. And there is a huge
15:38
gap. Something's a huge gap. Something's a
15:40
little less, but there is a gap between
15:42
the actual reserve resource to
15:44
the actual utilized resource.
15:46
And without changing the the
15:48
requested resource, cost
15:50
reduction cost reduction can
15:52
be achieved because as I
15:54
said, credit is works on the atomic box
15:57
determined by the the reserve spot.
15:59
the And that's what you actually do. We're
16:01
looking at the application ever from the
16:03
requested resources at point of
16:05
view.
16:05
Yeah. It's it's a little bit, you
16:08
know, So sometimes in in technology, we
16:10
see the same problems come up again
16:12
from from one generation or the other. So
16:14
we we used to see back in the
16:16
days of virtualization people were
16:18
buying really big boxes
16:20
to run an application, and
16:22
they vastly underutilized them.
16:24
But because they just
16:26
knew they had overhead. They they
16:28
worried more about, okay, I don't want the box to be
16:30
overrun. And Kubernetes
16:32
came along and and sort of told
16:34
everybody hey, we're gonna we're gonna schedule
16:36
your stuff. We're gonna automate it. And everybody
16:38
went, oh, that that's great. That's gonna be really
16:40
good. But it's like you said, Kubernetes
16:42
does exactly what you tell it to do. So if
16:44
you tell it to take a bunch of resources to
16:47
deploy things, it's gonna do
16:49
that perfectly well, and it's gonna make sure you
16:51
always have those resources And I think
16:53
what you're really highlighting is some of
16:55
the same problems we used to have with
16:57
virtualization, which was people
16:59
tend to over provision things
17:01
just because they're not sure how the application's gonna
17:03
run, or they're worried about memory leak or
17:06
whatever. And and so we are sort of coming back
17:08
to the same same challenge again
17:10
just in a new, you know, in a new
17:12
environment. And and given
17:14
how fast Kubernetes works versus, you
17:16
know, the old days of virtual machines,
17:19
you know, you you need systems more paying attention
17:21
to this than just people with a spreadsheet that look
17:23
at it once a month? Yeah.
17:26
Exactly.
17:26
Like, when you start
17:28
using Kubernetes, you're assuming that everything
17:30
will work out of the box. Everything will
17:32
work perfectly. Right. And nothing's
17:34
special with Appen. Nothing nothing, you know, intervention
17:37
from your side should be appen. But
17:39
as
17:39
we, like, change that around, like, the
17:41
the the usage of
17:42
relatives, where they standard like,
17:44
there are two two main challenges for
17:46
us to related users. The first challenge
17:48
is to set those requested
17:50
resource sets for each container.
17:53
because if not if not like
17:55
something that you can understand prior
17:57
to the the setting of this value,
17:59
you can't really know what should be the
18:01
value. So it's most of the and I'm
18:03
setting the copy paste from another containers
18:06
since sometimes they tried an error. Right. But
18:08
they tried an error. There is
18:10
some risk try to get there. I was remembering
18:12
you get throttling. So,
18:14
like, this approach wasn't really hard
18:17
because like, most of competitive
18:19
users want to utilize both
18:21
their vertical scaling concept and their result
18:23
of scaling concept. But I can
18:25
use anything in, like, like, you're leveraging
18:27
both of them to improve. So to
18:29
get, like, maximum cost reduction, it become
18:31
really complex. So this was, like, the
18:33
first challenge that we faced. And the second
18:35
challenge that that we saw that happening
18:38
is like there are different
18:40
persons in each company that
18:42
in charge of different areas. So
18:44
if you know, competitors allow allow us to, like,
18:46
move some of their of their ability
18:48
to their engineers so they
18:51
can and beat it to fast
18:53
microservices. They can deploy it really fast.
18:55
There are CICD tools that do it for
18:57
you. And the but
18:59
have the the like, this
19:01
whole area of move to to the
19:03
engineering, and they need to to set
19:05
the value the values of Like,
19:08
the CPU memory presses for each
19:10
container, there are other departments,
19:12
which is the DevOps department that's in charge
19:14
of the budget. So, you know, as
19:16
they engineer infecting the values, they
19:18
had their hopes engine the dev of engineer, dev
19:20
of department need to,
19:22
like, give the explanation. Why is the budget? Why is
19:24
the billings that high? Like, the
19:26
the this bill side of
19:28
become, like, rechallenge in
19:31
many company.
19:32
Yeah. Absolutely. Absolutely. Granite
19:36
just recently launched a
19:38
a product service SaaS service
19:40
called gmeistro, your product. Tell
19:43
us a little bit about about how it works. You
19:45
know, what are the what are the
19:47
the most common ways that people start using it
19:49
and then some of the benefits they see?
19:51
Yeah. Sure. So first and
19:54
foremost, like, the the
19:56
the as I said, the motivation guys in France was to to
19:58
solve those two big challenges
19:59
that the the company that the community
20:02
has. And
20:04
What is
20:04
it? Are are
20:05
a solution? The GMS
20:08
solution? What actually, that is,
20:10
like, continuous right sizing. continuous
20:12
rightsizing your calls and your workloads.
20:14
And we achieved really good results
20:17
around sixty percent cost reduction from
20:19
a rightsizing operation. And,
20:21
actually, it's a pretty straightforward
20:24
concept. We just remove the gap and
20:26
reduce the gap and get the gap
20:28
between the utilize resource set
20:30
to their action to the reserved resource
20:32
set. And the way we do it in
20:34
technical matters, we just deploy a
20:36
single pod into a cluster First,
20:38
we provide a visibility, a full visibility
20:40
into the Kubernetes cluster.
20:42
I'm showing all the deployment
20:45
information, the deployment digitization, whether we
20:47
need deployment, roll out schedule
20:49
set, all the queries object.
20:51
And second, a few minutes after, we
20:53
provide the first recommendation, state
20:55
of recommendation on outright side
20:57
of the workload in your capabilities.
21:00
And we provide you and we generate
21:02
those kind of recommendation. by
21:04
checking out all the peak usage
21:06
on your workload, taking
21:09
into consideration out of
21:11
memory and throttling effect that
21:13
already happened. and checking,
21:15
like, we want to make sure that your
21:17
that you don't need to compromise on your
21:20
competitive SLA. And that's actually how we
21:22
build the recommendation. We build the
21:24
recommendation for you. We show it to you.
21:26
And the continuous part of that, that
21:28
we allow you to apply this recommendation with
21:30
the click of a button, and we make sure
21:32
that every time your workload
21:34
change, evicted, pretty dynamic.
21:36
We right sized your workload again
21:38
or over again. So you
21:40
always, like, use the the the
21:42
list resources needed. to
21:44
end up the load that the workload need to be ended.
21:46
Okay.
21:46
Good. So it's it's not a not sort of a one
21:49
time scan and and, you know, it's a it's
21:51
continuous sort of it's
21:53
watching what's going on, making recommendations, and
21:56
and are the recommendations done
21:58
on a on a dynamic
22:00
basis? Does it sort of see what's there
22:02
and and fixes it for you
22:04
or or gives recommendations back to
22:06
the to the user and and they, you
22:08
know, they they can decide what changes
22:10
to make or what you know, what areas to
22:12
evolve? Yeah. So
22:13
it really depends on the user. It
22:15
can take those recommendation. We
22:17
provide the the required yummin
22:20
changes to provide the patch command to
22:22
to update the workload
22:24
deployment that the recommendations
22:26
will generate their own. But If
22:28
you don't want, we can apply them
22:30
for you for them, sorry, and
22:32
ensure that and I said it is continued
22:35
solution. We ensure that every time that the the vocalist
22:37
changes at the time passes and the
22:39
utilization part of change as well,
22:41
and there are, like, different peaks and
22:43
different spikes, We can take that into
22:45
account and make sure that every time and every
22:47
change that the world changes, we are
22:49
making sure they said that
22:52
the the workload remained
22:54
the right side.
22:55
Now do you find do
22:58
you find some optimizations or
23:00
are sort of more immediately impactful
23:02
or, you know, some some are more
23:04
focused long term. Like, how do
23:06
once once people sort of see what's
23:09
available to help them, you know,
23:11
how do they tend to to think through it? Is
23:13
it, okay, these are these are immediate
23:15
because they're just you know, they're they're too big
23:17
and bulky and then others are
23:20
you just you're making little tiny
23:22
incremental things over time. What what
23:24
tends to happen as they as they see the optimizations
23:26
presented to them. Yeah.
23:27
So most of the
23:29
time, people, when they feel recommendation, they want
23:31
to figure out how we generate those
23:33
kinds of recommendation. after checking your
23:36
recommendation of tool and making sure
23:38
that this is a quality recommendation, they
23:40
can apply. And most
23:42
of the time after applying those recommendations, they cluster when
23:45
they just cluster autoscaling, autoscaling,
23:47
direct magic, and reducing
23:50
the loads the loads amount.
23:53
Sometimes when the cluster is like
23:55
a static cluster, we need to change it
23:57
manually and we provide a recommendation to change
23:59
that manually for you as well. But more
24:01
than the time competitors does it magic by
24:03
itself. But I think the the
24:05
concept of rightsizing is,
24:07
like, after you right side your
24:10
workload one time, you need to make
24:12
sure to right side your workload every
24:14
time and continuously. Since if
24:16
your workloads wanted to try size in the
24:18
first place, and it can gather out
24:20
of your room, it can end up throttling, And
24:23
there is also a lot of waste in terms of resource.
24:26
But once you write in the
24:28
workload, we need to make sure that it remains
24:30
right side over time. So,
24:32
like, the first impact should be immediate. And
24:34
as the time goes, you just make sure that
24:36
you don't raise much more resources,
24:39
and you're running in the most optimized
24:41
way. Gotcha. Gotcha. And and
24:44
Jim
24:44
Maestro is is deployed as a as
24:46
a SaaS service. Correct? Yeah. Yeah.
24:48
It's a SaaS service. You
24:51
can you know, open our platform,
24:53
you know, from the environment
24:55
side, open the GMX platform, sign up
24:57
with the screen offering. You can use it
24:59
for free, check out your recommendation, and
25:02
start using it right now with multiple
25:04
cluster on multiple application as well.
25:06
Gotcha. And and will it work for obviously, it'll
25:09
work for workloads
25:11
Kubernetes workloads in the cloud? Will it also
25:13
work if a if a cluster is is
25:15
on premises? Yeah.
25:17
So if you have,
25:20
like, an Internet connection from your
25:22
over VPC from your own premises and you
25:24
can allow an outbound
25:27
networking. It will be no problem to to
25:29
connect our view management to your
25:31
platform. Okay.
25:31
That's excellent. That's excellent.
25:33
So we've talked a little bit. Obviously, we could
25:35
we could spend a lot a lot of time going
25:38
in-depth. For people that are are looking at this and they're
25:40
going, yeah, we're we're beginning to see this
25:42
this sprawl of clusters in our kubernetes
25:44
environment. We're starting to, you know, get some
25:46
bills that we don't love what are
25:48
some of the best ways for,
25:50
you know, companies to engage with not
25:52
only the the GeoMistro platform, but
25:54
also potentially engage with
25:56
your team. Sure.
25:57
So they can open the open
25:59
our website, but I dot I'll check
26:01
the different solution we have. They
26:03
said their runtime optimization solution,
26:06
and Paul, the ride saving solution,
26:08
Jim Mastro. And Jim Mastro,
26:10
it's pretty you you definitely need
26:12
to sign up and deploy our our
26:14
Jim Mastro pod and then
26:16
you already get any sort of a combination. You can use
26:19
it on multiple cluster. You can use it
26:21
on single cluster, whatever
26:23
they they desire. And they
26:25
are the solution that you just need to schedule a
26:27
demo with one of our experts.
26:29
Excellent.
26:29
Excellent. Yeah. It's the other thing
26:31
I think it's it's really interesting, you
26:33
know, you you granulate is an
26:36
intel company. Back in the
26:38
day, Intel used to make a lot of
26:40
investments to drive
26:42
more and more CPU usage, you know, video
26:44
and and other things. It's it's really interesting
26:46
to see them evolve to now looking
26:48
at, you know, how do we be really efficient
26:51
with the the CPUs that you're using, the compute that you're
26:53
using because, you know, it's
26:55
it's so easy now for people to
26:57
to drive more usage. There's lots of business projects
27:00
that drive more compute usage and, you
27:02
know, AI and and data science and stuff. So
27:04
it's it's really good to see them
27:06
being on the other side and giving people
27:08
access to tools that that make them efficient
27:10
as well. So that's that's exciting to
27:12
see. You you know, last
27:14
last question, you know,
27:17
what's the best way you know, give us an example, what are the
27:19
best ways for folks to to engage with your
27:21
team? What are some of the, you know, kind of earliest
27:24
ways to beyond just
27:26
signing up for the account? Do you see
27:28
what are maybe some of the immediate types
27:30
of things that you hear feedback from from
27:32
people like, hey, we just save a
27:34
bunch of money or, you know, hey, you know, we we were
27:36
surprised we we could optimize some things. What are some of
27:38
the early, you know, stories
27:41
you hear back from people using the
27:43
platform? Yeah.
27:43
So we launched our
27:46
platform around there months ago. We got some pretty
27:48
amazing feedback from the community.
27:50
And we have a Slack channel. like
27:52
a like a platform for the
27:54
community, which we ask the
27:56
questions from the community, we have we
27:58
answer for any support provided
27:59
over there, but if
28:01
someone need to contact us, he's a
28:03
great child to do so.
28:06
But we got great feedback.
28:08
We we we are already checked that platform
28:10
before releasing it to the to global
28:12
available. We checked that in our
28:14
internally customer current customer, and
28:16
we achieved amazing
28:19
amazing cost reduction around an average
28:21
of sixty percent It
28:23
depends on how much the
28:25
environment was over provisions before our
28:27
first solution. but most of the time it
28:29
was over provisioned, as I said.
28:31
And I think the
28:33
best way to contact us just
28:35
join the community, join the Slack community,
28:37
talk with us. You can send me a
28:39
private message. You can send me a link
28:41
in link in message. And
28:44
But this will be the best way to listen. Excellent.
28:46
Excellent. Good stuff. I appreciate the time
28:48
today. It was really, really good not only to speak
28:50
to you again, but also to learn more
28:53
about what granulates and what the
28:55
GeoMistro platform is doing. So folks, as
28:57
always, we've been highlighting
28:59
more and more around Kubernetes, not
29:01
only how to run run the
29:03
platform, keep you up to date with the technology. But
29:05
with that, we
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More