Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:12
Hey, everyone,
0:12
and welcome to episode three nineteen
0:15
of the weekly Google Cloud Platform Pod
0:17
webcast. This is Stephanie Wong.
0:19
And today, I'm here with
0:21
Drumroll, please. Stephanie
0:24
Wong. Yes.
0:26
You are all in for a treat because it's just
0:28
me here today hosting, but
0:31
hey, I'm here on every other week
0:33
or so anyway. And, you know, I think
0:35
my teammates needed a well deserved break.
0:38
They're either in the heat wave of California
0:40
that we just had and they're out on a pool.
0:43
or they went to Burning Man
0:45
or they just plain wanted to take extra
0:47
few days off with Labor Day so I don't blame them.
0:50
but we have some great content up here today.
0:52
We I, I should say, just
0:54
had a great conversation with Sean and Nishant
0:57
who are the product leads for our
0:59
storage products. And
1:01
if you're thinking storage really
1:03
well, we had a lot come out. We just had a storage
1:06
launch day or storage spotlight, if
1:08
you will. It's this event online. And
1:11
they both had some great presentations about some of the
1:13
biggest launches of this year so far.
1:15
And so they're here today to talk about
1:17
any and all things storage, block storage,
1:19
object storage, storage for GKE.
1:22
There's a lot going on, so we're gonna get
1:24
them in here in just a second. But
1:26
of course, Even though it's just me today, I'm not
1:28
gonna skip out on the cool things of the
1:30
week.
1:36
First cool thing in the week is something that
1:38
I cannot cannot forget to mention.
1:41
Google Cloud Next is happening. Yes.
1:43
virtually still, but
1:46
it is open and available to register
1:48
for online at cloud dot with
1:50
google dot com slash next slash
1:52
register yes, we're gonna have that link in the show
1:54
notes as well or just search CloudDEX
1:57
register. And you're gonna be able
1:59
to, for free,
2:01
join us live online along
2:03
with one million builders, dreamers,
2:06
leaders for Googled Next two
2:08
thousand twenty two, and we're all gonna learn and grow
2:10
together to Take on what's next.
2:12
Again, access is complementary. I am
2:14
going to be in a couple of the sessions. I'll
2:17
let you know in just a little bit about which ones
2:19
those are, but yeah, definitely don't miss it. We
2:21
have more launches on top of what we were talking
2:23
about today for the storage. Now, I
2:25
have
2:25
two other cool things because it's me and I can do
2:27
whatever I want today. So my
2:29
next cool thing is a really awesome
2:31
blog post by one of our
2:33
developer advocates for DataFlow.
2:36
Dataflow is one of our data analytics
2:39
tools for batch and streaming.
2:42
And dataflow recently announced the general
2:44
availability support for Apache beams,
2:47
generic machine learning prediction and inference
2:49
transform. It's called run inference.
2:51
So in this blog post, they take a deeper dive
2:54
on that transform capability.
2:57
And you're gonna go through a
2:59
actual full blown Example and walkthrough
3:01
of how you can use run inference to
3:03
transform and remove large chunks
3:05
of boilerplate data pipelines that incorporate
3:07
machine learning models. So
3:09
take a look at that. It's very detailed.
3:12
And if you are in the data engineering world,
3:14
you're gonna love this because it will span your capability
3:17
to be able to do machine learning in production.
3:19
And then the last cool thing of the week that
3:22
I wanna mention is just a fantastic
3:24
blog post written by Forrest Brasil. He's
3:26
ahead of content here. He actually had
3:29
me featured in one of the videos that
3:31
I wrote about how to get into cloud from
3:33
non traditional backgrounds, but that is
3:35
just one video in series. There
3:37
are a bunch of videos talking about how you
3:39
can get into cloud from IT. how you can
3:41
get in cloud from help desk and
3:43
how you can get into cloud without a degree.
3:46
Priyanka did that video along with forest
3:48
and couple others in our team.
3:50
So definitely don't forget to check that out. If you're
3:52
interested in getting a career in cloud, these videos
3:55
give you some actual tactical tips.
3:57
Forest has a lot of background in this area. So check
3:59
out the blog post
3:59
for a series of them. Alright.
4:02
Well, that's enough cool things from one person, I would
4:04
say. So Let's go ahead and dive into our conversation
4:06
about storage with Sean and
4:08
Nishant.
4:13
Sean and Nishan,
4:15
thanks so much for joining for another
4:17
exciting
4:18
storage launch day. Thanks, Stephanie.
4:20
It's great to hear back on the show. Hard to believe
4:22
it's already been a year. Yeah. Thank you, Stephanie,
4:24
for having us.
4:25
Yes. We've had Sean on last year for
4:27
the inaugural storage spotlight. And this
4:29
time around, we're having the second time around,
4:32
none less exciting than last year. Now, Sean,
4:34
I know this is your first time on the podcast, but why
4:36
don't you start with giving us a
4:37
quick intro of what you do here at Google, and then I'll pass it
4:39
back to Sean to do the same. Sure. Thank you, Stephanie.
4:41
Yeah. I'm one of the outbound product managers
4:43
for our Google Cloud Storage product.
4:45
So excited to be on and love to
4:47
chat about key exciting things we're announcing.
4:50
like
4:50
I said, Stephanie, it's great to be here in the storage
4:52
product management team here as well looking across
4:55
our portfolio of offerings. And as you said, we've got
4:57
a number of exciting announcements that we just made
4:59
last week.
5:00
we have a storage spotlight day.
5:02
If you all listening haven't checked it out, definitely
5:04
go ahead and take a listen. We've had
5:06
you both do some talks there and so
5:08
those are up on demand, and we have those links
5:10
in the show notes. But storage, why
5:13
are we talking about storage? Because we're
5:15
talking all the time about these services and
5:17
products that are coming out in the cloud. That
5:19
store just sort of automated and taken care of
5:21
for you in some cases. but it's still
5:23
so hugely important for a company to
5:26
really keep top of mind. So from your perspective,
5:28
Sean, why storage? Why are you in
5:30
the storage space? Storage
5:32
is one of those unsung heroes,
5:34
if you will. Everybody needs it, and not
5:36
everybody unfortunately thinks about it from the beginning.
5:38
And one of the great things about Google Cloud is
5:40
that we have a lot of experience
5:42
of delivering massive scale applications,
5:45
Continental even global scale applications, I think
5:47
YouTube and Gmail, etcetera, which is
5:49
all really finding good. And that's all built
5:51
on our cluster file system colossus.
5:54
This is just a ridiculously large capacity.
5:56
Can't even mention how big it is, but it's just like,
5:58
as big as you're thinking it's bigger than
5:59
that. Right? This is a foundation that we've been running
6:02
applications on for years. This is
6:04
now the same basis, if you will, the same technology
6:07
that we're now building solutions that enterprises
6:09
can use for their business. And it's a different delivery
6:11
model of Google delivering a service
6:13
to an application versus delivering
6:15
a storage service that enterprises can
6:18
consume whether it be blocked, file, or object, and we'll get
6:20
into some of those details. But this is all just
6:22
crazy technology stuff that really is
6:24
something that can be thought of an afterwards
6:26
in terms of I'm just gonna subscribe to so much performance
6:29
or capacity and really take advantage of a
6:31
managed service. However, there are a lot of nuances
6:33
that have to come into play when you design
6:35
a continental scale application that
6:37
there are differences and things you have to think about
6:40
with storage. As easy as we're trying to make it,
6:42
we're not done yet. And that's some of the things that we're gonna
6:44
talk about a little bit.
6:45
Yeah. I mean, to have a storage day
6:48
dedicated to is a testament to how much
6:50
exciting stuff is happening in this space. So
6:52
what about from you and Sean? No, I think
6:54
Sean kind of summed it really well. It is on
6:56
Sun Hero, but it does need
6:59
storage to function everything, especially
7:01
I kind of talk about storage in the sense of
7:04
working on human generated data and that's
7:06
transformed into non machine generated
7:08
data and the volume of data that's being generated
7:10
is far exceeding what we can manage. So
7:13
having capacity at scale and
7:15
a continent scale is a key attributes
7:17
that definitely excites me. I've been in
7:19
the objects store industry for over decade
7:21
plus and I don't think we're going anywhere
7:24
slower. It's just going faster and faster. kick
7:26
talks of the world and other capabilities generating
7:29
more content that we can store. So
7:31
I think that just keeps us exciting out in our
7:33
capability wise and especially focus in the enterprise
7:35
space. Yeah.
7:36
Just from a consumer standpoint, when I think
7:38
about how I'm constantly paying a little
7:40
bit more, when I have more Google photos,
7:43
backing up all my photos. I'm like, how is this trend
7:45
gonna keep continuing for Eons into
7:47
the future? So it's a big problem space,
7:49
and I'm sure in the last decade, you've seen it change
7:51
in the industry. so much from a technological standpoint.
7:54
So we have these launches this year.
7:56
Can you talk about some of the themes that we've
7:58
been focusing on since
7:59
last year's storage day? What have
8:01
we been
8:02
really doing in this space? Sean?
8:04
We've really been focusing on optimization.
8:07
This is clearly as you'd expect from product development
8:09
active. These don't happen overnight. Right? So this is
8:11
really long term planning that we've had in place
8:13
about how customers can optimize their environment.
8:16
And
8:16
that's optimizing not only the cost,
8:18
but also storage utilization and performance,
8:20
and how do we optimize their own resources? because
8:23
as you started off by asking about, it's just
8:25
a managed service. that's true, but somebody
8:27
still has to check a few boxes,
8:29
make a few mouse clicks, type in a few commands.
8:32
So, of course, we still have ways to go to make it
8:34
even easier than what we have been doing today. And
8:36
so as we talk to customers, there are really kind of
8:38
three overall pillars that we are
8:40
focusing on. That's number one, enterprise readiness.
8:43
companies that are gonna bring applications into
8:45
Google Cloud, how are we gonna deliver
8:47
enterprise scale applications, both in
8:49
terms of reliability, and file store
8:52
enterprise is an example with four nines of
8:54
regional SLA availability for those critical
8:56
applications like SAP and even GKE.
8:59
or high performing applications need
9:01
with files to high scale, where you have high
9:03
performance computing workloads that need
9:05
twenty six gigabytes per second of throughput
9:07
for that, or improving our block storage
9:09
with persistent disk extreme. So
9:11
there are things that companies just have to do to kind
9:14
of run of the mill, if you will, the day to day things,
9:16
but there are also a lot of the data driven application
9:18
and they're consuming all that storage you're
9:20
referencing for the photos and so forth with
9:22
cloud storage enhancements for analytics. And
9:25
if you look at what we've done to make it
9:27
easy for companies to deliver continental
9:29
scale applications with our dual
9:31
region option for cloud storage where
9:33
they can replicate between nine
9:35
different regions across three continents, all
9:38
to a single bucket, and all to the same API
9:40
access regards to storage class. is something
9:42
that's unique to cloud and unique to Google cloud.
9:44
And this is one of the things that we're doing for
9:47
availability of those applications in addition
9:49
to making sure that they can not only have a recovery
9:51
time objective of zero to be all failover for
9:54
one region in a single bucket to actually
9:56
add that option to replicate the data
9:58
what we call turbo replication with a fifteen
10:01
minute RP0
10:01
Right? So you have an r two of zero
10:04
and RP0 of fifteen minutes for
10:06
petabytes of capacity or tens of petabytes
10:08
of capacity is just unfathomable.
10:10
And then lastly, really around storage everywhere.
10:12
So this is really the notion of yes,
10:14
a lot of applications are coming to Google
10:16
Cloud, but there is still an edge presence
10:19
on premises. And that's where our
10:21
Google distributed cloud edge appliance comes into
10:23
play where you have companies like neuro
10:25
that have distributed on premises
10:28
workloads. They have their autonomous
10:30
driving vehicles collecting data all day.
10:32
They pull into the shop, they upload the data, the appliance,
10:34
and it's really a transition into getting that data
10:36
into the cloud for further training in
10:38
AI ML workloads. And it's not a transfer.
10:40
Now a one time transfer is more of a daily occurrence,
10:43
and so it's bridging that cloud
10:45
boundary, if you will, down to the customer's
10:48
data center.
10:49
while wrapping your head around what you just said,
10:51
twenty six gigs of throughput needed and
10:53
also having storage options with a fifteen minute
10:55
RPO for petabytes of data. Can we
10:57
just sit with that for a second? Because it is pretty impressive.
10:59
Just thinking back to where we've started when it
11:01
comes to storage, like, when the first storage
11:03
area networks were developed and where we
11:06
are today with having multiple users and
11:08
the type of latency needs that we have
11:10
today,
11:10
it's quite impressive. And and
11:13
Stephanie, those are the things we talked about largely
11:15
last year and have delivered just this year. So we're not even
11:17
talking about the new stuff this year's story spotlight
11:19
yet. This is all things that we've been we've been
11:21
busy this year even ahead of the announcements last
11:23
week. So I think
11:24
that's a perfect segue because I was gonna ask about some
11:26
of the key launches and what are
11:28
we talking about this year? Maybe start
11:30
with things that kind of fall into the enterprise readiness
11:33
space.
11:34
It's really around I would characterize it as a block
11:36
storage. Right? If you take the typical application, I'm gonna
11:38
run SQL Server, I'm gonna run SAP
11:41
and Google Cloud, how would it make it easier
11:43
for customers to optimize their deployment? so
11:45
that's where we announced the next generation of persistent
11:48
disk called Hyperdesk. So
11:50
Google Cloud Hyperdesk is really the option
11:52
that gives customers three levers to
11:54
turn, if you will, iOPS throughput
11:57
and capacity. And so you're going to
11:59
start to see this
11:59
rollout next quarter in Q4 this year.
12:02
particularly focused on the higher end workloads
12:04
as it'll be called hybridistic stream. As the
12:06
name implies, it's gonna be Phil's most demanding
12:08
database workloads where you've got
12:10
to get the sizing right. Right? Coming back to the what
12:13
you said opening, Stephanie, about sizing
12:15
application that what makes storage interesting is
12:17
that people have to think about it from the beginning. And
12:19
historically, They've gone through a long planning
12:21
cycle about iOPS and trying to figure out what
12:23
the application is gonna need from the
12:25
business, not only in day one, but also
12:27
day three sixty five. and how does that change
12:30
over time? And so with HyperDisc, we're really giving
12:32
customers that option to optimize and
12:34
make it easy to provision iOPS throughput
12:36
and capacity on day one, but as the
12:38
application changes over time, tune
12:40
those knobs. Make those adjustments as needed without
12:43
going through lot of arduous tasks of
12:45
migrating data around. And the other
12:47
is really around files to enterprise with GKE.
12:49
You know, we're seeing more and more stateful workloads
12:51
running in Kubernetes, and that
12:53
really depends upon availability
12:56
of data because
12:56
you can't have a stateful workload that
12:58
you can't number one backup or have
13:01
multiples pods have access to the same
13:03
data. right, for failover and availability reasons.
13:05
And so now that same four
13:07
nine SLA of regional availability with file
13:10
store enterprise, we've brought to GK
13:12
already, but now we're actually optimizing how
13:14
customers can maximize a storage utilization
13:17
and carve up hundred different shares of capacity
13:19
for hundred pods to be able to write
13:22
to the same data at the same time. And
13:23
that really changes the way customers and the flexibility
13:26
that gives customers to deploy stateful
13:28
workloads in GKE? Yeah,
13:30
that's been an initial challenge I think
13:32
with containerized environments as having to have these
13:34
modularized architectures that need to
13:36
access the same storage in order to have state
13:39
full applications be deployed on Kubernetes. So
13:41
it's like we've matured a lot in
13:43
this space, and storage has been a huge part
13:45
of that calculation. Nishan, is there
13:48
anything else that tested
13:49
out to you in this storage day for enterprise
13:51
readiness? Oh, yeah. There's a lot of citing stuff
13:53
that we've introduced in our cloud storage
13:55
world. And Sean talked about just volume
13:57
of data that we have to manage now. Customers
14:00
are taking advantage of our cloud stores at a confident
14:02
scale, but at the same time, they're worried about
14:05
cost. So the three pillars that we
14:07
talk about cost performance and capacity.
14:09
One of the key aspects is cost. How do I
14:11
manage my cost at billions and trillions
14:14
of objects or petabytes at
14:16
bytes of data. That seems to be a daunting
14:18
task for our customers. And often, they
14:20
don't know what workloads are coming
14:22
to our object stores. They're just
14:24
turning applications or bringing in applications
14:26
from business use case perspective. So
14:28
managing that whole aspect of
14:31
what application create what kind
14:33
of workload and managing what data should
14:35
go and what class of storage tends
14:37
to be pretty daunting. So we are now introducing
14:39
what we call our auto class capability
14:42
for cloud storage, where it's really
14:44
our easy button. You click it and forget
14:46
it and set it. And now based on your
14:48
access patterns, we'll ensure that
14:50
their data is sitting in the right storage class
14:53
based on your access, and we'll make sure
14:55
that we're optimizing your costs along the
14:57
way. So a lot of our customers have to
14:59
start to figure that out at a get go where
15:01
they have to understand the application workload
15:04
and start to plan that. And so we
15:06
wanna take that planning away or
15:08
have to reduce that aspect of
15:10
planning. So autoclaves really
15:12
helps sets us apart from our competition
15:15
and also helps our customers. Now
15:17
I think of it in two ways. We do
15:19
give our customers flexibility to
15:21
be the what we call the advanced users.
15:24
Do it yourself. They can build their own
15:26
patterns and usage and decide which
15:28
storage class they wanna retain the data in.
15:30
that's perfectly fine. So you continue
15:33
having those advanced users, but the ones
15:35
that just wanna have simplicity and
15:37
ease but still manage their costs on
15:39
their behalf, and Autoclast does that
15:41
for them. So from that perspective, the
15:44
capability gives them a great
15:46
scale. but also the flexibility of
15:48
just simplicity that we can provide. It'd
15:51
be hard to achieve that on your own.
15:52
And I have a lot of thoughts about what you said because
15:54
when it comes to cost optimization with storage,
15:57
we actually had an episode on FinOps,
15:59
which is a kind of growing area, especially with
16:01
regards to cloud. And So it's this intersection
16:03
between finance and operations. And so
16:06
the more that we can automate these things and do
16:08
these recommendations and activists' related
16:10
features we have to optimize your
16:12
storage, which is can be a larger operational
16:15
burden for OpEx. This is great.
16:17
I mean, I haven't seen it across industry,
16:19
and I think you just mentioned it might be one of our unique
16:21
launches this year or two, so that's super exciting.
16:24
And when it comes to the fifteen minute turbo
16:26
replication and having multi regional single
16:29
buckets. I heard this phrase the
16:31
other day. It's like when starts are thinking about
16:33
choosing a cloud provider, choose a provider
16:35
and first start to think about multi region,
16:37
not multi cloud. And that's what you really
16:39
need to prioritize when it comes to having
16:42
that enterprise readiness and availability
16:44
before you start to think Okay. Let me just choose across
16:46
all these providers. It's like, can we get
16:48
one that really gives us the capability to be multi
16:50
region?
16:51
Spot on and And then lastly, another
16:53
amazing feature that you introduced is really
16:56
around our storage insight. Now when we talk
16:58
about the volume of data you're generating,
17:00
As much as we can automate, there are certain aspects
17:02
we cannot. So we do want to give
17:04
insight into the customer what type
17:07
of data they're storing. So our initial launch
17:09
of storage insight is gonna be ging this
17:11
year or early next year as well, which is
17:13
giving them knowledge of what capacity they're
17:15
storing. In those trillions of objects?
17:18
What kind of data are they storing? Can we give
17:20
them the visibility into that? Are there
17:22
any patterns in that storage data
17:24
that they can then facilitate And also
17:26
other aspects could tie into you talked about
17:28
fan ops and compliance space. Customers
17:31
are looking at insight of what type of data
17:33
I'm showing in my buckets. Am I
17:35
giving access to customers that shouldn't
17:37
have access to that data? Should that data
17:39
not be publicly accessible? There could be
17:41
PII data in there? So really all
17:44
those aspects, giving our storage and site
17:46
service to begin with, can give them
17:48
kind of directional guidance, not just
17:50
from a managing capacity perspective,
17:52
but managing information. So we really
17:54
wanna grow from, of course, pod
17:56
storage. It's all about storing data at massive
17:59
scale, but we wanna get to managing information
18:01
for our customers and then giving them insight
18:03
how they can operate their business better,
18:05
how they can manage their application better,
18:08
how they can optimize their application for
18:10
best workload. So for example, they
18:12
have analytical workloads. We talked about
18:14
dual region, how we can co locate compute
18:17
and storage together. So getting that
18:19
depth of knowledge from their data, they can then
18:21
decide what's the right storage class
18:23
for them or right storage operations
18:25
they wanna do and cool features and functionalities
18:28
that we're introducing when to set it.
18:30
And then also thinking about from a security
18:32
perspective, some of the challenge we've heard
18:34
from our customers hey, am I managing
18:37
all my data across with the right
18:39
keys, encrypted keys across the board? Are
18:41
we rotating them enough? So Having
18:43
that global policy level
18:45
of information across all their buckets is another
18:48
key attributes that they can take advantage from our storage
18:50
insight service. there's
18:51
a lot happening clearly. And recently,
18:53
we just had an episode on GKE because they
18:55
just turned seven. And we have much
18:57
more exciting features coming out with that product,
18:59
but I know you were talking a little bit about what
19:01
we have for storage when it comes to GKE
19:03
Shawn. What are we announcing around
19:05
GKE and the theme of storage everywhere?
19:08
Yeah.
19:08
Happy belated birthday to us. Right, Stephanie?
19:10
Yeah. So I talked about
19:12
files to enterprise. So that's kind of on the storage
19:14
side for those stateful workload deployments. But the
19:16
other part that goes along with this is have
19:19
to be able to recover an application that you really care
19:21
about. If you can't recreate the application or
19:23
the data easily, you've got to back it up. Right?
19:25
And we've been doing this for long time in
19:27
VMware on prem or GCVE in
19:30
Google Cloud or GCE. But
19:32
now we're actually delivering this for Kubernetes.
19:34
And this
19:35
is actually something that is it's our
19:37
own development. We developed this to
19:39
be integrated directly within the cloud console.
19:41
So now you've got a couple mouse clicks choose
19:43
which workspaces or pods you wanna protect,
19:46
where you wanna protect them locally or in another
19:48
region. And then in the recovery side, you have
19:50
that same granularity of I'm gonna recover
19:52
everything. I'm gonna recover portion of my
19:54
environment. And you really have to
19:56
have that forward thinking about protecting
19:59
the applications having the file
20:01
store data being available across multiple
20:03
pods for availability. We're
20:05
using persistent disk with GKE.
20:07
Right? You back up that data with backup for GKE.
20:09
And so it's a really changing of a mindset
20:11
about being able to run more and more different
20:14
applications that are typically more important,
20:16
if you will, in terms of stable, but not necessarily.
20:18
but it really is supporting that customer journey
20:20
throughout their transformation and optimization process
20:23
because not every customer is gonna go directly to
20:25
GKE. They're gonna have a portion of things that are gonna
20:27
run-in VMs. and others are gonna be running
20:29
containers. Yeah.
20:30
That's a good point. It's usually very much so
20:32
heterogeneous environment, and so it needs to be
20:34
adaptive in terms of being able to support these
20:36
different deployments. So I know
20:38
we mentioned security just a moment ago
20:41
with storage insights, but is there any more you
20:43
can talk about when it comes to the security of
20:45
data?
20:45
yeah, it really comes back to the heterogeneity.
20:47
You mentioned Stephanie in that backup for
20:49
GKE is obviously protecting that environment.
20:51
However, you've got to think about everything else.
20:53
and this is where we're announcing Google Cloud
20:56
for backup and DR. This is also now
20:58
directly integrated into the cloud console.
21:00
For those of you familiar, this is from the Actifio
21:02
acquisition. that we made little bit ago,
21:04
but now it's fully managed service. So there's
21:06
nothing else for a customer to install. Just
21:08
point click, which applications you wanna protect,
21:11
where you wanna protect them and what that policy looks
21:13
like. And so for GCVE with
21:15
VMware or GCSE and VMs,
21:18
databases and applications, we can now protect
21:20
applications running in the cloud consistently. You
21:23
have the protection for GKE applications. So
21:25
you're really providing that consistency for
21:27
security and data protection. across
21:30
all of the environments, across the heterogeneous deployment
21:32
you're mentioning. But then the other thing that with
21:34
backup and DR that we're announcing is
21:36
that, yes, the in cloud protections there,
21:38
But if you're running VM or on premises, you
21:41
can use the same capability to protect those
21:43
VMs on premises and back them up to Google
21:45
Cloud. So as you're thinking about
21:47
the potentially a hybrid deployment and
21:49
a migration strategy of what's gonna be moved when
21:51
into Google Cloud, you have that option to backup
21:53
from on prem to the cloud within the cloud?
21:56
You
21:56
have options. Yeah. I love it.
21:58
We talked a little bit about cost optimization
22:01
earlier in the conversation, but I wanna circle back
22:03
to it just because I know storage, as I mentioned,
22:05
such a fundamental building block, but there's
22:07
also ballooning storage, which we said, and
22:09
that comes with ballooning cost and OpEx
22:11
especially in today's age is tightening up for a
22:13
lot of companies. So what are we doing to help with
22:15
cost optimization?
22:16
Yeah. So I would say couple of things,
22:18
Stephanie, on the operational side. Nishan
22:21
talked a little bit about auto class where we've
22:23
now given them the easy button, if you will. So
22:25
they only can apply this auto class policy
22:27
to an entire bucket very simply.
22:29
And that can manage billions or trillions of files
22:32
and move data down. And if you think about
22:34
the cost of standard all the way down
22:36
to archive, Archive
22:37
is a fraction of the cost of standard.
22:39
And as cheap as cloud storage is,
22:41
if you're taking, you know, ten, fifteen, twenty percent
22:44
the cost of standard, that's a significant cost savings
22:46
over time. Right? that's number one. Number two,
22:49
I've talked about hyper disc and how giving
22:51
customers three knobs to turn between
22:53
IOPS throughput and capacity. so they
22:55
can right size that. But then also, we're introducing
22:57
storage pools for hyperdesk. And this
22:59
is gonna be the notion of, think
23:02
of thin provisioning now in the cloud.
23:04
So we've done this on premises for years where
23:06
you provision block storage and you over provision
23:08
and you actually are using less physical storage.
23:11
Same notion here. We're now giving customers that
23:13
flexibility. to optimize the storage utilization
23:16
with hyper disc and storage pools so
23:18
that they can actually, as you'd expect, save
23:20
some money and and actually rightsize how those
23:22
applications are consuming storage. And then also
23:25
with with file store enterprise and multi shares,
23:27
carving up file store instances to
23:30
be more granular and more increase that storage
23:32
utilization with smaller shares
23:34
just helps to right size again those pod deployments.
23:37
with GK.
23:38
Very nice. I love the themes of rightsizing and
23:40
auto classing. This is a trend
23:42
that is definitely super valuable.
23:45
And of course, I know that for many
23:47
of our lunches, we don't do this alone.
23:49
We take so much feedback from
23:51
our customers to help guide where we wanna
23:53
take the products next course.
23:54
This is not an exception. So
23:56
were there any customer use cases and stories
23:58
that you have from some of these
23:59
launches? Yeah.
24:00
No, Stephanie. That's valid many launches
24:03
we have prior. We do work with closely with
24:05
our customers. And similarly here, we'd be DevIS
24:07
is one of our customers that they're in the research
24:09
industry. They aggregate research data, and
24:11
then they provide that to the end user or the applications.
24:14
They're from analytics perspective or
24:16
research aspect perspective. And
24:18
one of the challenges that they generally have
24:21
is they don't know when that research
24:23
data is important. Now, they have to
24:25
retain that data forever. but
24:27
in early stages of research that data
24:29
is fairly active, but then eventually becomes
24:31
dormant. So they wanna be able to have some sort
24:34
of capabilities, i e, auto class, And
24:36
based on the access patterns where they don't have
24:38
to continuously pull the users, are you done
24:40
with this data? Just migrate or move
24:42
that data to the right storage class and
24:45
start saving them costs. And
24:47
one thing that struck us when we were talking
24:49
to them is it's more about human
24:51
failure or accidents that happen
24:53
where when we do move the data
24:56
to the lower storage cost,
24:58
but, you know, when you access that data,
25:01
it starts to to hear more
25:03
cost to you because there's operation
25:05
charges there. So with our auto class capability,
25:07
we automatically move that data to standard
25:09
storage and ensure that on subsequent
25:12
reads, you're not paying that heavy penalty
25:14
of retrieving that data. So not only
25:16
we're automating, getting used to the lowest
25:18
cost tier, But at the time of access
25:21
and when you wanna get that data, we'll
25:23
move it to the right storage class or the more
25:25
of the hotter storage class and ensure that you
25:27
can use that data at any given time.
25:29
Throughout this whole thing, you know, the application
25:32
doesn't need to change. It's a single API.
25:34
And even at our lower cost
25:36
storage, the latency to read that data
25:39
is milliseconds. So from the end
25:41
user perspective, they don't have to change their architecture.
25:44
They don't have to change their application. They
25:46
get the same functionality they were getting
25:48
regardless of what storage class are sitting in.
25:50
So from their perspective, it
25:52
made their life as a storage admin
25:54
or the owner of the infrastructure, so
25:57
much simpler that they can then focus
25:59
those engineering resources in the
26:01
part of the business that mattered is really
26:03
collecting that research data. One
26:05
of the things the customer said once this
26:07
thing is GA and ready to go, they
26:10
expect ninety percent of their data will be managed
26:12
under this capability. So which is huge
26:15
saying from their perspective, and I think they
26:17
can take huge advantage of this capability as
26:19
it comes out. Yeah.
26:20
No. That's amazing. I was gonna say that
26:22
I'm sure the red flag that people
26:25
weigh when they hear out of close. That sounds great,
26:27
but if it's gonna affect how easily
26:29
I'm going to be able to pull that data
26:31
back down, if there's cost differences,
26:33
if it's gonna be difficult for my application
26:36
to be compatible with such differences,
26:38
and that's gonna be a big decision factor.
26:40
But if it's easy and seamless, then
26:42
why not? Right? That's
26:44
great. So Can you step
26:47
back real quick, Sean, and summarize some of
26:49
these launches and where they fit into our broader
26:51
portfolio to help
26:53
our customers secure their data, simplify
26:55
operations, and optimize the cost.
26:57
If you take a look at what we've done, it's really across
26:59
the portfolio of the storage that we have.
27:02
lot of different applications are gonna come to the cloud,
27:04
Not everything is gonna be leveraging object storage.
27:06
Not everything is gonna be leveraging block storage. So
27:08
you have to really think about what we can do to optimize
27:10
deployments as customers are thinking about their enterprise
27:13
readiness or the data driven applications
27:16
or if you look at storage everywhere. And
27:18
so if you think about block storage, introducing
27:20
HyperDisc. The next generation persistent disc
27:23
give customers knobs a tune, the performance
27:25
and capacity. We've announced capabilities
27:27
around files are enterprise to right size that storage
27:30
utilization within stateful workloads for
27:32
GKE and multi writer deployment scenarios.
27:34
We've got the data protection side of things.
27:36
backup for not only GKE, but also
27:39
Google Cloud backup and DR for GCVE
27:41
and GCE and other databases. then
27:43
we also have cloud storage. And the auto class you're just
27:46
talking about of making it easy to tier data
27:48
back and forth between the different storage classes
27:50
without the additional charges. So optimizing that
27:52
placement and giving customers the information
27:55
that they need to make informed decisions about
27:57
what objects are where it was storage insights. And so
27:59
We certainly am busy again across the portfolio
28:02
of block file, object, and data protection
28:04
this year. Howard Bauchner:
28:05
Yes, you certainly have. And to anyone
28:07
who's questioning like storage. We're talking about storage.
28:10
Yes. There's so much that has to do with
28:12
storage. It's so fundamental still. And so
28:14
thank you so much. Is there anything that you wanna mention
28:16
before we start to wrap up? I think
28:18
you mentioned at the top of the show, if you didn't
28:20
have a chance to see stores spotlight, it's available
28:23
on demand. Yes. For
28:24
anyone who is listening out there, We
28:26
have the links in the show notes, storage spotlight,
28:28
second year in a We're gonna keep with this trend because
28:31
clearly there's a ton happening. And, Sean,
28:33
and Sean, I just wanna thank you
28:34
again for coming back on. Now thank you for
28:36
the time. Yeah.
28:37
Thanks Stephanie. Alright.
28:39
I wasn't wrong. Right. tons of stuff happening
28:41
around storage when it comes to optimization. cost
28:43
optimization capacity optimization, performance
28:46
optimization. We talked a lot about
28:48
rightsizing, auto classing,
28:50
file store enterprise hyper disc. We
28:53
have lots going on around the theme of simplicity
28:55
and ease of use with data protection. We have
28:57
GKE backup and Google
28:59
Cloud backup and disaster recovery. and
29:02
I love the use case that they went
29:04
over when talking about redivis. So
29:06
I'm really glad that we got Sean back on and included
29:09
Nishant this time around. It's gonna be an annual
29:11
transition that we just have storage, spotlight, and
29:13
have them jump on because I know a lot of you
29:15
might not have the time to check out
29:17
the full on demand video of that happening.
29:19
And so the podcast is such an easy way
29:21
to get the product team on and give you a great
29:24
summary of some of the exciting things that
29:26
are happening in the storage space. Now,
29:29
I know we always talked about what we're working on, and
29:31
so I just wanna give a quick update that I
29:33
am doing some exciting video content with
29:35
one of our latest partners, and
29:38
I'm not gonna give too much away, but you're
29:40
gonna see it at next a
29:42
cool partner announcement, and so I'm gonna
29:44
hopefully be able to showcase that video
29:46
soon. And then in terms of next, I'm
29:48
also doing two sessions. One
29:50
is for networking, which if you've
29:52
seen my videos, you know that I love talking about
29:54
networking. And this session covers how
29:56
you can simplify and secure your network
29:59
for all workloads
29:59
in the cloud. So we're gonna covering
30:02
all of the announcements related to
30:04
networking. And then I am
30:06
also in a session for
30:09
our partner ecosystem, where I bring on
30:11
Rubik and Exavim to talk about how you
30:13
can take on security in the
30:15
cloud using some of their tools
30:17
alongside the analytics capabilities
30:19
in Google Cloud. So check
30:21
that out. I know it might not be everyone's
30:24
cup of tea for the sessions I'm in, but we got a
30:26
ton coming from our VPs, GMs.
30:28
Obviously, Thomas Kurian, Priyanka's, in
30:30
some too, so definitely come here from these
30:33
leaders. And I'm gonna keep
30:35
reminding you about next for the next several weeks. Sorry.
30:37
I'm not sorry. But for
30:39
now, This is Stephanie Wong, and I will
30:41
see you all next week.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More