00:00
Welcome. Good afternoon. So I'm happy to that a few of you are here and to listen today's session which is about uh pure cloud blockto and EVS Elastic VMware service and new service from from AWS. Uh, My name is Pavel. You can, you can hear that I'm not from the US. I'm, I'm based in Prague,
00:25
so in our Prague, uh, R&D hub is our pure storage Prague R&D hub, of course. Uh, I am a technical product specialist, so I'm responsible for our cloud solutions and, uh, I am working for Pure for, for 1 year only, so I'm quite new but before Pure I was working for VMware for more than 6 years, so.
00:49
It was during the old good times before a broadcom acquisition. So I have a special guest today with me it's. So I'm Ron Wooddell. I'm a senior partner solutions architect here at AWS, um, working primarily with RSI and ISP partners to build solutions and, uh, help our customers migrate to the cloud.
01:13
Prior to, uh, AWS just over 34 years ago in the before times, I was actually on the customer side as a infrastructure architect and I hold my. One of the few folks left that have a VCDX with VMware, so. It's so cool. OK, so First, I would like to focus on uh on CBS.
01:36
So what, what it is, what is the architecture? I have a question to you, uh, is there someone in in the in the audience who who is uh who is who has experienced with CBS already, or, or no, no one, or do you know some, some of you what is CBS and how it works? A Little bit. OK, so it'll be educational.
01:58
Let's go. OK, so first, uh, before I jump into the architecture, I would like to look at, uh, what how is the data in the cloud different between in in comparison to on premise so. If you look into on-premise environment and compare it to cloud and look at how the data are stored, so you, you, you know very well that within within on prem we have typically
02:24
highly reliable and available arrays. Those arrays are has some enterprise features. It, it, it can doin provisioning, it can do the duplication, compression, it can do snapshots, things like that. In the cloud, typically if you, if you, if you would like to store your data in the cloud, you you those those services because cloud is a service, those,
02:50
those storage is typically has lower availability, but it has higher durability from efficiency point of view in the cloud the storage is provisioned as a tick, so you, you need to provision the whole capacity. And no, no one doesn't care if you, if you use all the capacity so we need to, you need to choose from, uh from capacity you, you need to select the correct uh or
03:17
right performance from cost considerations uh in the cloud you need to consider things like egress, egress charges. Typically resources can be wasted so you can you you need to provision a lot of, a lot of capacity to get the the the correct performance or or things like that. So, so that's the reason why I built uh CBS this is CBS cloud block
03:41
store. So, so it's our 100%, 100% software offering which we are providing in the cloud. So we are not shipping our flash arrays into into cloud data centre. It's our. Purity operating system which is running in the cloud and it's using the cloud infrastructure, so the, the hyperscalar infrastructure which is provided by,
04:03
for example here from AWS. Uh, CBS, uh, by its design, it brings high higher efficiency, so we are, we are using our features which are available with our flash arrays. We are, we are using our management so we can, we can, we can leverage our.
04:25
Pure pure fusion features and things like that. We enable some hybrid mobility scenarios we can. Connect CBS with for example with with on-prem flash array to do some mobility or migration tasks we can bring some better protection to to cloud because of our our architecture, our operating system.
04:50
CBS is running the same operating system as our flash arrays, so all features which are available today in in our flash arrays are available in CBS. So our grade data reduction, thin provisioning, so you, you don't care, you don't care about about the the capacity or how how many capacity you need to you need to provision.
05:12
We, we can do our or we can use our replication features. We are, we can use our snapshots. We are, we are using our encryption within within the CBS. So everything that you know from from on-prem on prem world, so it's it's available in the cloud. One thing is not.
05:32
Possible it's not possible to use uh use file file services. So CBS is only block, block storage, so it's it's a block storage running on top of cloud cloud infrastructure on top of cloud manage discs, for example. Uh, if you look a little bit deeper into, into the architecture. Uh On the left there is, there is our flash array which which you,
06:02
you, I guess you know very well. So in our flash arrays we have, we have some controllers, we have some, some flash modules and VRAMs, and we are, we are trying to, to keep the same same as as the CBS, but we are, we are of course to be limited with something. Which need to do different. So for controllers we are,
06:25
we are currently, we are using dual control architecture the same as as flash array, but we are, we are running our our controls with EC2 instances. We are using 2 EC2 instances currently C5 and the large one instances which which act as a primary and secondary controls. For, for NVR RAMs, we are, we are, we are using IO2 volumes,
06:51
so, so those are, those are used for, for as a, as a right buffers. And as a as a capacity flash modules we are, we are using instant stores. So, so in the architecture we are, we are creating, uh, we are, we are saying virtual drive. It, it, this drive is each virtual drive has 11 small, small easy to instance with
07:16
with uh IO2 volume connected and the instant store connected we are, we are. Using or creating some virtual shelf, so within the shelf there are 77 virtual drives, and we are, we are also using some rate HA on top of that so to to get the the the.
07:41
Durability and availability. Because, because instant store are, are very fast but uh sorry, but are not uh uh persistent. We are, we are also mirroring all the data which is written into, into, into the instant instant stores drives. We are mirroring the data to S3 bucket.
08:03
So those S3 buckets is used to. To to get some uh higher durability and in case of virtual drive will be, will be lost or we can if we if we lost the the the whole virtual shelf, the data will be rehydrated back from the S3 bucket and the all all this will be, will be built, built as a as a new as a new shelf.
08:31
We are also uh using in the architecture uh Dynabodibi. This is a very small database which is used for, for some metadata and some for, for, for configuration parameters, for example. Uh We are, we are, we have two SKUs available in AWS, so there is one which is, which is smaller, which, which has lower performance and uh this,
08:58
this SKU is uh is uh. The capacity of the SKU is limited to 30 terabytes as an initial capacity. We can do capacity upgrades by by adding new, new virtual shelf, so the capacity will be doubled if if customers need that. So, so.
09:19
We can double the capacity from 30 to 60 terabytes us usable data after the data reduction. The second one, a scale which is more capacity capable and is more performant, is, is able to is starting from 120 terabytes and it can double double the capacity. So we are able to do the capacity upgrades. It it is non disruptive so it so it will add
09:50
new new shelf, new shelf, and new virtual drives. We can also do uh purity, non-destructive upgrade. So if, if, if we have a new version of Purity, it, it, the customer can do the, do the upgrade through uh through, through CLI or through uh pure one, sorry.
10:11
So So this is the current architecture as I mentioned, we are providing the block, block storage so you can connect, you can connect through the controls. You can connect using ICAS or we are supporting NVMETCP. You can connect this storage to to your workloads or to your, for example, AWS EC2 instances or to EVS.
10:41
So this is, this is for the architecture. Are there any questions so far or not? Is it good? OK, so, so. So that's the reason of why we are, we are able to to lower the cost of of cloud storage because when when I mentioned on at the beginning when you are.
11:04
Provision in cloud storage, you need to provision as a as a as a tick as a tick, so you need to. Get the whole capacity, attach the capacity to your VM, and then you are you are you are storing the data so. Cloud box store can can reduce the the cost of of cloud storage by by our data reduction ratios, so by using our compression, the duplication, thin provisioning those those
11:32
features can dramatically decrease or lower the cost of of of the cloud storage. We can, we can do the, as I mentioned before, we can do the snaps, clones, things like that. We can, we can do some, some other safe mode of protection between, between the data, data deletion and and uh.
11:51
Things you, you know, from, from uh our flesh arrays. So this is, this is for the. Beginning about for you you don't know how CBS is uh working and how it's made and now I will, I will hand over to to Ron to continue. Thank you. So yeah, I, I was mentioning this before we
12:14
started the session, um, you know, lots of upheaval in the market with Broadcom's acquisition of VMware and one of the. Um, Issues that came out of that was VMware cloud on AWS no longer being able to be sold from AWS's perspective and, um, it is a fully managed service by, that's owned and operated by Brockcom.
12:37
So this time last year, AWS we, we kind of had some internal meetings and discussions and say how can we build our own VMware service on our hardware but allow our customers and partners to have full control over the service to kind of set us apart from other hyper scalar, uh, VMware solutions. And what came from that was our elastic VMware service, which is currently in public preview and it should be GA,
13:03
uh, soon. So Essentially what it looks like is a fully deployed VMware cloud foundation 521, uh, environment in a customer's VPC running on EC2. Uh, I4I metal instances. This isn't in a shadow account. It's not in a sidecar data centre. It is inside.
13:26
It's running on the same I4I metal that anything else in our AWS that uses I4I. And lives in the customer's VPC, so there's a full amount of control and access and then low latency connectivity to other AWS services, including, uh, Pier's Cloud block store. So, oh, that doesn't look pretty, does it?
13:50
So again, what we're calling out here is that it is fully controlled by the customer. When we deploy the VCF stack, we take all the secrets, all the route access to the ESXi hosts, SDDC manager NSX, put in Secrets manager, and then hand it off to the customer to. Go forth and virtualize all their workloads or migrate from on premises up into the cloud or from other hyper scales and the benefit again is that there's less change control.
14:18
There's customers have access to use whatever tools that they've been using historically, um, you have full route access to the ESXI host so if there's custom vibs or controls or anything like that that you do on premises today, you can do it in in EBS. Um, these are at GA will be and during public preview, I4I metal. Our goal is to certify all of our X86 metal instances so that customers have a
14:47
array of choices when it comes to sizing their environments whether it's CPU memory or disc. Um, particularly when we're partnering with folks like Pure to deploy and maybe a smaller compute cluster and then offload a lot of the storage because obviously VCF deploys VSAN, um, but that may not make the most sense for customers, um, in the long term or if they have a large on-premises infrastructure that's based on pure
15:13
storage. It makes sense to use cloud Block Store to accelerate that migration up into the cloud or maybe have a pilot like DR situation inside of AWS while you maintain your on-premises infrastructure. Uh, consumption models follow the same standard AWS consumption on demand 1 year, 3 year at GA will be bringing your own subscription.
15:36
Um, we are pushing towards the licence included model, uh, as a fast follow, um, and then in terms of choice, you know, if customers don't want to manage this, that's where AWS is gonna lean in with our SI partners to deploy and have managed services on top of EBS. So What's it actually look like if we zoom in a little bit more because there are a couple of
16:00
new pieces that we've had to create inside of AWS to run off the shelf VCF on top of bare metal, um, things like Route Server which went GA. Oh, that doesn't build out. OK. So there's another secret little guy over here. Um, so the route server, we needed something that would allow NSX to advertise routes to the
16:26
BPC route table, you know, it's not something that existed. So last month we GAd a service called Route Server that essentially acts as a BGP pair to our NSX tier zero that we deploy for you and we'll fill in the route table on a customer's VPC. So if you create an overlay segment. It'll pop up there. That way VPC knows where to send those bits,
16:48
right? If someone wants to talk to an overlay, it's one way and it says, OK, go up into NSX, you're happy. The other thing we created are this concept called EBS subnets. When you deploy VMware on-premises today and you create a distributed switch for your VM kernel interface and VMotion, what's one of the things you got to put on there?
17:08
Vance. What doesn't exist in cloud networking? Villains. So what we did is we created some uh new networking construct that can only be deployed and, and, uh, used and interfaced with through the EVS that essentially allows those VLANs to exist. So you have VLAN 1020, 30, 40 up,
17:29
there's a 10 total networks there and we separate those like you would see any standard VCF deployment. You have management, you have VMotion, you have VSAN, you have your TEPS. You have your NSX uplink, which is essentially where all the traffic from outside of the VCF environment will ingress and egress is through that,
17:49
uh, NSX uplink, which goes to the edge pair that we create. And then we have your standard subnets that you have in VPC, and this is how the service itself interacts with native services. So when you have EBS and it wants to create a network, it'll go ahead and interact via these. And again, these are in the customer's VPC.
18:11
So if you want the S3 end point, you want cloud block store, you put it in the EBS subnet. You can icecuzzy up into the ESXI hosts or NVME over TCP. It's all VPC native networking, so it's low latency, high throughput, you know, each I4I host has 275 gigabit ENIs attached to it,
18:30
plenty of throughput to to provide a highly performant network and serve your storage externally. So this is kind of a high level diagram like what we said, you know, you have our cloud block store which Pavel already described, right? You have your virtual controllers, your virtual discs from a customer's perspective, it looks like any other pure array.
18:51
It just happens to be running on ECTU in the customer's account in the availability zone, and then it interfaces with uh it has two ENIs in that private subnet that we mentioned over here. And then integrates with the management of the ESXi hosts as you would see, right? So it's gonna show up as external storage, you know, as an ice ice cuzzy data store and then.
19:17
You're good to go. It's, it's pretty much that simple. You can, you can install our vSphere plugin so because you have full access to to EBS cluster, you can install our vSphere plugin and operate all, all things like uh data store creation, uh, snapshottings and things like that. You can, you can deploy ours yourself.
19:37
Cloud block store has has the same GUI as a. Our flash arrays so because it's the same operating system, so you can directly go to CBS GUI and configure, for example, Iazi connections and VMECP connections and really one of the, I think the differentiator here is really cool is that this really is feature parity from on-premises into cloud for the customer again, same purity OS same off the shelf of
20:06
vSphere. This isn't edited. It's not locked down. This is exactly what the customer is using today. It's the same. Uh, the bits that make everything work and so when you're looking at a customer that has compelling events like a data store or a data
20:24
centre consolidation, a, a new CIO comes in with that cloud first mentality and says go, I don't care why or what, just go. Or, you know, they have a broadcom licencing renewal and once we have licence included, that becomes a moot point, right? So the customer doesn't have to take the time to retrain, they can leverage the same storage experts they've had on premises today,
20:47
the same VMware experts, and just happen to live in the AWS data centres. Yeah. So for the for the benefits and use case already mentioned that, but currently, currently customers are considering what to do next with VMware infrastructure. So Broadcom discontinued a lot of products. It increased, increased a lot of the prices for licencing.
21:11
So there are some, some possible ways what to do next so customers can stay with VMware, they can optimise their infrastructure, reduce the cost somehow, maybe, maybe. Disable some some servers or maybe power up some server. Uh there are some options they they can they can switch some to the new virtualization
21:32
providers. So you, you heard some session about Newtonic port work, some some others, other providers which can do the virtualization OpenSheet for example. Customers can modernise VMs with Cubertis so they can move away from VMware to Cubertis to containerized, but it's complicated. So the fourth option is to move the cloud.
21:56
So this is this is the use case for EBS and together with CBS it can. Bring a lot of advantages in in terms of uh the the storage storage, yeah, and, and to call out, right, like the, the benefit is you've moved into EBS in the AWS infrastructure right now you can take advantage of all those other native services if you want to do modernization with.
22:21
Uh, you know, EKS or ECS on our platform or even, um, you know, we look at those other virtualization providers like Newtanix or Rosa, which are also partnered with us and run in our data centre in the customer's VPC. It, it opens up all the options of the cloud while at least addressing in the short term some of that pain, whether it's data centre closure, you know,
22:44
move to the cloud edict. So there there are some, some uh sample use cases which, which are which are which can be done with EBS together with CBS. So you can easily easily migrate migrate to cloud from on-prem is VMware you can, you can easily migrate to to EVS.
23:03
You can if if you are, if you are using our flash array on prem and the CBS is in the cloud, you can, you can migrate using, using our, for example, replication. You can, you can migrate without, without need of refactoring or changing, changing anything in your in your VMs, so you just need to, you just need to move the volume or you can, you can use HCX for example to to migrate the
23:30
VMs as a life motion. This is, this is possible, yeah, that's a, that's a good call out. I didn't mention that earlier, right? It's off the shelf VCF, right? You have a VCF licence that comes with HCX today, so you can do that layer two stretch from on-premises into the cloud.
23:45
Do replication with your cloud block store and your on-premises array and then you don't even have to worry about changing IP addresses. You can use uh EBS and CBS uh as a disaster recovery. So for disaster recovery purposes, for example, your primary workload is running on-prem. You are, you are doing DR to cloud.
24:04
You can, you can use EVS uh as some environment to if you need some. Temporary capacity you don't have the capacity in your data centre. You need to do short term capacity. You need to increase, for example, at the end of the year some some more capacity in your in your VM VM cluster. You can, you can, you can rent EBS.
24:26
You can connect CBS and you, you are, you are, you are. You can, you can use this environment as a as a temporary capacity for yeah, especially once we have licence included in on-demand, right? Think about those DR tests. You replicate your your storage with CBS. It's living there. You're not paying for compute.
24:43
You spin up your DR test. You spin up a temporary, you know, 4 node BCF cluster, attach to the cloud cloud block store, do your DR test, tear down the compute, and move on. Customer no longer has to pay for a whole separate physical infrastructure, use it on demand in the cloud.
25:00
And We are, we are supporting all those services for, for data mobility so we can, we can do active DR, we can do a sync replication between on-premise and cloud. We can do cloud snaps so you can you can offload offload your snapshot snapshots from on brand flash array to AWS S3, for example, and to rehydrate that to CBS.
25:26
We can do active cluster with CBS, but this is not possible between on-prem and cloud because of latencies. But if you have two CBSs in AWS running, you can, you can do active clusters active clusters between those two instances. The, the for me the, the largest use case and the biggest benefit is I don't already mentioned that that this this VCF there is, there is primary storage for so for customers
25:55
they they need to, they would like to go to to EVS and if they need more, more, or if they, if they have some high capacity uh requirements so they need a lot of storage. There is an example, if you, if you need, for example, 150 terabytes of, of usable storage. You need to, you need to deploy 10 EDS nodes so you can imagine how many it will cost the
26:24
licencing and the VA licencing alone, right? Like 10 years ago I think that you know we all heard HCI was the future. It's gonna take over the data centre. I think in the reality, a lot of folks have come up and realised that sometimes you do need to scale, compute and storage independent of each other, right? And that's where, since this is VCF it's a,
26:44
it's a hyper converged methodology to deploy out vSphere, but at the same time you may need to scale storage independently and that's where, you know, us partnering with Pure. Really takes some of that pain off the customers, especially when you start looking at VCF and VCN licencing include that cost alone, not even the compute,
27:04
just straight up licencing cost. So if you if you don't need a lot of compute power, you can easily decouple decouple storage from compute. You can, you can connect CBS as an external storage to EBS and use it. 11 CBS instance can deliver.
27:21
The same capacity as all those 10 nodes. So, so you are you are you are safe and yeah, and that's the eye for eye at GA right as we certify more metal instances, especially ones with smaller storage platforms, that's really where you'll see CBS really shine as the external storage provider for you. OK, so, so if we, if we just do, do some summary and and key
27:49
takeaways of what is the main benefit and what, what, what, what can, where, where can I help you with so. It can dramatically increase or decrease decrease decrease your your your your your cloud infrastructure capacity. Yeah, so, so this is, this is the, the main benefit for me.
28:19
So it has, it has the same management so it, it, it can be, it can be connected and to your on premise you, you can build your hybrid hybrid environment, uh, use, use CBS as a secondary secondary storage. So, so for, for example, for disaster recovery or for capacity expansion of, of your on-prem environment or you can use it as a primary as a primary environment which is
28:43
running in the cloud. You can, you can use some, some other. Features like like our our ransomware protection, some some data mobility things and yeah, I, I don't think we can short sell the benefit of that operational consistency from on premises into the cloud, right? That's one of the biggest tripping points for
29:06
customers and partners who migrate from on premises. It is a different language operating type. Uh, functional type, everything you do in the cloud is different and having some operational consistency, I one of the slides said restart stalled migrations, and I can't stress that enough is a lot of migrations start,
29:26
everyone hits the ground running and then they realise, I don't know what a security group is. I don't know where to look at my performance metrics. I don't know. Period. Right? And that's where, you know, I, I was, I was a customer before I I joined AWS and I, I lived that life and I remember when the first Pi Ra
29:48
hit my data centre and I no longer had to manage a giant massive multi-rack Hitachi. And it was ridiculously user friendly that I could take some junior new hire admin and say, go create me data stores, right? And so you take that operational consistency you have on premises today and all the innovation that Pure has done with Pier One. And say, you can use it on AWS and you can migrate faster and you don't have to retrain
30:15
all of your staff tomorrow. I, I just, I, I don't want to sell that short because I think it is a massive win when you're talking about cloud migration.
