Grey Beards on Systems - 51: GreyBeards talk hyper convergence with Lee Caswell, VP Product, Storage & Availability BU, VMware

Episode Date: August 25, 2017

Sponsored by: VMware In this episode we talk with Lee Caswell (@LeeCaswell), Vice President of Product, Storage and Availability Business Unit, VMware.  This is the second time Lee’s been on our show, the previous one back in April of last year when he was with his prior employer. Lee’s been at VMware for a little over … Continue reading "51: GreyBeards talk hyper convergence with Lee Caswell, VP Product, Storage & Availability BU, VMware"

Hey everybody, Ray Lucchese here with Howard Marks here. Welcome to a sponsored episode of Greybeards on Storage monthly podcast show where we get Greybeards Storage assistant bloggers to talk with storage assistant vendors to discuss upcoming products, technologies, and trends affecting the data center today. This Greybeard.Storage podcast is brought to you today by VMware and was recorded on August 14, 2017. We have with us here today, Lee Caswell, VP Product Storage and Availability. So Lee, why don't you tell us a little bit about yourself and what's new with vSAN? Well, thanks a lot, Ray. And Howard, it's a delight to be with you guys and to be on a gray bear show.

It's always a pleasure. So it's been such an exciting year. I've been at VMware now just over a year. And to watch what's happening with vSAN first, right, just this unstoppable movement of taking basically servers and using those to what I call serverized storage as a starting point. And then how you go and build out the modern data centers in there.

One of the really interesting things for me that I've found is that a lot of customers are looking at, they want a modern data center, but they're trying to figure out how to fund it. And so in this environment, it's like, how do I go and buy $300? And so one of the things, the reason I think why vSAN has been moving to the past, through the VMware Reseller Channel and through all of our administrators, is that it's resonating. It's very similar to what happened in the early days where people consolidated servers at care number, right,

as a way to go fund shared storage, right? Have a little money left over. What they're doing right now is they're using vSAN in place of shared storage, and they're using that as a way to go and fund their path to the modern data center. And that's why this market is taking off so quickly. You're saving money and getting a better alpha. Make sense? Yeah, yeah. So you're getting a lot of adoption of vSAN then.

Well, it's the fastest growing market in IT today. It's growing faster. Now, it's not as big as the cloud, obviously, but it's actually growing faster. If you look at it, we're growing 150% year over year right now. And the market itself, it looked like it doubled last year, growing about 60% this year. I mean, markets slow as they get bigger, but these are still stunning growth numbers. We're still adding over 100 customers a week, which in storage terms, you know, that's big numbers. But remember, we have 8,000 customers now,

and VMware has 500,000 customers. So we're still less than 2% penetrated in the market. God. Yeah, I mean, stunning, right? By market there, you mean the hyper-converged market, right? Hyper-converged market, that's right. Yeah, I mean, I find it interesting that the hyper-converged markets become significant enough

that vendors are starting to come out with systems that fit HCI kind of only if you stretch the definitions so that they can call them that. It's a little bit like web, remember web 2.0 and, you know, cloud washing and things like that, right? Well, I mean, my favorite was software-defined storage where at least one vendor said, oh, yeah, the product we've been selling is hardware for 15 years. That's software-defined storage. Okay.

Well, here's an interesting thing, right, is the move, and I certainly saw this. I came here from that app right in the words, and a very good storage company. But what I did notice is that when you move from a hardware business to a software business, that's not to say that there's not a lot of the value in what is a storage company in the software. But the move to a software business model is striking. It's not a little bit different. I mean, it is a massive change. First, the numbers are a lot smaller when you start. Just a software component in HCI is about 25% of the overall value of the product. All of a sudden, that means you've got a business which is like, hey, I'm doing all this work,

but now I'm actually getting just the software piece of it. Now, for investors, what happens is the benefit is you get a much higher multiple on software companies than you get from hardware companies. Well, because you have the margins. The incremental cost of software is zero. Exactly. And so this is an interesting dilemma for companies that built a hardware business. It's very difficult to move from a hardware business to a software business. So what we're seeing is companies that are building sustainable business models like VMware and Microsoft are able to go and use the hypervisor footprint that they have, add in the storage components, and then have a business

that financially makes sense at the end of the day with, like you said, higher margins, but lower overall aggregate dollars. Make sense? Yeah, less top line, same bottom line, because you have higher margins. Right, right. People who don't get the difference between top and bottom line get confused. Well, what we've done over time is markets tend to disaggregate parts of the value chain that you can go and get more effectively someplace else. I mean, that's basically the knowledge. And so in this case, the server hardware, one of the things we've been able to do is we we are integrating with server hardware but we don't

have to go and manage any of it inventory relationships and all that i don't have to worry about flash pricing i don't have to worry about you know what's happening on on any sort of like just in time sort of thing on the other hand right what we've got is we've got this huge influx of new hardware if you love hardware i, HCI is the place to be. I mean, you can flash and you can use it. I mean, it's all there. Oh, and you guys have done a nice job recently with moving some of the server hardware management

into your update process. Well, I've always hated HCLs that said, you have to have this version of the firmware on the HBA, and this version of the firmware on the drive, and then left you alone for and how do I update that stuff? I have to stand in the hot aisle

and boot from a USB stick and run DOS? Exactly. Well, it's kind of funny where we have this integration that looks a little bit like the early days of x86 servers, like the BIOSes. This whole idea that, you know, how do you build together a system, right, that is a storage system that's going to provide the level of reliability and predictability

that you'd expect out of a battle-tested fan. And doing that with combining software and hardware, one of the things we did find out, by the way, is there are customers and partners who prefer to buy an integrated appliance where all that work is done for them, and there's an integrated lifecycle manager. And that's why we went and jointly built the VxRail product with Dell EMC. And, you know, it's our belief, right,

that every server company right now is leading with their value option in HCI where they're like, hey, let me pull all the pieces together so I can get, like, the most integrated product. And, you know, at Dell they're leading with VxRail, which is based on DCN. Cisco's doing the same with HyperFlex. We've got HPE doing this, you know, their SimpliVity appliance. And Lenovo's a little confused. Well, stay tuned here, right? Because what you're finding, right, is the appliances

that are a piece of the market, all of those players that we mentioned, right, also have a vibrant vCN Ready node business today. Yeah, I would say the vCN Ready stuff is the other side of this game, right? Yeah, so the build part, you know, for anyone who wants to build, you know, my experience right now is that there's

buyers who say, you know what, I want the integrated appliance. Then you have like a big retailer that stumbles on a thousand store opportunity or, you know, a 500 node data center environment. And then what they do is they come back and they say, I'm going to evaluate the appliance, but I might want to go in. Maybe that's a HP shop. Or maybe these guys could go in. I know exactly what the workload looks like.

And then they start to tune this, right? So there's going to be builders and buyers, in my experience, right? We'll see. I mean, that's why we're continuing to work really closely with HPE, with Lenovo, with Cisco even, because we're able to go and drive integrated news to servers with vSAN for people who want to build solutions at scale. Yeah, it's funny how HCI is, in some ways, the pendulum swinging back to direct attached storage.

Oh, God. Well, if you look at, you know, outside the large systems world, if you look at the Exchange and SQL servers and Windows file servers and web servers of the world, before vSphere, all of those servers ran off their internal drives. And there were years where the big difference between a server and a desktop was it had a RAID controller and 12 drive slots not the processor that's right and then vSphere came along and we saw vMotion and we all fell in love with vMotion

and vMotion required shared storage and all of a sudden we started buying Equalogix and Clarion's to put vSphere on so that we could have vMotion. That's right. And now, well look, it's going back. Storage is moving back into the server. Now we've got a storage layer that's a lot smarter than Veritas Volume Manager that's turning that shared storage into a common pool for all the servers to use. But from the physical point of view, it's really similar.

It really is. And what it does is it allows you, you know, because you've got Flash, right? I mean, what made this possible? Flash allowed you to have caching across nodes that approximated the performance of a scale-up traditional system.

And 10 gig ethernet gave us enough bandwidth and low enough latency. And now when you take that and then you add in the new CPUs coming out, right? So we're partnering with Intel on the new Skylight Corollary or the marketing team,

the unscalable processor. And we're using that next server refresh as a way for customers to say, yeah, you might be thinking about a server refresh, but why don't you think about supersizing that with UCAN so you can build HCI. And now, you know,

people aren't throwing away their SANs. I mean, we don't recommend that. But what they are doing is they're saying, you know what, if I just want transactional storage that supports vMotion and I can start putting this at scale and it's going to be faster, not the same, but faster than my existing hybrid system using all Flash, I mean, why not, right? Save half the cost.

Manage it with my vSphere administrator. It's a very compelling argument. It is, and my problem is, of course, people take that argument and go, and it'll work for everything. Yeah. And there are times when I really would rather have VVols than vSAN because I'm a firm believer in different horses for different courses. Well, you know, some customers ask me, they say, well, I hear what you're saying.

Tell me when you wouldn't use vSAN. Yeah. Yeah, that's a good question. Yeah. Tell me why. Let me give you some examples of places where storage as we know it can be and should be managed separately.

One is if you've got a big archive, for example, inactive files or largely inactive files that you want to have basically be scaled independently of any compute, it makes a lot of sense right i mean that's a great market for a filer right or something where you say i really want to go and have something where i'm managing something other than a vm or if i've got a specific application with a

vm or or a even a bare metal server where I want to go and scale performance by line specifically, right? And be able to have the knobs and really say I've got someone who I am going to pay to be able to manage the storage directly. There I'm scaling capacity or performance, again, independently of a VM. Now, I mean, these are areas where, you know, there's storage-centric discussions and storage-centric management. I think those are going to last for a long time.

But what we're finding is that wherever you've got business moving fast, where I've got to basically scale up, scale quickly, change my environment fluidly. Those environments, boy, HCI does a great job of using storage generalists. And as you pointed out, the thing you need to go and make sure is you've got all the management tools that you'd expect so that these generalists can now go and manage, debug, review the storage attributes that up until now have been looked at by specialists, right? Yeah, and to some extent, we have to imbue the virtualization generalist

with the paranoia of the storage guy. Yeah. So where do you see something like HCI in the overall scheme of investments and data centers with respect to the cloud and of investments and data centers, you know, with respect to the cloud and other options and that sort of thing, Lee? Well, here's how I'm viewing, you know, and Dell EMC is actually an interesting company to watch because when you watch how they're moving their sales motion, you know, up until last year, they were just saying, hey, let me go and talk to you about SANS,

and then I'll fall back to telling you about HCI if you've got a cost problem. I'm paraphrasing, right? But now what's happening instead, if you look at how CanSafeCash for example is blogging, is they're leaving with HCI, and then they're basically using storage,, sand storage, right, as the exception. I think it's really interesting because what happens is, as Howard pointed out, somebody is going to go look and say, hey, listen, like this is an environment where I really, really need to have sand-based storage. And great.

Now you've got someone who's both credible or, you know or able to go have that discussion, able to go manage a different solution. And there you'll have the storage elements come together and you can go manage it that way. But what we're finding right now is that customers who are thinking at all cloud-like, which usually means flexibility, not low cost, but flexibility. The HCI environment is one that appeals very quickly, and an investment strategy looks like this. I used to take five years to go and scale my storage, figure out what the scale looked like, put together that five-year time frame. It was a nine-month selling cycle, install cycle. And once I was done, I started planning for the next one. HCI is very different, right? I can get in for half the cost, which means shorter sale cycle, and the risk of

being wrong is actually pretty low. Let's say I misconfigure this. Well, add another box. I can add more capacity, more performance. It's a really simple way to go and say that, you know, the risk of being wrong in storage has historically been very high. And as a result, these systems, right, have largely been overprovisioned. People overbought like crazy. No. They did. Yes. And so now you come back and you say, hey, I can go and buy like a smaller amount and then I'll just go scale it as I need to because it's simple.

I just add another box, and guess what? These boxes are five figures instead of seven. Having spent years working with folks that did scale-out storage, the problem we always found was the finance guys. It's like you have $4 million to buy a vmax every three years and i know you could spend a million dollars a year and save a million dollars but that means we have to move money from this year into next year and we don't know how to do that you know and so what's happening in the field as a result is we're seeing fan refreshes pushed off and using HDI to go

fill the gap in the meantime. I don't think I quite answered Ray's question about the cloud because what we're finding is as storage guys, I mean, we know the only strategic decision about the cloud is the data. LVM in the cloud

or DevOps is a transaction. It's a credit card transaction. You do it, you cloud, right, or DevOps, in a sense, is a transaction, right? It's a credit card transaction. You do it, you turn it off, and no harm done. Yeah, but transferring my four terabyte database to AWS takes a week. Exactly. I mean, I was with some very non-technical people, and everybody understands this. They say, hey, listen, you know, email me a video and you know what I'm talking about. And people know. They're like, oh, wow, you're right.

It takes a long time to move data. It strategically matters where it lives as to whether I can access it quickly. We know about sovereignty issues where it matters a lot about who owns that data. Who has the rights to the data? And then from a performance standpoint, you can get IOPS out of the cloud, but they're expensive and they're not guaranteed. So when you start thinking about

how to engage with the cloud, we're looking at the way to be relevant in the cloud is to have a common software stack that runs on-premises running above. And that's precisely what we're doing with AWS. I announced this last year.

They tuned very exciting VMware coming up with our VMC product. So we're working jointly with Amazon. It's not just a one-hand clapping. This is a joint integrated development about running the full VMware software stack all the way from vSphere to vStandard NSX, all of that working together in a way that takes care of the updates and is simple, pushed out to customers where you can now get the multi-tenant access to the same infrastructure that you've got on-premises. And it's the same data structure.

So you've got all of the ability to have a common data services across the hybrid cloud. That's how you'd be relevant. I don't think anybody knows what the, you know, what this path to the cloud is, but depending on your, on how aggressively you want to engage the cloud, we've now got an option for you to go and have this all the way from your on-premises tier into the 8,000 partners we've got, and then into the leading public cloud in the world.

So, Lee, the vSAN in this environment operates in AWS with AWS Backing Storage and AWS Compute, that sort of thing? So, here's how the first implementation, right? You know, stay tuned for the, you know, we'll be announcing this, you know, coming up at VMworld. And so all the details will be coming out about how it's actually implemented. But the

storage is vSAN in this case. And so we've got the full stack now running natively on AWS equipment and then being able to be accessed or purchased on a consumption model by VMware customers. Yeah, the one funny thing about the cloud from the infrastructure's guy's point of view is we know less and less about applications. And so flexibility becomes more and more important. True, right? I mean, this is one of the reasons HDI by itself hasn't been widely adopted in the cloud to start.

Because in the cloud, you actually have a very good idea, generally, of the applications you're going to build your infrastructure for. Look at YouTube, for example. You have one application. You can build infrastructure for a specific application. Where HCI has really done well is when you have mixed workloads. And especially with Flash, you have workloads that require capacity, performance, they're working together. That is an awesome environment for being able to have a mixed use case environment for enterprises that want to have the same access that they have on-premises.

In fact, I might go so far as to say this, right? I mean, the real, what I think of as the really interesting thing for customers to look at over the next two to three years is we have two very different views on emerging, right? VMware is taking the complexity of what we have on-premises, simplifying that, integrating that, and then making that available in the cloud. That's a way for all the developers that have been relevant on-premises to now be relevant in the cloud. Microsoft's taking almost the opposite view. Microsoft's coming from a cloud perspective and trying to bring Azure and putting something that has been cloud-like, now putting that on-premises.

And the problem is that the on-premises version of that doesn't have nearly the knobs and tools that developers on-premises are used to. And so these are the ways, you know, it'd be interesting. These is how I see it. It's interesting viewing that, you know, in terms of both companies' founding stories and histories. It is, right?

Microsoft wrote compilers. They were a developer company. And so it's, we're going to create this new platform for developing cloud applications, and you can run it in the cloud, you can run it in your data center. Right. You know, and VMware is, you know,

we're an infrastructure company, and we're going to give you the infrastructure you know and love in your data center, in the cloud, so you can keep using it. And so we're, it's precisely how I view it. Now, what's interesting, I think, is our relevance for developers has increased as we've gone and embraced containers, as we've gone and increased cloud-native apps, as we've worked with typical cloud foundry.

And now we're taking all of this developer, we're adding that developer relevance. And distinctly important is we're committed to this being a multi-cloud offering, cross-cloud we call it. And so as we look at the Microsoft competitive environment, that looks to me like that's going to be a single integrated stack of one cloud. We'll be coming at that with a cross-cloud effort where our view and certainly our data shows that the majority of

enterprise customers will be embracing more than one cloud. Well, Microsoft's attitude dating all the way back to Bill has been we just want our fair market share, and that's 100%. All right. There you go. All right. Well, gents, this has been great. Lee, is there anything else you'd like to say to our audience

before we talk off? We're really encouraging our users and VMware enthusiasts to look at the next server refresh as an opportunity to engage HCI. Find an application that you would like to go start. A lot of our customers start off in test and dev or management infrastructure.

DDI is a good use case, right? They're outside the data center. But very quickly, right? Nearly 70% of our customers are using HCI and vSAN, whether it's a vSAN ready node from all of our 15 server partners or VxRail. And they're now using that as database or business critical application infrastructure. And so, you know, find a way to get started because this level of flexibility and the

ability to fund it using the savings off shared storage not only comes around once in a while and, you know, it's a great opportunity to go and show how you can be relevant with low risk. So I look forward to seeing everybody at VMware. Howard, any last questions for Lee? No, I'm looking forward to the big party in Vegas as usual as well. As we all are. See you there. Well, this has been great, Lee.

Thanks to you and VMware for sponsoring our podcast today. Terrific, Ray. Thank you so much. And Howard, always a pleasure. On our next monthly podcast, we'll talk to another data center technology person. Any questions you want us to ask, please let us know. That's it for now. Bye, Howard.

Bye, Ray. Until next time.