Intel Optane Pmem as a Service- Dell EMC VxRail
Transcript
(upbeat music) >> Hey guys, it's Keith Townsend, principal of the CTO Advisor. You're watching the battle of the beards. No, no, seriously. You're watching another Intel sponsored conversation with Dell Technologies. We have on the other end of the country, Greg Colburn, director of HCI systems at Dell Technologies, Greg, welcome to the show. >> Hey, thanks Keith for having me. >> So Greg, other than beard care, we're going to talk VxRail and we're continuing this conversation. If you're joining the entire playlist, we're coming off of Dell Technologies World 2020 Virtual Event.
And we're talking about VxRail, specifically the value of VxRail over commodity, more commodity systems. And Greg is going to share some thoughts with me around VxRail. Greg >> Yeah absolutely. So what I'd really like to talk about today and a little bit about my background. I'm not an infrastructure person at heart, from my technical background, I actually come from the application side of things. So for me, whenever we're talking about these sorts of platforms, it's really about the applications themselves and how do I support the applications and how do I streamline that?
And really for the application developer, for the end users, the hardware, I want that to be an afterthought right? And I think that's really what we can get at when you start talking about VxRail and what it brings to our customers. So when you think about VxRail and you think about hyper-converged, a lot of times you're thinking general purpose workloads and things like that. But the reality is with the flexibility that we have built into the platform, there are limitless opportunities for us to do very different types of applications, all the way from your traditional business applications, your LPP database workloads, all the way through to HDC, to big data analytics and things like that.
And the reason that we can do that is because of the deep integrations that we've got with some of our partners like Intel. So when you look at things like, you know, the Intel second generation XEON processors that we have, we've got the two socket and the four socket versions available. When we think about things like on the storage side of things, where we've got the Optane SSDs available to us, Optane PMem starts to bring in some very interesting sorts of workload capabilities into VxRail.
Then the cool thing about it from making the infrastructure invisible to those applications and users is that the streamlined nature with which we worked with VMware to integrate VxRail into the hypervisor layer and into the management layer, the management of the environment itself is virtually seamless. So regardless of which workloads are run, I can have a block of VxRails doing a number of different things. And it doesn't matter from a workload perspective, the management of it how I do my upgrades, how do I do my scalability?
How do I expand the systems? It's all the same and it seamless. And you can really only get that in a tightly coupled integrated system that we've developed in VxRail. And what I really want to do is kind of highlight a little bit around the the broad scope of the applications we can support. So I'm going to take something that's on one end of the spectrum all the way down to another end of the spectrum, just as a couple of examples.
So for instance, if we start thinking about, some of the high performance requirement applications, SAP HANA comes to mind right? When we start talking about the in-memory database capabilities and what we can do leveraging that Optane PMem that I just talked about for these high intensity- >> So Greg, I'm going to hit pause just for a second here. I hear this SAP HANA story time and time again, with the Intel Optane PMem story. This video is sponsored by Intel. So SAP has become the poster child for PMem.
Talk to me about more practical things like I'm a SQL shop I've heard the story around SAP, PMem, Intel Optane memory, reduce costs, closer to DRAM type performance, but most of my customers are predominantly SQL shops. They have Microsoft SQL, Oracle databases kind of running the business. Talk to me about the value of VxRail and PMem for those shops. >> Sure, so I don't have a slide on it, but I can certainly, we can talk about it a little bit here off the cuff.
Thinking about SQL server specifically. And I really liked the story, because Microsoft has really evolved what they've done in the SQL server space since 2016. So they started in 2016 and they were looking at in memory databases. And how do we start doing some of the high performance database workloads all the way through to 2019, where they've really got some really interesting things that play very well into the VxRail story, particularly when you start talking about things like Optane PMem and things like that.
So we look at how they use, a couple for instances, hybrid buffer pool. That's something that they introduced really in 2016, they've enhanced it over the past couple of years. But what that allows me to do is focus on read intensive workloads. So if I'm working on workloads where I'm doing a lot of reporting out of, it's not really a heavy write intensive that allows me to leverage PMem, as basically the storage space for those clean pages. So if you think about how SQL server does its reads and writes.
SQL server doesn't do anything direct to disk, right? It does everything in memory, and then it flushes the changes down to the disk as it frees up cycles. So what that means is if I've got read intensive workloads and I'm calling pages into memory off of this, but then I start having any writes that come in, you can end up doing what they call disk thrashing. So where I'm or buffer thrashing where they're going through and having to load things back and forth, and it creates performance issues.
When you think about what PMem brings to the table in that, and hybrid buffer pool specifically, it allows me to flush or keep my clean pages in PMem for access and I can use the faster DRAM for making those changes. So, as I do updates, as I do inserts, as I do deletes, that's handled by one side of the buffer pool, whereas the reads all come out of PMem, and it gives me a much better performance when it comes to read workloads. Even beyond that, when you think about some of the enhancements, Microsoft has made that we actually love, because it makes our platform ideal for SQL server workloads, their support for Linux specifically, when you start thinking about PMem, when you're leveraging SQL 2019 on Linux, you can format PMem as DAX which allows me essentially to keep my database and log files actually on PMem as persistent storage.
So now all of a sudden I'm reading, writing, including my logs on, basically ram speeds, which is absolutely amazing, the performance that you can get out of these high transaction databases. So if you think about an online retailer, that's doing high volumes of workloads during the holiday times and things like that, this is a very real enhancement. And the cool thing for us is one, we support these technologies right? So VxRail, obviously deep integrations with the Intel side, deep integrations with VMware side, but from an operational standpoint, so it's great for the end user, it's great for the SQL DBA, but it's also great for the folks that are operating this because I still, getting back into that, making it transparent to the workload, transparent to the people that are using the system.
I manage it the same way. I expand it the same way, the feeding is the same. If I've got upgrades that I need to do, because of the integrations that we've put into the system there's no specialized knowledge. It's not like, oh, those systems are database systems. Think about like an Oracle Exadata. I don't have to worry about having somebody who knows Oracle Exadata to go do that. I can get high performance workloads on these LTP databases on VxRail and manage it the same way that I manage everything in my databases.
>> Well, I love that you basically complimented the conversation I had earlier with Bill Leslie, you guys hadn't talked about this. I actually pushed him on this very same topic. You just gave me all of the technical detail. You guys should go back and check out that interview that we did with Bill Leslie So I interrupted you on talking about this database, SAP HANA use case. We can skip the HANA conversation, we've had it before, please continue. >> So at the other end of the spectrum, and this is gets to be a little bit more general purpose but I want to bring it up specifically, because of some of the announcements that came out just yesterday at Dell Technologies World.
So when you start thinking about the use case around VDI you know, in our current environment, in our current world, with Corona and everything else, the uptake in customers we've seen looking for expansions to, or brand new VDI based systems has been tremendous. It's amazing. It's amazing how quickly transformation can occur. Really, it's one of those I guess, what do they, necessity is the mother of invention. Is that what they say? It's something along those lines right? So when you look at the integrations that we had in place, that support those VDIs.
So obviously in a VDI world, GPU is a big thing. So we've got a number of different GPU builds. One of the things we see specifically, and this is why I called out at the beginning, when I was talking about the Xeon processors that we offer the two socket and four socket nodes, unlike some of our other workloads, when we're looking at VDI systems, we see a strong preference towards the two socket processors. So when you think about the configurations that we have to offer, we can build out configurations to meet those customer workloads, whether they're, you know, I'm really building out something for an order taking task force and I need lightweight VMs, or whether I need to build something that's a little bit more heavy duty.
Maybe I throw some Optane SSDs in there or things like that for some of the performance if I've got more heavy duty users than just order takers and things like that. But there's flexibility built into the system, both from what we bring to the table as Dell, as well as the integrations that we've done and what we offer from an Intel component standpoint. And the coolest thing about this is when we start thinking about what was announced yesterday. So you've probably seen over the last couple of days, there's been a lot of talk about project maintenance.
So this whole, you know, basically everything as a service, you know, you start thinking of infrastructure as a service, how does that work and things like that? Well, from a VxRail standpoint, we just announced yesterday that the Dell Technologies Cloud Platform, we are going to start offering this as a service capability, but it's going to be focused on workloads, which is fantastic. So take a step outside of Dell Technologies for the moment and think about when I go to a public cloud provider, whichever public cloud provider it might be, they've got a menu of offerings and the offerings are typically based on workloads, right?
I can order a VM of X size to do this certain thing. It's not highly customizable but there's enough offerings that I can typically find something to meet what I need. It allows for that rapid provisioning, it allows for the ability to get up and running quickly without going through these expansive sizing or sizing exercises and things like that that we typically see when we're buying things from an on-prem perspective. Well, now we are going to have the ability to go in to the Dell Technology Cloud Portal and essentially order things based on workloads.
So I can go in and start looking and say, hey, I need a system that will run, you know, 500 VMs that look like one of these sizes, click go. From the time I click go and give a little bit of information about how I want the licensing to look and things like that, from that time to I'm up and running with the capability to run those, 14 days. No deep dive sizing exercises. No, it's, and we're having offerings, not just from a VDI perspective, but from a HTC perspective there's a lot of LTP offerings.
But I'm going in and I'm ordering based on the workloads I'm running. How many of those workloads that I'm running, it's not a full-blown, here dump the performance data from the last two years, analysis, most customers usually have a handle on what a lot of their workloads look like. And it gives them an easy button to subscribe to that, you do that as a service on-prem. which is- >> I'm looking forward to hearing more about APEX, it sounds like it's a higher level abstraction, or a higher level of consumption.
So, you know, if I want VDI as a service, and I want that on premises, which in today's time is, like you said a necessity. If I've gone into the public cloud, which we did this in the CTO Advisor data center, we used a hybrid cloud model to expand out to the capabilities that we need during the pandemic. And then as we think about long-term, you need to pull these workloads back in but there's the uncertainty of what that demand will be, APEX and consuming VDI as a service from Dell on top of VxRail seems like a good compromise between those two environments.
I'm looking forward to finding out more. >> Yeah, it's awesome. And I'll get a link to you after this. The session was a really short one, about 10 minutes long but really packed with detail on what those offerings look like. And really that's what I really wanted to come talk about today. When you look at VxRail from an application perspective we can provide a cloud service that is performant for virtually all workloads. We've got a number of different configurations, whether you have a bespoke use case that you really need to go do those deep deep dive sizing exercises on, or whether you're looking at something that's going to be more cloud consumption based, where you're ordering by the block of workloads.
So I want 50, 60, 100, 500, you know, instances of this kind of VDI. We can do that. We can meet the customers where they are from a workload perspective, and we can do it all while keeping that same management layer keeping those same infrastructure upgrades, and expansion and scalability. It's the same across the entire environment. So from an enterprise architect perspective it's glorious, I don't have to maintain silos of different ways of managing things because they're all different. It's all the same.
The workloads are what's different and that's what's important anyway, >> Greg, I really appreciate you coming on we appreciate until Intel sponsoring yet another CTO Dose. I'd love to hear your feedback, is Dell providing VDI as a service or just everything as a service something that you want to see? Is that meeting your needs when it comes to your hybrid cloud infrastructure? com. Talk to you next CTO Dose. (upbeat music)
