10 on Tech Episode 010: James Brown on HCI in the Field

Share with your friends


This week I’m joined by James Brown, who is currently a consultant with Nutanix’s professional services division. He’s helping customers evaluate and deploy hyperconverged infrastructure all the time. James spends most of his time where the rubber meets the road – when customers go from discussing hyperconvergence as a concept to deploying some equipment in their own data center. As such, James has a unique perspective on the adoption of hyperconverged infrastructure and insight into what adopters really experience. Enjoy!

As always,  the transcript of our chat is available below the player if you’d rather read than listen. Make sure to subscribe on iTunes, Google Play, or Stitcher

Show Transcript
James Green: Hello and welcome to another episode of 10 on Tech. It’s he James and James show today. I’ve got my friend James Brown, who’s joining us today to talk a little bit about hyperconvergence. He’s recently, in the last year or so, started working with Nutanix to get on board with some of the exciting HCI stuff they’ve been doing and I wanted to get his feedback, from the field. I know he’s doing a lot of work out there with real customers, so I want to hear about what’s going on out there, what tips he has for people who are starting to check out hyperconvergence or are maybe just moved over and getting used to hyperconvergence. Thanks for joining me, James. This is not going to be confusing at all.
James Brown: I know.
James Green: Let’s talk about, as you’re out there, working with either new customers, or prospects, or that kind of thing, I’m sure they have a lot of questions from the get-go. Is this going to do what my old platform does? Is it going to be better than my old platform? Will it have the same features? All of those kinds of things. With regards to hyperconvergence, specifically, what do you say to those people?
James Brown: It’s kind of the interesting part, because a lot of them, they’re sitting there and they’re based in their old three-tier at architecture. They’re either fibre channel, iSCSI, you know, that fun stuff. It’s kind of like, “I’ve ran with this for years, I’ve done this for years, but I’m seeing the new shiny toy out there,” as I’ll call it, but until they actually put it into their infrastructure, put it under a little bit of load, you’ll never know what it can truly do.
  I was actually just at a client yesterday, just flew back in from up in Texas, they were actually running a VDI workload and we’re actually bringing up a desktop. We’re just basically staging one directly out of the box, on another hyperconverged system, and bringing that up and I was actually cutting 15, 20 seconds off the actual, overall, boot-up time into Windows. It kind of goes back to the same thing that I’ve always said is, until you see it, until you test it, you don’t know what you’re getting until you see it.
James Green: Wait, you took their existing desktop image and spun it up on a different platform, changed nothing else, and shaved 15 to 20 seconds off of boot-up?
James Brown: Not even that, it was actually a fresh image without actually going through the standard Windows 10 tuning techniques, and actually shaved 15 to 20 seconds off of it.
James Green: Oh, wow. Cool. One thing I wanted to talk to you about today, again, specifically with regard to hyperconvergence, is the storage bit. There’s still a lot of people out there who are very credible, and I’m not saying whether they’re right or wrong, but they say that the traditional storage array architecture is still the way to go, and that there are advantages to the way that hyperconvergence is aggregating storage, and it’s a good thing, but the standalone, shared storage array is still the best. Could you speak to that a little bit?
James Brown: Sure, I could do that. Basically, we’re still looking at the old three-tier architecture. They’re still using something as a network backbone, as in fibre, or iSCSI, or stuff like that. As you’re reading and seeing these new pieces coming out, with the new all-flash arrays  and pieces coming through, it’s getting the speeds up there, but the piece that we’re going to get to is when NVMe and these pieces start coming out, we’re going to run into bottlenecks between the fibre channels and the iSCSI’s, even the 10 gig link. Eventually, they’ll come out with a 40 gig, but that’s still going to be underpowered, so we’re kind of getting to the point to where it’s like keeping data as local as possible, or actually back on the servers …
  I come from the time where it was actually dumb terminals to start with, when I started my stuff, going into the service, where the powerhouse is, and then it’s like, “Okay, now it’s the desktop.” It’s going into the sand. We’re basically making this complete turn all the way back again, and going, “Okay, we need to actually put the storage back onto the servers, almost into a dumb terminal scenario again, and start all over again,” just because of the advances. It’s always been that the storage has been the complete bottleneck on the motherboard, so with NVMe and pieces like that coming, data locality is going to be a huge thing in the near future.
James Green: Sure. Now, you mentioned, while you were talking about that piece, that currently, and also as it evolves, networking is going to be a big piece to this story too. That’s regardless of whether we’re talking about hyperconvergence or more traditional architecture, either way, the storage networking is important. What are you seeing out there with regard to that, and what do you think is something that people need to be thinking about?
James Brown: I’m going to go from the Nutanix stand point, just because, usually, our deployment goes on two 10 gig nics, for the redundancy piece. I’ll kind of dig a little bit further, but even the true workloads that we’re seeing with some of these SAPs and Oracles, and pieces like that, we’re not seeing the 10 gig links that are being fully saturated, as of yet. In the near future, when we’re starting to talk about these big data pieces, Hadoop, and pieces like that, these mass amounts of data that are coming through and just trying to hit these systems, you’re talking 10 gigs is going to get saturated, the 40 gigs are going to get saturated. I’ve even heard rumors from a couple of my networking friends that 100 gig is actually ready to go. I don’t know if it’s actually hit the market yet or not, but it’s getting very close.
James Green: Yeah, it seems like we always have this competing cycle, in the data center, of which piece is the bottleneck now? Today, is it the CPU, or is it the disc, or is it the network. Next year it’ll be a different thing, and we’re always just kind of shifting from one to the other. There’s some technologies that we’re working towards right now, that might kind of alleviate some of that pain, but, for the time being anyway, it seems like we’re still just switching back and forth from banging ourselves on the hand with one hammer to banging on the other hand.
James Brown: Yeah, all about performance tuning, is really what it’s going to come down to. Once these pieces come into play, to where we get rid of these bottlenecks, then we’ll have to actually start and look at the real cues that … Depending on the people listening, as the develops guys, and the people out there are always going to frown when I say, “Well, we have to go back to the application now and we’ve got to start tuning your application.” Because we’re going to be running on the fastest hardware possible, now we’ve got to tune the application to make it just as fast.
James Green: Let me ask a practical question about that. How do you help them get started with that process? When you say, “We’ve got the infrastructure tuned. I think you guys need to take a look at the application,” are there any tools you provide them, or guidance you give them on, “Here’s how you can start optimizing the way this application is making calls to the database,” or whatever it is?
James Brown: Okay. One of the biggest things, and like I’ve said, I’ve learned from the DBAs that I’ve worked with, and actually, one of my last companies was a huge enterprise application. We were actually running multiple databases in there and trying to go to an all SQL database. With reading and learning, even I was at that case to where I didn’t know anything, but dealing with even Microsoft SQL, learning that Microsoft has an issue with having one drive, basically, writing, even in the virtual world. Creating multiple drives out there, writing for tempdb logs, and data, and logs themselves out there, and also creating directs paths for these, has actually created a performance of 10, 20, I’ve even seen upwards of 120% increases of reading millions of rows of records at any time. Making small changes can also make a huge impact, just on the database connectivity, or the set up, but going into the software and start making calls and setting reference points for these calls to be against, that’s going to also help in the long run.
James Green: Yeah, it’s interesting that you mentioned in that example, it was just changing the architecture of the virtual machine, just a little bit, and that just makes me think about how many various different levels of the architecture where a small tweak could make a big difference. It could be the storage itself, or it could be the storage network, or it could be the hypervisor, or it could be the virtual machine, or the operating system, or the application itself. There’s a wide array of possibilities and even one small change could make, like you said, a 120% performance over the old configuration.
James Brown: Correct.
James Green: Let me jump back to networking, just for a minute. You were talking about saturating those 10 gig links. I used to be a consultant and I have actually seen a lot of similar things in the field. For the customers that you’re seeing, right now, is 10 gig really a requirement, and if it is, how close are they coming to actually utilizing all of that? As a follow on, when you get done answering that, would it be possible to do it with one gig links?
James Brown: Like I said, I’m coming back to what I know, just because I know the Nutanix product just so intimately. We actually do have multiple 10, 15 node blocks that are actually running on one gig links. Is it possible, yes, but it really depends on the workload. Normal workloads, like virtual machines, that I wouldn’t call enterprise apps, but probably small to medium sized businesses, those seem to work perfectly fine. Once you start talking the big stuff … I’ve also got a health care background, so Epic, McKesson, pieces like that that are running huge databases in the background, VDI, Citrix, Zen, going into VMware Horizon the same way, you’re going to start seeing that, because these pieces are going to start requiring higher bandwidth. Just like VDI, when you go out there and you’re streaming video, yeah, it’s streaming the video, but it’s also putting the video on top of that. That’s being done on the clients, too. You’re getting that extra workload there.
  10 gig, as of right now, I wouldn’t consider it a complete requirement, depending. If you’re a customer that is medium to large, we’re talking VDI, 10 gig is pretty much a pretty good requirement, especially with 40 gig being released as the backbone, as of right now.
James Green: Makes sense. Well, as we’re winding down here, I’ve got one more question for you. I’m a family man, and I know you are too, and to be honest, that’s part of the reason why I got out of consulting, is because there comes a point, in consulting, as you’re working with bigger and bigger customers, and high value accounts, they’re more distributed and you start having to do a lot of travel. It was just kind of becoming a lot for me, and I was away from home a lot, and I didn’t necessarily like that. I know that you do a fair bit of travel in your new position, and that’s something that you have to deal with. What are your thoughts on balancing that role and being a family guy?
James Brown: That’s kind of an interesting one, and James knows this, but just for the rest of everybody listening out there, is I’ve actually got six kids. With that being said, most people’s mouth is probably already dropped and go, “Okay, how do you really do it?” The good piece is I’ve got a loving wife that sits behind me and she looks at it and goes, “This was your logical next step. Where you were at in the middle of Iowa, there was probably no where else for you to go.” She’s very supportive behind me. My kids are the same way. They’re exactly the same thing, understanding that I’m probably traveling anywhere between 50 and 60% of the time during the week.
  The good piece is the company that I’m at, I know that travel is just kind of the first start of my cue. I know that going down, that things are going to change. There’s going to be other HCI vendors that are coming out. There’s going to be other bigger and bad things that are coming out. With the company that I’m currently at, if I don’t want to do consulting anymore, there are so many other departments that they’re willing to say, “Okay, if you don’t want consulting, we’ll find somewhere else for you to fit in.” That’s just kind of one of those piece of is, the family life is good, you’ve got to have the support behind you, but the other good piece is as long as you have a company that’s backing you saying, “Okay. Do this for us for right now. As we keep growing, we’ll find you somewhere else to go.”
James Green: Sure. Yeah, that support is critical and not everybody has that flexibility to say, “I don’t really want to do this anymore. Could we do something different?” It’s cool to hear that they do that for you.
James Brown: Right.
James Green: Well, thanks for your time, James. I really appreciate you coming on and appreciate you being here.
James Brown: Yeah, thank you. It was a blast.
James Green

James is a Partner at ActualTech Media and writes, speaks, and consults on Enterprise IT. He has worked in the IT industry as an administrator, architect, and consultant, and has also published numerous articles, whitepapers, and books. James is a 2014 - 2016 vExpert and VCAP-DCD/DCA. Follow James on Twitter

No Comments

Post A Comment

Web Analytics