Datrium BuildDay CEO Interview
Transcript
[Music] hey how's they want to ski towns it from the CTO advisor I'm coming to you day to at least they to of CTO visor interviews at day trium in Silicon Valley what we're doing the VMV Baalbek build a with day troops dvx system amazing opportunity to see the behind the scenes of how to actually install the unit how it integrates in a traditional customers environment speaking of customer environment I'm joined by the CEO of data buyer browse the VP of Product Management to shower hey
thanks for joining me so first off there's a lot of similar products in the market to your solution at least in concept you guys I love it that you've come out with this term old big convergence I've actually seen it in other places in the market people latching on but you can either I've learned over the past couple of days there you go quite different first off what problem are you guys solving with the open convergence concept versus what we've seen in the market wouldn't come
the hyper-converged it's we're trying to simplify tier one so if you and and kind of as a secondary thing we also bundle in data protection in a way that's very deep in our contextual so at the the tier one side that solves three basic pains the first is you know you have a big sequel database and you want predictable low latency very hard for hyper-converged to do yes so kind of scale up server problem second you want a bigger pod to manage you don't want small
clusters you don't want the sort of cluster rigidity of hyper-converged toward you know it's half a dozen monolithic server types that share i/o and i/o intensity you want to be able to mix and match kind of and grow to say 50 or 100 servers without having to worry so much about how the sub cluster is constructed you might third you may you want to pull in any kind of server type you know brownfield whatever makes and match all of these things are classic sort of Tier
one deployment problems for servers hyper-converged doesn't get you there we do so Brian I understand that message is pretty straightforward I have a brownfield I have stuff in my data tailings that I that up they're still depreciating I'm gonna need to use it years into the future I need some type of control plane I can put over it stall this data protection we taught that Devon about the different use cases people have been using the data protection portion of the dvx for when you're talking to
CEOs who are under constant pressure to move to some level of cloud how is that conversation what how did you help to explain the value of date rooms dvx beyond the low-level details of the solution well there's a fundamental question at the top of that of what what workloads are you going to put a sort of off prem into a cloud environment of SAS or you know something you build on amazon and what are you going to keep on Prem that's a sophisticated question and it
depends on the the you know price performance the how much you want to outsource for example the early markets for Amazon or mostly outsourced data center markets and it's a you know it's better than what IBM used to do right but for the for the people that want something on Prem they want it to be as simple as possible because they don't want to have a lot of operating expense in dealing with it and in that spirit a couple of changes have been going on one
is that arrays are declining because you always need storage management as a unique specialty right San Management convergence can resolve that hyperconvergence especially if gives you VM granularity management and that that means you're not dealing with lungs and that's simpler it means as you add house you add speed automatically although sometimes in spotty ways so so that transition has been happening on Prem independent of us that people are moving away from arrays toward convergence and hyperconvergence as that movement happens it turns out their exposures in
that set of opportunities that really hasn't resolved for Tier one it hasn't resolved for bigger databases with consistent latency with you know scale-up hosts with bigger pods all these things are classic enterprise problems you don't want lots of things to manage and you know you take your traditional application and something like ASAP or even I even just a traditional distillation of Sapa SAT Hana which s AP is extremely restrictive on how you run storage compute next to analytic computing so you have to separate those two
out and your open converge system seems kind of perfect for that traditional data center workload that may not go into the cloud at any point right so we're you know focused in this interesting way on these two sets of problems one is how to do a sort of bigger pod much more server flexibility much more consistent low latency but we're also tying it together with built in architectural II deeply built in VM backup and data protection that also replicates to clouds unless you you know store
them so that's a very also a very interesting new thing if you look at the the landscape of data protection there are you know the analysts would tell you about 15 percent of enterprise doesn't use backup software today in three years from now that'll be 30 percent that's an enormous change right that's not well covered but the reason is VM granularity data protection is getting more and more built into on-premise infrastructure yeah and the plug for the CTO visor des separate we're doing a backup is
dead hanukkah's for these very reasons 15% enterprises are not using backup backup is a legacy concept I think in itself so - sure talk to me about what talk to me about what customers are seeing the feedback about the product as they're starting to dip their toe into cloud functionality yeah what's what's the number one use case and demand for feature yeah sure thing great so happy to so I think as Brian said the focus for date trium has been to really simplify the life for
customers and administrators and that's exactly what we are hearing from a lot of our customers my number one comment from customers is I don't have to think about it anymore and so what we find is a lot of customers who were spending there are 10 hours a week 12 hours a week managing their storage are now being freed up from all that time because they've designed something that's super simple to deploy but also simple simple and scalable entities pretty much self managing to a large degree
right from a cloud perspective I think the pillars of what daydream is truly focused on is truly simplifying convergence not only on-prem but also including hybrid cloud as a big element of that convergence and so what we are hearing from customers if they want to keep a long-term retention of copies of snapshots that they are keeping on their on-premise site they can use the daydream cloud software to essentially keep those copies in a public cloud on a paper core basis you know pricing in a much
much more elastic and you know as a service manner they don't have to worry about the cloud we take care of all the difficult aspects for them and simplify the clouds that they can just consume so and customers are very freedom of this responsibility I love the comment that I'll pull Alastair from V brown-bag earlier you want to ask them hey how's the built going and what's the results what's the early impressions you know what you just don't have to manage this office what he basically
said it's amazing in the sense that is freed up from just his installation what he has to manage day two so it's about day to operations so Brian can you talk to me about day two operations what are some of the CIOs and CTOs are talking to what are they come to talking to you about when it comes to day to experience with dvx well it's kind of as as too short service as Alistair said but you know our our it's our demo is kind of
like watching a Corona beer commercial there's nothing to do there's nothing there like storage management goes away even even hyper-converged cluster administration where you setup you know do you want D duper or compression or whatever on a particular shared storage range of hosts that goes away because everything's always on we've built this in such a way that we can be fast and maximize every efficiency technique all at the same time so we just leave it all along we just did a benchmark with IO mark org
and the evaluator group where they the il mark test is a test of how many VMs you can run with consistent low latency in the shared store and and they basically keep adding VMS until it breaks mm until latency goes too high we broke the current hyper-converged benchmark by about 10x we did 8,000 VMs compared to 800 with a I think that was V Sam with obtain what's so crazy of our while we beat the prior best benchmark for an all flash array of the IBM
V 9000 by 5x and this is in our case we did it with encryption in software global deduplication inline compression and a racial coding RF sorry you know to to failure tolerant racial coding which are our defaults all set in all these other in the benchmarks we competed against that was all turned off they did like raid one no data reduction it's a great car yeah it's it's ludicrous how scale it is a stock car but it's not really stock yeah so so taking all performance
management off the table yeah just by saying you know configure your host to whatever speed you need and then you're done that's that's just it breaks a lot of other products so I'm just quickly add to that we did all that with two thirds the latency of those other offerings so even with that performance or latency is were way better than the alternatives yeah I was also impressed with the path size so sure we talked about like just we talked about the large fill your domains
versus the small theory of the base or large parts versus small pods talk a little bit about that build how big was that cluster with not cluster that dvx yeah absolutely so I kind of answered that in two ways so this particular environment was about 128 hosts or compute nodes that's about the size of a dataset that's exactly right and that was coupled with ten of our all-flash data node products and so that was a result we got with that but then more interestingly what we
find with a lot of our existing deployments as well is that we've had you know number of customers who have used over 30 hosts with a single data node in production and so this is a proof point for the actual you know delivery on the promise of the open conveyance architecture we're having stateless host allows us to scale elastic me to a massive amount of degree and even with that provide customers with a reliable and predictable performance that they expect so Brian I'll give you last
word what do you want to leave the CTO senior architect loveable audience what is the last message you want to leave off to them in this broadcast the reason to talk to date 3m is if you want to simplify ur on-prem architecture with a bridge to the cloud and then simplifying your on-prem architecture your infrastructure you want to get coverage of you know all your apps but especially your tier one apps it's a thing we can do better than anybody with a mixture of host types
from you know wherever you want to get them and tie-in complete and robust VM data protection policy based VM granularity or container granularity so you don't have to buy separate backup you know software and storage products with encryption and down so with that said I want to thank you for again tuning in to the CQ advisor daily dose with special sponsorship by date room and thinking the V brown-bag team for inviting the CQ advisor to tag along to their little build a fight scene behind the
scenes even though they left me in a room with the DV X and there was a red cable and I really wanted to pull it I did talk to you next CTO daily dose
