Grey Beards on Systems - 0100: GreyBeards talk with Colin Gallagher, VP Dig. Infra. Prod. Mkt. @ Hitachi Vantara

Episode Date: April 21, 2020

Sponsored By: We have known Colin Gallagher (@worldc3), VP, Digital Infrastructure Product Marketing at Hitachi Vantara, for a long time and he has always been an all around smart storage guy. Colin��s team at Hitachi Vantara are bringing out a brand new, midrange storage system and we thought it would be a good time to … Continue reading "0100: GreyBeards talk with Colin Gallagher, VP Dig. Infra. Prod. Mkt. @ Hitachi Vantara"

Hey everybody, Ray Lucchese here with Keith Townsend. Welcome to another sponsored episode of the Graybeards on Storage podcast. This Graybeards on Storage podcast is brought to you today by Hitachi Vantara and was recorded on April 17th, 2020. We have with us here today Colin Gallagher, VP of Digital Infrastructure Solutions Product Marketing at Hitachi Vantara. So Colin, why don't you tell us a little bit about yourself and what's new at Hitachi Vantara Digital Infrastructure. Thanks, Ray. Good to talk to you again. I am a 15 plus year storage industry veteran. Came out of business school, went into storage, told myself I was going to stay for two

years and ended up staying for a decade and a half or more. But I'm excited to talk about what we've got new at Hitachi Ventura about major updates to our mid-range storage portfolio, both in terms of hardware and software and purchasing options. Yeah, so tell us a little bit about what's new in mid-range. I'm proud to welcome into the VSP storage family the new virtual storage platform E990. It is an all NVMe high performance storage rate, really designed to help businesses accelerate mission-critical applications and take that step affordably into NVMe. It's pretty unusual to see mid-range with all NVMe

configuration these days, don't you think? Are you seeing a need for it? I guess that's the question I should ask. Yeah, that was one of my first questions when I saw the announcement. I'm like, wait, hold on. Mid-range system, all NVMe? That seems like a mismatch. Not really. I mean, I think NVMe has sort of started out as a high-end technology, but it is migrating down, much like Flash migrated in 2009, 2010, 2011 down as well. So we are seeing customers in the mid-range looking to deploy it. A lot of it is about future-proofing. Customers know that NVMe is where the future of storage is going to be,

at least for the foreseeable future. And so they want something down that way. I will admit that the E990 sits at the high end of our mid-range portfolio. So we do know it's what we call the bridge range that sits between a you know, a medium to large scale midsize customer and a small enterprise customer. So it is, you know, it is coming down. But to your point, it isn't, you know, it isn't something you're going to put in a remote office branch office, you know, to support a terabyte or two. So does the solution have pretty much the same levels of functionality as your bigger solution systems and stuff like that? Yeah, that's one of the joys about working at Hitachi Ventura is that we have a common storage operating system and management across the storage portfolio,

be it our small 2U remote office or edge data center products versus as same as the VSP 5000, which we introduced in October, which is sort of the most powerful, scalable enterprise storage around, all run the same storage virtualization operating system. And what we were able to do with the E990 is we took the NVMe enhancements we made for the VSP 5000 in October, as well as a bunch of other enhancements, and we brought those down to a dual control architecture to design the E990. And so talk to me a little bit about performance. And NVMe has got serious performance levels of functionality, right? Yeah.

The E990 delivers tremendous performance and predictable performance with NVMe as well, right? This part of NVMe is not just about the total IOPS, but it's about delivering consistent latency as well. And so we support, we deliver up to 5.8 million IOPS with the VSP-E990 and as latency is as low as 64 microseconds. 64 microseconds is better than any latency out there in existence, right? It is.

It actually, it's slightly better than the VSP 5000 series from which we took the architecture, but we get a little bit of extra performance boost when we only have to focus on two controllers instead of the full scale-up. God, this is impressive. Talk a little bit about some of the changes that had to go in in order to support.

It seemed like there was a lot in the cache space. So, yeah, so there are a bunch of cache enhancements as well as a bunch of other resiliency enhancements that we did in the E990 really to help maximize availability and performance simultaneously. We eliminated dedicated cache boards and caches now distributed across the individual nodes for faster access. We also eliminated write-through modes so we can keep latency latency sensitive applications online um at all times and a bunch of other things so for example we have we've increased our flash rebuild times by 80 versus our previous mid-range systems so this is fat faster flash rebuild is that is that what we're

telling me particularly yeah 80 did you say 50 or 80 80 80 okay and that you know some of that's due to the sort of faster devices obviously but there's other functionality that's been introduced in order to do that, no doubt. No, totally. And a lot of, I mean, we did a lot of cache optimization. We did a lot of code optimization for performance and availability with the VSP 5000. And we took a lot of those improvements and brought them into this platform. So, Colin, when I think of the VSP 5000, I was a former VSP 5000 customer in a previous life. And I think of a really, really big box.

And when I think about this mid-market array market, I think of much smaller boxes, you know, around 4U. And the capabilities that you're talking about, I usually think of in these larger boxes. Talk to me a bit about packaging. It still matters. Well, it does. And I think that, you know, there are particularly as we move to, you know, more and more digital business, I'm thinking of the environment we're living in today. Most of us are operating entirely digital now, for better or for better or for worse. As we move to more digital business, performance becomes more and more important because that's what drives a consistent and superior digital experience. And so customers who might have said, I can get by without top-notch performance in the past are now actually demanding it because that's what they're basing their business on no matter what size customer they are and so by delivering that performance but

in a smaller package with some of the things that a smaller customer may not need like no mainframe support but it allows those customers to to deliver that consistent and superior experience without having to to sacrifice and so that that smaller package, does that include a certain number of drives? Did you mention the size of the unit? I don't know if you did. He did. He said for you, so it is for you. Sorry.

No, I was questioning whether or not it was for you. I'm thinking, am I going to get all of this in for you or not? It is for you. We do support an expansion, Jesse, as well, but it starts at a for you. And yeah, I mean, it scales from as low as six terabytes to up to 16 petabytes of virtualized capacity and that virtualized effective capacity. And we can start as low as four or eight drives and then scale all the way up to 96 within the box and then add additional virtualized capacity. Because, again, all of our DSP products support storage virtualization. So, Colin, talk to me about, you know, total footprint, because this matters in this mid-scale market.

I don't have racks and racks of space. How much can I get in a single, like, rack? You can get up to 96 NVMe drives in 6U. Oh, wow. Yeah, and you can make the marketing guy do the math calculations right now on their call. So what is it, 10 terabytes each? So that's almost 960 raw terabytes.

Is that kind of number? Yeah. And then talk to me about data reduction. There's something about data reduction in here, right? Yeah, totally. We have what we call our advanced data reduction available on this platform as well. Again, this is technology we designed for the 5000 series and brought here. As we all know, NVMe isn't the cheapest storage media on the planet.

And so you need efficient data reduction to help make it more affordable. And so what we have is a really interesting and smart data reduction capability that allows us to leverage AI to decide how to do the best dedupe. When you design a storage system, a lot of times you have to choose which algorithm and how you're going to do it. We actually have a selection of algorithms we can

apply in different ways to the data as it comes in. So the data is written to cache. It is compressed immediately. Sorry, give me a second. And then the AI kicks in and decides whether to do inline, post-processing, or offload processing for the deduplication. Yeah, yeah. So, I mean, when you say best, I'm thinking best compression, best deduplication, but what you're talking about is best performance. Well, the best balance of performance and efficiency, right? Because, you know, if you take, if we have a highly, you know, highly random, you know, workload that comes in, right, that could, you know, trying to do that inline could slow the system down. So that's something that's best handled

in post-processing, and our algorithms will detect the randomness in the I-O stream and shift that to post-processing. Talk to me. So, I mean, Hitachi has always been known for rock-solid reliability and availability, bar none, and that sort of environment. Does this mid-range solution inherit a lot

of those characteristics? Absolutely. It's built on the same architecture as we've built our other platforms, you know, and it leverages the same software as well. So we bring our 100% data availability guarantee to the platform as well. It's a six nines platform designed for high availability, and we guarantee that for all of our customers that, you know, we will not lose any of your data. Again, I think six nines is like enterprise class level stuff here for mid-range. This is pretty impressive. Yeah. Again, it gets back to, you know, the expectations of what a mid-range customer needs have changed over the years, right? You know, it's, you know, it's, it's, it's, you know, Clayton Christensen's innovators dilemma, right?

You know, the needs always continue to scale up. And so that's what we're seeing happening in the market again today is that not only the demanding performance and availability that used to be reserved for the, you know, the highest tiers. So when I went back in my SAP days, we usually look at these mid range systems and we throw my nose up to them for these very reasons. I couldn't get five lines out of them, let alone six nines. That's an amazing claim. Talk to me about how Hitachi Ventura has been responding to this expanding market, not just from, I'm going to throw a curveball at you, not just from the storage capabilities, the speeds and feeds, but helping customers take advantage of this because you're pushing a product down into a mid-range market that's going to get the attention of, you know, application teams like SAP or big Oracle heavy databases, enterprise class services,

I may not always have kind of the skill set to take advantage of this. What are you guys backing up from a services perspective? You know, so we deliver a full set of services. And I think that's something that, you know, distinguishes Hitachi Ventura from some other competitors in the market who sort of brought NVMe out there

is because customers do need a full set of services capability. You know, not just installation, but consulting services, migration services to get on the platform, right? You know, it's not just about buying the box, it's about getting your data from your old box to it. Yeah, I've found that we've had the same boxes side by side for years, because that people underestimate how hard it is to actually migrate the data from one platform to another. Yeah, so Colin, could you talk a little bit about if have you made enhancements

to the operational characteristics of these solutions? Yeah, so we announced Hitachi OpsCenter alongside the VSP 5000 in October. And Hitachi OpsC Center was a focus for us on a new way of managing your data storage infrastructure. And I'm using those words specifically, and I'll explain why in a second. But, you know, because we announced it, it applies to all of our virtual storage platform arrays.

But because we announced it alongside the VSP 5000, people assumed incorrectly that it was only for enterprise customers. So as part of this, we were talking about Op Center again. There are some improvements to it as well. But what we really want to share is what Op Center brings to any customer, including mid-range customers, is the power of AIOps. And what AIOps allows you to do is to leverage machine learning and AI to better manage not just your storage arrays, because we certainly deliver that as well, but the entire data path. And so OpCenter is designed to, it has your basics, it's a suite of products of four. It has OpCenter Administrator, which allows you to do

your basic element management. It has new to the suite OpCenter Protector, which is our copy data management capabilities. And those are particular to the storage, but we also have in it Hitachi App Center Automator and Hitachi App Center Analyzer. And those allow you to automate the entire data path as well as analyze and monitor the entire data path. And so those are really cool things that allow customers to do things like manage a petabyte of storage with a single person, right? Because you're no longer being able to having to fine tune manual knobs within a system and configure zoning separately and configure, you know, virtualization or monitor virtualization separately. You can do that from all from one console.

So these sorts of things eliminate a lot of manual steps and configuring solutions and stuff like that. They eliminate the majority of manual steps. You know, the example I like to give is, and again, this is something that applies to mid-range customers, you know, configuring a shared data store for ESXi. Typical activity, right? Yeah, typical activity.

Typical activity, we've all done it. Even the marketing guy did it briefly. I realized I wasn't the best administrator and got into other things, but I'm not the best coder either. That's another failed career for me. But there are multiple steps, and it's not just provisioning the storage. You may have to do zoning work.

You may have to make host configurations. There can be up to 54 manual steps in doing that work, and it's spread across multiple different vendors' infrastructure. Maverick and storage and VMware, et cetera. work, right? And it's spread across multiple different vendors infrastructure. Fabric and storage and VMware, et cetera. Exactly, right? Yep. And what we can do with App Center is we have, you know, 50 plus built-in templates for automation with App Center. One of those templates is for provisioning an ESXi data store.

And so by executing a single command to leverage that single template, we can provision that entire path and eliminate all 54 of those steps. Not only does it speed up the process, but it makes it more error-proof because every manual step you add to a process increases the chance for human error. So Colin, APIs are all the rage these days. So, you know, me being able to perform this provisioning via Terraform and all the cloudy terms and tools. Is there open APIs to be able to access these capabilities? Yep. We provide a full set of APIs,

sort of both, you know, northbound as well as east-west. You know, how do we manage that, the monitoring of the fabric or the configuration of the fabric? We do that through APIs. The same thing for virtual machines, the same thing for even some application capabilities as well. And those APIs are available for integrations for the other platforms. Exactly. That's what we do. We have a great set of tools in AppCenter, and we think we have a great GUI, but we know that customers may have standardized on other tools to manage their infrastructure. They may have other things in their environment. So we expose all these analytics and automation

and management capabilities up through a robust set of APIs. So you can plug these into a bunch of other tools should you need it. We actually show a demo of leveraging a chat button service now using the APIs to provision and monitor storage with just a few simple clicks. You mentioned Analyzer as another component, I guess, of App Center. What does Analyzer do for you? It's about really providing telemetry analytics across the data path and then allowing you to action on them accordingly. It really falls into two major capabilities. The first one is root cause analysis and resolution. It provides you an end-to-end

view of the infrastructure. It allows you to highlight and monitor capabilities. You can set thresholds for particular resources or workloads, IO response times, cache rates, capacity limits, et cetera. But then once a problem is detected, you can actually go in and use it to quickly diagnose the problem to eliminate potential false positives, get a historical view of what's going on across your entire data path.

So again, it leverages those APIs and pulls in switch information, pulls in some server information. And we have AI built in here as well. And we have an AI heuristic engine that will actually help identify the cause of the performance problem and suggest problem resolution for most common issues.

You mentioned App Center was across the VSP product portfolio as well as other digital infrastructure. You want to talk about what the other digital infrastructure might be in this case? Yeah, sure. I mean, obviously, you know, it's the VSP storage systems, obviously, but it is, you know, fabrics from Cisco and Brocade, right? Multiple types of virtual machines, applications such as Microsoft and Oracle, all of these capabilities, you know, all the things you would normally interact with on a day two or day one operation managing a

storage environment. So we're talking about digital landscapes. I hate to be the jerk that brings up pricing. Packaging, like packaging matters in today's day. Everyone is talking about pay-as-you-go, OPEX, et cetera. How has Hitachi Ventura transformed and allowed me to buy these mid-range systems, especially at the target market. I can see easily this sitting side by side or next in my colo next to a public cloud that's consuming this storage platform. But payment, how I pay for it matters. You guys experimented with

some type of pay-as-you-go model? Totally, yeah. And we are also announcing EverFlex from Hitachi Ventura. EverFlex is a full set of choices for how you can consume infrastructure. EverFlex includes your traditional purchase options, lease options. It includes a utility model, which is that cloud-like model that you're referring to, and a full as-a-service consumption model as well. So how does that measure? what type of tools do you give me to manage and view my capacity usage? Yeah, two things. So when people talk about cloud-like consumption or pay-per-use, they can mean one of a couple of different things because we all know cloud is a little bit nebulous and overly broadly used. But there are two ways in which people can be thinking about it. One is sort of that utility

basis where they pay, they would get the equipment they would normally buy in a CapEx model, except they're paying for it on a usage basis on a month-to-month basis as well, right? So they might want an E990, and instead of paying upfront and amortizing that over five years, they pay for what they use that E990 on a monthly basis. On a capacity level? On a capacity level, exactly, yes, on a capacity level. Then there's the other sort of more advanced cloud-like model, which is a true as-a-service, where you're actually not buying a specific equipment. You're actually buying a service level.

And we provide that as well with EverFlex as a service. And there it's about meeting an SLA, regardless of what the underlying infrastructure is. So we offer you both. And SLA is like IOPS and data availability and response time stuff? Exactly. Yes. Yep. You're kidding me. Okay. I expect you to say something far less capable than that. So I'm actually impressed. No, they're two. And again, you know, customers fall into different buckets.

The reason we have these four, you know, Everflex capabilities is because, you know, there's a continuum of consumption choices and some customers aren't ready to move off CapEx. It just doesn't fit their financial models. Some customers are actually experimenting with all models. And some customers who are sort of born in the cloud and then have moved on-prem are looking for, you know, exactly what they used to buy in cloud, where they can buy a certain amount of IAPS per month. Though obviously on-prem, it ends to be a little more affordable. This is pretty impressive for a mid-range solution,

but it's really across the board, right? I mean, these sorts of functionality, this EverFlex functionality, right? Yeah. And particularly mid-range customers are looking for more of that utility model, pay as you go. They don't either have the capability or they don't have the financial ability to forecast four or five years out anymore. Their businesses are too dynamic. They're too changing. And what we find is that when customers actually move from a five-year CapEx purchase to a utility purchase, they can actually reduce their spend on infrastructure by 20 percent because they're not making, you know, judgment calls about what they're going to need five years down the road and overbuying capacity. Yeah. Bigger, bigger costs and stuff like that. All right. Well, this has been great. Keith, any last questions for Colin? No, I'm just impressed with the pay as you go model. I did not expect it to be that mature. Good work. Well, I was going to say, Keith, there's one secret thing I've been holding back from you just to, you know, always save the best for last.

It's not just about, you know, hardware and software, because like, let's be honest, there are a couple other couple other you know utility models out there in the market right um what but what we can actually do is we can actually bundle not just the hardware and software in a utility model we can bundle a wide variety of the services we mentioned offer before additional solutions and truly allow you to purchase all of that on a monthly utility basis so like installation services or consulting services and that sort of stuff? Installation, consulting services, monitoring services, right? Exactly. Yeah. Yeah. I was a full disclaimer. I kind of expected some high level stuff. I was, again, a Hitachi customer before, and I was used to buying services in this model. I'm just impressed with the granularity in which they've gone with it now. So it's pretty impressive. Right. So Colin, anything impressive. Right, right.

So Colin, anything else you'd like to say to our listening audience before we close? No, I just think we've got great information on HitachiVantaro.com about all of this. You know, we did a great introduction here. Thank you guys. But obviously there's more information. I've got a great technical blog posted, so go check that out. And happy to answer any questions, you know, in the Twittersphere. I'm at World C3, World C, letter C3, number three on Twitter.

Okay. Well, this has been great. Thank you very much, Colin, for being on our show today. And thanks to Hitachi Vintar for sponsoring this podcast. Next time, we will talk with another system storage technology person. Any question you want us to ask, please let us know. And if you enjoy our podcast, tell your friends about it. Please review us on Apple Podcasts, Google Play, and Spotify

as this will help us get the word out. That's it for now. Bye, Keith. Bye, Ray. And bye, Colin. Thanks, Ray. Have a good day. Thanks.

Thanks. Thanks. Thanks.