Today's leading minds talk Data Storage with Host Enrico Signoretti
Phill Gilbert is the Primary Storage Product Line Manager at Hewlett Packard Enterprise. He has worked with some extremely bright, capable engineers and architects to design and build products and solutions. As a product manager, he has taken that experience and technical capability to truly understand customer problems, communicate and brainstorm solutions and work to deliver products that exceed his customer's expectations. As a manager, he uses his experience and compassion to drive the team to success.
Enrico Signoretti: Welcome, everybody to a new episode of Gigaom’s Voices in Data Storage. I'm Enrico Signoretti, your host, and today with me I have Phill Gilbert, product management lead for HPE Primera.
Phill Gilbert: That's right.
Okay, Phill, we are here at the HPE store and this is the last day of the event and we went through everything about HPE Primera. Also we had the chance to see all the announcements we added this week. Before going into the HPE Primera, it's quite interesting for me looking at HPE and the evolution it is having, especially on the storage side. It's been a very exciting–I mean, almost all the product line-ups were great. We see a lot of news coming out. The company's pretty active on this side, right?
Yeah, there's a big focus on innovation at the moment and actually building not just new things but completely different new things, so a big focus on–the phrase that's often used is ‘think unconstrained.’ If you have that blank sheet, what could you do, and then work out how to get there. I think this week is demonstrating a lot of that.
Yes, also what I notice, especially here on the show floor–excuse us for the noise. I didn't tell it at the beginning, but we are actually on the show floor, so this is why you probably are hearing those little bits of noise. What I saw here on the show floor is a lot of integration between on-premises storage and the cloud, so this hybrid thing that is going on with a lot of products like HPE, e-cloud volumes, for example. It's pretty interesting. Also Anthos integration, so you're doing a lot of things to make your customers more hybrid.
Absolutely, I mean, in reality, our customers reach out to us and tell us these are the things that you need to be able to achieve. One of the big things we have to do is make sure we listen to our customers because they're the ones who know what it is that they need. If they reach out and say “We need to be able to do cloud volumes,” we need to be able to integrate those kinds of things, we can actually build solutions that really meet what it is that they need to do on a day-to-day basis.
Let's start talking about HPE Primera, your child somehow now.
Sure is, yeah. (laughter)
That was the most important announcement of the week. It's a new storage system positioned very high for primary workloads, and so tell us a little bit about it.
Sure, so it is–so essentially what we've done is we–as I said, we're like this blank sheet of paper. What we did was say if you could build anything, where would you start? We looked across the platforms that we already have and see how can we take the best of each of those and put that into a single platform. If you want to build a platform for Tier 0, for mission-critical applications, that market already exists, so you've got to offer something different. You got to have something unique in some way.
For us, we can take everything that we know about, for example, the monitoring experience, the simplicity of the Nimble platform. I mean the experience when you interact with that. You take the intelligence that we've built from InfoSight, that modeling, the ability to predict, the ability to adapt in a real-time and take the underlying architecture from one of our most successful mission-critical platforms that we already have like 3PAR, which delivers that availability.
Now, if you can take the concepts of all three of those and wrap them into a single platform, you can create something that sits in the Tier 0 space and offers the availability and performances expected there. If you can take the simplicity of Nimble and bring that into that space and then take the intelligence that runs in InfoSight and say you're going to deploy that on the array itself and make the array more intelligent, more capable, then you're building something truly unique.
I want to start from something that I think particularly–so in the last few years, HPE acquired Nimble and 3PAR and other companies. Each time, you decided to design everything in-house. It's an evolution somehow but also a revolution for the platform. Instead of buying something on the market, you decided to go to the design table.
To build something instead.
Yeah, and so it's also a warning for your engineering teams.
Absolutely, and I think acquisitions are something that quite often happen when you have a hole in your portfolio that you need to fill. That's not quite accurate. I'm going to start that one over again. When we acquire products, that's usually because we either–that's not even the phrase I want to use.
It's quickly–to close a gap.
Yeah, exactly. Oftentimes with an acquisition, it's all about speed, and it's about plugging the gap quickly . In our portfolio, that's not a big yawning gap that we have at the moment, and we wanted to do something genuinely different. Also, the acquisitions we've made–if we go back over the last nine years and we have three products, Nimble brings InfoSight, and then we have other acquisitions we've made along the way in terms of storage and if you can learn everything that you've got from those acquisitions, then you're making the most of what you've got. We already have happy customers in every single one of those, and if we can make them even happier with a brand new platform, that's the goal.
Right. The other thing is Tier 0. What do you mean for Tier 0? Sometimes, Tier 0 is all about performance but in this case, there is much more on HPE Primera than just performance.
Yeah, and that's why we've married with–Tier 0 is mission-critical. They're those two things together. It's difficult in a world of defining storage because it means so many different things to so many different people. What we're getting at with the Tier 0 is we're talking about your most business-critical applications, those things that cannot tolerate downtime, that need not just high performance, but they need predictable high performance You need to be able to guarantee that the performance–and I don't just mean IOPs and bandwidth and latency. I mean the whole package. It has to be predictable. If you call for the availability, the predictability, and then provide the intelligence that just makes everything a bit better, then that's what it's about.
In fact, you provide also a 100% reliability guarantee, which sometimes you think about as a marketing thing but actually there is something beyond it, right?
Yeah, quite often these guarantees, when you read the terms and conditions, you find out that what you're getting–if for whatever reason you're claiming on the guarantee, then what you get is a month or three months or something of elevated support, for example. That's quite unsatisfying if you, for example, bought the best level support because there's nowhere to go from that. The other thing to mention is I often have quite complicated terms and conditions.
What we wanted to do was give something that came to everybody. If you buy this, you will get this. We wanted to show how confident we are on delivering this by saying not–we're not going to give you support credits or something like this. We're going to give you essentially cash back. We're genuinely putting our money where our mouth is and saying if, for whatever reason, you do experience an outage, we'll give you something meaningful back in response.
We're talking about a new platform that borrows a lot from 3PAR. If I am a 3PAR customer, what is the best way to approach this new platform and with this, I mean, what is the best way to integrate primary with my existing 3PAR and my...
That's interesting because as you can imagine, we spend a lot of time talking to our customers about what it is they want. Initially, we were talking about having a management platform that was unique to Primera, but a lot of our customers said, “Well, we spend a lot of time and money investing in SSMC, for example, from a management perspective. We don't really need a product with this deep learning curve.” Then we went back and said well, actually if we can make it work inside of SSMC, so you can have your existing platform and then buy something new, but manage it in the same experience, that really breaks down the barriers for adoption and makes it much easier for those customers to integrate into their environments. We did a lot of work to make sure that was going to be possible.
Okay, so is there a migration path from 3PAR to HPE Primera? Do you expect customer moving from this platform to the other?
Yeah, absolutely. I mean, the idea with Primera is to target obviously new customers. We want new people to come out and say “This is wonderful,” but obviously we have an existing set of customers who are going to look at this and say “That's the thing that I needed.” We needed to make sure there was an easy and reliable way for them to migrate to this platform. Initially we have a set of tools that you can run and they will look at your existing 3PAR system. They'll look at your Primera system, and they will orchestrate the migration of your data into Primera.
Another thing is about this Tier 0 is performance. How do you compare the existing 3PAR with the new Primera, for example? What is Primera bringing to the table?
Yeah, so that's an interesting point. When you look at the Primera system, it's 4 unibyte. It's tiny, really. If you were trying to compare it with one of the 3PAR platforms from an all-out performance perspective in the thing and nothing else, your closest approximation, assuming you want to do maximum performance and that kind of stuff, would be somewhere around a 2850, the biggest platform we have at the moment. In terms of footprint, that is probably around–to give that thing the maximum performance, you're probably looking at the best part of something between 20 and 40 euro, I would imagine, under 20K, and we're now doing that in 4U in Primera. That gives you an idea of the difference.
Okay, but you didn't mention performance. Oh, no, sorry, I'm saying the same performance as a 2850 and the same footprint. I'm giving you a comparison between the two.
What about scalability? Because actually looking at the slides, you have plenty of options for this system.
Oh, yeah, so in terms of connectivity, each controller can support up to three, four-port adaptors, so that's 48 ports of host connectivity in 4U, so that's pretty dense. In terms of scale, in terms of a capacity, it supports up to four petabytes of effective capacity or Flash, and obviously from an enterprise scale perspective in terms of volumes, it's everything you would expect. It's, depending on the model, tens of hundreds of thousands of objects.
Can we expect in the future that there would be tools to make Primera compatible with Nimble and Nimble maybe in remote offices and remote installation that can replicate data through the Hyper-V. Does that make sense?
Absolutely, we've got–we have a technology called Peer Copy. It's enabled by RMC, and that allows you to replicate between a Nimble system and a Primera system. You can do snapshot-based application-consistent, all-crash consistent, if you want, replication between those two different platforms. That absolutely is a big goal. As you said, a big use case for that is remote offices with Nimble platforms in them and a big central database with your Primera.
This is pretty cool, actually. And when will this system will be available?
Yeah, of course. It was available for order in the beginning of July and it will start shipping in August.
Okay, so very soon.
What can we expect then for the future of the product line? I mean, so Primera is very high-end but starting is also very small for units, small controllers also, looking at the various models that you will propose. What can we expect from now on on the product line-up? I mean, I'm sure that Nimble that is also very successful will stay forever. What about 3PAR?
Yeah, so the 3PAR platform is an interesting one because it's been very successful for us. We have a lot of systems out there and a lot of extremely happy customers. There are no intentions at the moment to end of like that platform; it's still got a lot of life left in it. We still have a healthy road map for new software features for that platform, and you're going to see those coming out over the next six to nine months.
Ah, very good. What about the platforms that you will support on top of Primera? I mean, VMware for sure with all the companies like eVault and stuff, but what else for the other environments?
Yeah, so I mean, in terms of who we support, we support all the big operating systems that you would imagine that are there, including things we have support for Oracle Linux and [IBM] AIX. For mission-critical platforms, this is the kind of hardware that they've got out there. In terms of applications, we support everything you would imagine and also of course we have support with RMC for data protection to provide application-based support protection for things like SAP Hanna and Oracle and those kind of things, so we've got integration into all those platforms as well.
Okay, back a little bit to the hardware because performance today means NVMe in most of the cases. I suppose that internally, we have a lot of NVMe in the system.
Are you going to expose also NVMe on the front-end?
Yeah, absolutely. We know that–it's an interesting one because we know NVMe over fabric is coming. It's unstoppable. It's coming, but our customers have very much been on the side that what they really want right now is fiber channel connectivity as their main option. Some of our customers have said they need iSCSI connectivity as well. The area of focus needs to be on those two first.
Most of our customers have said they are looking at NVMe over fabric, but they have no immediate intention to deploy it. What they want to make sure is we do have the ability to put it on the platform. One of the tricks that we actually have up our sleeve with this platform is that while we're not exposing it, the hardware and software are actually built for and can support NVMe over fabric. All the components are in the operating system already and the hardware supports it.
The system is ready. It's just a matter of time to get it in the field.
3PAR has also now query, is a hybrid system. You can support hard drives as well as Flash. What will happen with Primera?
Yeah, so Primera is designed differently. It has a different architecture, and it has a different design goal. It's not a hybrid platform. What it is is an all-Flash platform. When we work–I was thinking about this blank sheet of paper, and initially we said let's just build an all-Flash platform because that's what we want. We went around and spoke to customers and they said “yeah, we need an all-Flash platform, and that's where all of our most important data is going to go, but we do also have some other data that doesn't have high performance requirements but does have high availability requirements, and we need somewhere for that data to live.” It's somewhere between production data and archive data, so it's not super cold but it lives in the middle.
Availability but not performance.
Exactly. Nobody wants to be back in the world of a supple interior. All of our customers have said to us very clearly, we don't want that back again. We built a platform that is an all-Flash platform that offers a tier of spinning media for largely archive storage, but it offers a consolidation to a single physical platform.
I see. Is the platform ready to support new technologies like Optane, for example?
Absolutely, yeah, so not only is it built to be able to support those technologies, they are almost an integral part. You mentioned the architecture and having NVMe inside. The architecture itself is being built with–not just with NVMe in mind. It's being built for NVMe, so a whole internal architecture from a hardware and software perspective is built for massive parallelization. That means that it is internally a completely non-blocking architecture. Any I/O can go anywhere at any point and it never has contention. That's really important when you're talking about technologies like storage class memory because they offer huge parallelization. If you want to leverage that, you've got to have the right architecture.
That leads me to another question that is also related to performance . One of the major features of Creeper was the ASIC. The ASIC was loading a lot of storage abrasion. What is going to happen with Primera?
Primera has inherited the fundamental ideas and concepts from 3PAR and along with that comes the ASIC. The ASIC is a totally different architecture. It's nothing like 3PAR architecture. However, the ASICs do still have offload engines. Now the main use for the ASICs is actually to form a cluster, so that's how we're able to do this massive parallelization. They do have offload engines for things like grade and for de-duplication. Now, in addition to the ASICs, the Primera platform also has Intel's QuickAssist Technology, QAT, and there is a QAT kind of like an offload engine or a co-processor, if you like, and it sits in every node, as well. That can do additional tasks, so that can handle things like compression, encryption, those kind of things, and it can do those at a very, very high speed. Not only do we have the ASICs, that handle all the really low-level functions we've had, but now we've got the ability to offload and accelerate additional technologies as well.
Very good. We covered almost everything, I think; the new product, what it gets from 3PAR, but there is a very, very important thing that it gets from Nimble, which is InfoSight. The product is new, but you are the leader in the storage wars with InfoSight. It's a very successful analytics platform. How does InfoSight change to reflect what HP Primera brings to the customers?
Yeah, so that's a really great question because one of the things we're aware of is we've done this–so in InfoSight, we have an amazing analytics engine. We have a lot of information about performance and how performance works, how it's handled, and being able to predict things. That is one of the main reasons we're able to offer the 100% availability guarantee, but the performance aspects are fascinating. We took a similar concept and put that for the existing 3PAR platforms. We've put that inside SSMC. That's what we call performance analytics and workload analytics. Those two platforms allow us to report to the storage administrator what the saturation level is the array. You could now see how busy your array is. It gives you actionable insights in terms of if you change this, you'll get faster, or if there's a bottleneck, this is what it is. Those are absolutely critical things to troubleshoot performance.
What we looked to and we said, “Wouldn't it be great if we could actually use that analytic data? Wouldn't it be awesome if the array could actually understand how busy it is? because that opens some new doors that you can't do any other way.” We've taken the engine that runs behind that, the model that we've learned in the cloud, and we're deploying that on the array itself. That means the array is now aware of itself. It now knows how busy it is. It can understand the workloads that are running on the array. We're just beginning to open the door into a whole new world of possibilities and things that the array can now do for our customers that at the scale they now live at, it's unrealistic for anybody to be able to do.
Okay, so one of the things that we have to check out when we look at the HP Primera is the InfoSight then.
Absolutely, the intelligence aspect of this cannot be underrated. It is quite staggering what we have initially and what we're going to have later. It's going to be phenomenal.
What happens from the end user point of view? Are they two different InfoSights or they are integrated in a single point? If my infrastructure has 3PAR as well as Nimble and then I add HPE Primera, what happens?
Essentially from a InfoSight perspective, InfoSight is our software-as-a-service portal, so you can log onto InfoSight. In there, you can see your entire portfolio. You can see your Nimble platforms, your 3PAR platforms, and your Primera platforms in a single place. That's very powerful because that gives you a global view of your entire storage structure. Of course, we've added Align into the mix. We have other platforms coming in there as well.
We have with this running on the node itself, this means that we can start offering that level of detail in real time on the actual array itself. The array–the Primera platform now has its own management console, and we can expose that data inside this native management console, inside SSMC, inside InfoSight, so we now have that across the board. It doesn't matter which management point or analytical point you use, you've got access to a huge new wealth of data and we're going to use that to make your life better.
That's great. I think we can wrap up now because I think we got a lot of information about HPE Primera. I'm very curious to know more; maybe also our listeners, so where can we find more information on the web about this new product?
Yeah, absolutely. The easiest way to do it is just go to your favorite search engine and just look for HPE Primera. There's already quite a lot of information out there and over the following weeks, we're going to be uploading more information, things like text batch, those kind of things, so you're going to get access to a lot more information. Keep checking; there'll be more stuff coming soon.
Do you have a Twitter handle in case somebody wants to know more?
Absolutely. Mine is a bit weird. It's my name backward, so it's @GilPhilbert, G-I-L P-H-I-L-B-E-R-T.
Okay, Phill, thank you again for taking the time to talk about HPE Primera with me.