This is a transcription of a discussion between Wikibon Chief Analyst David Vellante and CTO David Floyer of the August 2011 HP announcements of Peer Motion, a peer-to-peer functionality for both its 3PAR and LeftHand storage systems, and a new high-end V-class 3PAR storage array. They finish the discussion by talking about the pending disruption in the tier 1 storage market from new flash storage technologies and vendors. The discussion was webcast live on Siliconangle.tv.

DV: Good evening everybody. This is Wikibon’s David Vellante live from our studio in Marlborough, Massachusetts, and I’m on with my colleague and analyst extraordinaire David Floyer. Hello David.

DF: Hi there.

DV: We’re going to talk about the 3PAR HP announcement. LeftHand had some announcements. So HP had some big announcements today. The company continues to evolve its line of high-end storage. About a year after its acquisition of 3PAR, HP’s cranking away, new Vseries announcement, federated storage. So David let’s dive right into it. What did HP announce today, and what’s your assessment? What are the highlights?

3PAR V-Class

DF: The highlights of what they announced today were the new V-models of the HP/3PAR storage that’s got more connectivity, a faster processor, and a lot more cache inside the box. That’s the V800 and AV400 at the top end of the HP/3PAR series, and that’s a boost in that area. And they’ve announced federated storage. And then they’ve also announced a couple of small items. They’ve added more thinness to their story. They’re now even thinner. They save more and more storage. But I think really that is more putting right things that they thought they had before and just tweeking things to make them even more efficient..

DV: So basically bigger, faster, better. Does this essentially replace the T-series?

DF: They’re strongly saying it’s a total line. You can take the T-series, you can take the F-series, and you can take the V-series. And they all have different price-points. They haven’t announced it, but I would be very surprised if they didn’t have a significant reduction on the T-series to provide space & room for them to move it as a total product line.

DV: So realistically it replaces the high-end of the T-series.

DF: It will, but for the time being they’ve got all three in their product line.

Peer Motion Federated Storage on 3PAR and LeftHand

DV: Okay. Another big part of the announcement was federated storage.

DF: That’s the highlight to me.

DV: We’ve got a slide (displays slide). This is actually the HP version. They call it Peer Motion and claim it’s the first storage federation in the market for entry, midrange, and high-end storage. So what is federated storage? Why don’t we start there.

DF: Federated storage is about connecting the storage arrays into, if you like, a complex. And they can connect it both within the data center and between data centers within the metro environment. And then with that they can dynamically move workloads between the different systems. That’s pretty awesome, and they have some very nice use cases for that.

DV: Great, let’s talk about the use cases. What are the main use cases? I know you wrote an article today, and you’ve got a lovely graphic.

DF: [Chuckles} I don’t have months to create these graphics beforehand. It’s just a very simple one to show what’s going on here.

DV: Before you get into it: So David wrote a piece today, it’s on Wikibon. if you go to Wikibon.org you can find it. It’s “HP Peer Motion:” You’ve called it a “Must Have Feature for HP 3PAR and HP LeftHand Arrays.” So what are the use cases? Why is it must have?

DF: The first use case is the most cost-effective of all. We did a study at Wikibon a year or two ago, we were looking at the cost of migration for a new array going into a data center. And the cost is somewhere between %50,000 and $100,000. and even at the $50,000 level that works out to be about 43% of the purchase price of a new array is going into migrating that data off the old array and putting it onto the new. So you have to allow time for that. The average time is five months, and it can take up to 12 months. You have to have space in the data center, you have to have extra time on the leaf, you have unutilized resources. It’s really a very expensive proposition just migrating from one to another. On top of that you have lease extensions that are coming up on you that can add up to a lot of stress and a lot of overtime. So the key advantage here is that you can connect this new array to another 3PAR array, for example, if we take the example of two 3PAR boxes, and you can set up the dual pathing between the two arrays, and then the arrays themselves manage the dual address, and they manage the movement of data between the two systems. And it’s [David’s graphic] showing that it’s nearly empty on the one side and being filled up on the other side. And in the process you can choose to just leave it in limbo or you can say, “Now this has been fully moved over” and it moves the address of the volume across to that new [array], and you can disconnect at that point.

DV: So you can migrate data and then tear down the old array without taking an application outage.

DF: No application outage. You can do them[migrations] when you want to. You have to always plan any migration like this; you are not going to do it all in prime time. But it is much, much more flexible. And as you say, you don’t have to take down the system at all; you don’t have to schedule those weekend nightmares when you try to get everything across. These storage arrays have a huge amount of data on them, and the disks aren’t getting much faster. They take a long, long time to migrate that data across, even across a fibre channel link between the two systems.

DV: So it’s a perpetual migration capacity between two systems. Now my understanding is that the HP homogeneous. It’s not a heterogeneous virtualization engine. Can you talk a little about that?

The Competition

DF: Yes, it’s homogeneous. And they’ve also announced it for LeftHand as well, which again is a great edition. So you’ve got one more at the entry level with the iSCSI and one at the more high-end block storage. So nice coverage there. And both of those are homogeneous. There’s chatter about at sometime doing a connection between them, but there’s a lot of complexity in that. But homogeneous in many ways is goodness. If you are doing this and your application’s running you really want to test it 100% between the two. The only other alternative way is – there are two other manufacturers who have stuff out. There’s EMC has federated live migration, which is a similar migration technique that became available in June of this year. And Compellent has something called “Live Volume”. They were actually the first to get it out there and working. So there are a couple of other solutions out there in the marketplace.

Virtualization engines can also do the same job. For example IBM’s SVC is the most common, the Hitachi VSP can do the same job, and EMC has Invista and VPLEX as two ways they can use. The disadvantage of that is that you have to virtualize it and keep it virtualized all the time. And you can’t at the end of it devirtualize it and have it stand alone, working on it’s own.

DV: So that gives us a sense of the approach HP’s taken. They call it peer-to-peer. So there’s not an abstraction layer like you would find in an SVC and your mentioned Invista, which is no longer marketed. Those engines put an abstraction layer in between, and are appliance-based. HP’s is embedded capability, is that correct?

DF: Yes, the best thing to compare it to is actually VMware’s Storage vMotion. And the great advantage is that you now have the ability to do Peer Motion, and I think that’s a great name for it, between the arrays, and move stuff around dynamically, without any virtualization overhead. And that’s very significant under VMware.

DV: Yes, I was going to ask why use an embedded peer-to-peer capability in the array versus just doing it with VMware, and you’re saying the overhead of the hypervisor is onerous.

DF: The overhead of the hypervisor and the I/O through the hypervisor. And as a result the really mission-critical applications haven’t been migrated across to VMware in anyway like the amount of less mission-critical. So this is a way of being able to move mission-critical workloads, really important stuff, large volumes, very efficiently, high speed, both metro distances and local distances with a great deal of confidence that they are going to get there in one piece. And it’s excellent. I think it really extends the capabilities of the storage administration to move stuff around.

3PAR Versus the Tier 1 Vendors

DV: Okay, so we’ve talked about what it is, the use cases, we’ve talked some of the competition – I guess you mentioned EMC, Compellent – or similar products, I don’t know if they are head-to-head competition – EMC, Compellent, and VMware. And the last question I had for you, David, is: We’ve seen 3PAR emerge from startup mode, do an IPO, reach $200 million, and was sold for close to $2.5 billion. Just an incredible run. And 3PAR always was targeting the high end of the marketplace – some of us called it Tier 1.5. is 3PAR now a tier 1 market player? In your mind have the joined the ranks of the IBM DS8000, the Hitachi VSP, and EMC Symmetrix?

DF: Well, they certainly are close in terms of functionality. They’ve got the federated storage. Now, with Peer Motion, that brings them into the fold. One of the great advantages of joining the HP fold is that they could get direct marketing of their boxes right across the world. That’s another important capability that the tier 1 vendors all have. The real differentiation of tier 1 is that when the largest financial institutions, running their absolutely critical workloads, choose 3PAR to do their three data center replication type applications on that. And they’re not there. It takes a long time for these cautious people with these very high value applications to do anything different. 3PAR isn’t actually offering anything extra, other than maybe being slightly cheaper. So the bar to be considered a tier 1 supplier is pretty high. You’ve got to have been there and shown it and been part of that club to really be considered a tier one storage vendor.

DV: So we’re talking about 3PAR’s potential. It looks like they’re knocking at the door. Would you agree with that?

DF: Absolutely. Definitely knocking at the door. My only concern is that by the time these type of arrays get there there’ll be a new type of array, the flash-only array….

DV: Before we talk about that: So 3PAR’s knocking at the door. Is there anybody else that you know of?

DF: Oh, no, there’s nobody else. They are at the top of the 1.5, definitely. They’re at the top, straining to get out of that. But there’s no doubt that their faster performance in the boxes and a wider performance capability than NetApp, for example, or LeftHand, or Compellent, Equalogic, or any of the similar 1.5 type boxes.

Solid-State Flash Storage -- Tier 1 Disruption

DV: Okay, so it’s kind of a four-horse race at this point for the tier 1s. Okay so, you mentioned – we love disruption here at Wikibon/SiliconAngle, and you mentioned these all-flash companies. One that comes to mind is SolidFire. We’ve been briefed by those guys. There are others that just came out of stealth – I think Pure just came out of stealth.

DF: Pure Storage just came out of stealth today, yes.

DV: And of course there are others. We’ve covered Fusion-IO very closely. You just wrote a piece describing the different areas. Fusion-IO memory class storage, the all-flash block-based guys like SolidFire for example, very disruptive to the Tier 1 space. The interesting thing bout SolidFire is they’re targeting cloud service providers explicitly, trying to enable new classes of applications, which I think is unique positioning. Some of the other guys are building up different channels, direct sales force and going after that. But none-the-less they’re disruptive to the block-based storage guys, aren’t they. Can you talk about that a little bit?

DF: Absolutely. They’re now competing on a cost basis. They’re still not competing with the tier 1s because of the functionality within the system.

DV: They’re competing on performance, would you agree?

DF: Oh, definitely they’re competing on performance. And they will get there in the next three years because of the performance and the consistency of the performance that they can offer is just absolutely stunning. And they don’t need any of the complexity of storage tiering or anything like that. They have gotten rid of the slowest piece in the hierarchy, which is the disk itself. And that makes huge improvements in consistency and variability. They have an excellent story. They brought the cost down by great efficiency. Some great innovative designs are happening right across the board. SolidFire has a very strong team, for example, design team. So yes, to me the next two-to-three years is going to be the future of high performance computing in the data center, along with Fusion-IO at the very high end doing the extension of data in memory.

DV: So you just did an analysis of the Flash Memory Summit, and you did one of your “courses for horses”.

DF: This time it as the other way around – it was ‘courses for horses” rather than “horses for courses.”

DV: But in there I think you had some projections on the market size of these various classes of arrays.

DF: Well, yes; it’s very early days. But we put one up just to get the conversation started, because this is really going to separate out three classes of solution. In terms of number of Gbytes, the Fusion-IOs and other people on the motherboard itself, for example, we’re suggesting that that will be 3% of the actual Gbytes but in terms of money that’s likely to be 20% of the actual spend.

DV: So from the data in your chart, what you are saying is that from a storage capacity perspective if I sliced the pie, flash storage on the server like Fusion-IO is 3% of capacity. Flash only arrays will be just over 10% of the capacity. Traditional storage arrays, mostly SATA cheap and deep, are going to be 86% of the capacity. But when you translate that into spend, and you’ve got relative pricing per Tbyte if the traditional SATA-based arrays are the baseline of “1”, flash only will be 6X that baseline – this is by 2015 – and memory-class flash will be 12X. that translates into an enterprise spend of 45% – less than half – for the SATA arrays, and then 35% for the SolidFire-class flash-only storage arrays, and then 20% on the Fusion-IO class. So you’re really projecting by 2015 a major disruption to the spinning disk landscape.

DF: Absolutely. By that time the SAS high-performance disks will … they’re never dead, nothing is ever dead in this industry, but they may have a home in very wide striping, very wide transfers of sequential tape between tiers, which actually would make them look more like a tape deck, wouldn’t it. That would be an interesting final resting home of high-speed disk.

DV: Okay, great. We’re out of time. David, thanks very much for joining us today to review the HP LeftHand/3PAR announcement, new V-class stuff, federated Peer Motion, sounds like great capabilities. Last word from you: What advice would you give users in regards to this announcement and the other things we talked about.

DF: This announcement? Get Peer Motion. It’s fantastic, and it’s a must-have feature. On LeftHand it comes free; on the 3PAR boxes it’s about $6,000 to $25,000. So to me this is a really good deal, and it’s going to save a huge amount of heartache and planning and cost in the data center.

DV: Okay David, thank you. This is David Vellante and David Floyer signing off from the Cube inside Wikibon’s headquarters. Thanks for listening, everybody. Talk to you soon.