Interviews

posted on 01 April 2008 07:40

We had the idea that a fourth wave of storage offerings is coming along, from suppliers such as Panasas, Atrato and Pivot3. As Atrato has just launched and is looking at the video survelliance market we thought it would be neat to ask Pivot3's founder and chief marketing officer, Lee Caswell, about Atrato, video surveilance and this fourth wave of storage idea. (Pivot3 has recently announced an 'ecosystem' of some thirty or so video surveillance suppliers.) Here's the result:-

B&F: Has Pivot3 anything to say about Atrato and its Velocity1000 product? It is also aimed at the surveillance market, amongst others.

Lee Caswell: Many people think of the video surveillance storage market as one market, but it’s useful to break it into two broad market segments which I’ll call centralized and distributed, to better understand the user requirements and where products will be successful.

In the centralized model, from one hundred to several thousands of cameras are connected by a LAN to large-scale central storage. The LAN, typically Gigabit Etherent, provides high local bandwidth and the storage must provide both massive bandwidth and capacity, scaling from hundreds of terabytes to tens of Petabytes. Typical installations include gaming environments, airports, railways, seaports, border initiatives, and correctional facilities. Because cost is a key buying criteria, these customers largely rely on external DAS today although they could clearly benefit from the use of block-based SANs for storage consolidation, virtualized management, and better availability.

For the distributed market, from one to tens of cameras are located in distributed environments that are connected by a WAN. Video is generally stored locally and occasionally sent over the WAN. The local storage has both constrained capacity and bandwidth requirements and is largely driven by cost. Typical installations include retail outlets, financial institutions, educational environments and healthcare. DAS is used extensively in this market and there’s little benefit to networked storage or high-bandwidth storage.

Each of these segments accounts for roughly 50% of the market and Pivot3 is targeting the centralized market where the values of capacity scaling, massive bandwidth using cheap aggregated GE links, virtualized management, parallel rebuilds and low-cost from standard hardware makes us an obvious play.

It’s hard to see how Atrata, from my knowledge of the solution would a good fit for either market. While the simultaneous stream performance is attractive, I certainly wouldn’t see their solution as competitive in the centralized market because of the use of low-capacity notebook drives and the sealed box architecture, design choices which prohibit the solution from achieving video surveillance cost targets. Atrata list pricing of $7/TB, or 5x the going rate for video surveillance solutions, would seem to confirm they would be better suited for high-bandwidth markets with lower capacity requirements and less cost-sensitivity.

B&F: Does Pivot3 agree that there is a new and fourth wave of disk array storage companies emerging?

Lee Caswell:

1. Traditional SAN/NAS array vendors - EMC, NetApp, HDS.

OK so far.

2. Virtualised and less expensive more intelligent arrays - 3PAR, Xiotech Magnitude, Pillar Axiom 3.

Lower-priced maybe, I wouldn’t jump on lower cost. It’s still all custom hardware.

3. Virtualising and clustering and higher performance NAS - Isilon, BlueArc, ExaGrid

Yes. I’d put OnStor, PolyServe, IBrix, Spinnaker in this category too. From our block perspective, file virtualization was a great way to start virtualizing at the file system without having to re-architect the underlying RAID algorithms. That’s the challenge we took on.

4. Clustered, high-I/O newcomers - Panasas, Pivot3, Atrato and Xiotech with its coming ISE product. Would Pivot 3 agree with this classification? How would Pivot3 describe the attributes of this fourth group?

Sure – it’s an interesting way to put the group together. Common elements of the fourth group would be pay-as-you-grow, scalable bandwidth and capacity, virtualized management, high availability by sustaining node failures. Important differences to consider will be true peered RAID architectures versus clustered RAID pairs. The elegance of the solution will directly affect the solution’s ability to scale performance (without requiring extensive intranode communication), sustain node failure (without requiring mirroring), and supporting the use of standard server-based hardware (without requiring dual controllers and their cost) and standard Ethernet networks based on iSCSI.

By the way, I specifically mention server hardware as an important distinction for Pivot3 because server hardware has nearly 2 orders of magnitude of scale economies over storage hardware, even “standard hardware”.

B&F: Will IBM/s XIV technology and EMC's Hulk/Maui technology fit in this 4th wave category? If not, where?

Lee Caswell: Yes on XIV although it’s not very capacity-efficient. Couldn’t say on Hulk/Maui.

B&F: What application areas are better suited to 4th wave technology than existing storage suppliers: video surveillance, video delivery, HPC, web 2.0, 'cloud' storage services?

Lee Caswell: I see all of these as being great target markets where conventional RAID solutions with expensive dual controllers with fixed head-end pipes attached to many drives won’t be a good fit. As important, extreme availability is not the key customer requirement, so start-up solutions will be put to use. Again, there’s probably a great chart here showing the following attributes:

o Video surveillance (large capacity, high bandwidth, low cost)

o video delivery (large capacity, high bandwidth, low cost)

o HPC (small capacity, extreme bandwidth, low cost)

o web 2.0 (large capacity, moderate bandwidth, moderate cost)

o 'cloud' storage services (large capacity, moderate bandwidth, low cost)

B&F: Will the fourth wave's offerings evolve into general purpose ones?

Lee Caswell: I don’t see these fourth wave offering’s displacing general business applications anytime soon. The sizzle of fourth wave attributes isn’t going to overcome the weight of features and buttons available for conventional arrays. But all of the interesting growth over the next 10 years is going to come from these new markets.

B&F: Will the existing storage majors lacking 4th wave technology invent their own, OEM it or buy it in?

Lee Caswell: If history is any judge, they’ll claim it’s unimportant until they lose some sizable deals and then they’ll open their wallets.

^{[Chris Mellor.]}