HP tempts EVA customers with low-end 3PAR array honeypot
EVA replacement gives them a 50% boost
HP is building a low-end 3PAR array with data migration software to convert the EVA customer base to a 3PAR customer base. We hear HP will use SAS disk drives instead of the more expensive 3.5-inch Fibre Channel drives.
This will make it easier to substitute 3PAR's V-, T- and F-Class arrays s for the ageing EVA mid-range, dual-controller arrays in HP's storage product arsenal.
It will be priced at or below current EVA pricing levels and ship in November.
HP added a new high-end, the P10000 or V-Class, to its 3PAR line in August last year. Another year, another array sounds right. We understand that the new array was going to be called the P7000, to complement the EVA or P6000, the P9000/9500 (Hitachi VSP), P10000 and so forth.
However it begins to look as if HP is moving away from its Pxxxx classification naming scheme to product names instead such as StoreOnce, 3PAR, LeftHand, etc. On that basis the new 3PAR array will be called the "something"-Class where "something" is a letter in the alphabet before F.
At the same time as the P10000 launch, HP announced Peer Motion, a way of federating different storage arrays and having them function as a single linked entity with distributed volume management. At launch it included 3PAR and the LeftHand or P4000 iSCSI arrays, with the capability of moving data between arrays.
Our information is that a developed Peer Motion will incorporate EVA arrays and there will be single-click, snap-copy, data transfer from an EVA array to the new 3PAR array. The EVA arrays are long-lived and reliable but limited in functionality and getting outclassed by newer technology arrays such as EMC's VNX and Dell's Compellent line. EVA revenues are declining and HP needs to provide a migration path for its EVA customers or watch them walk away.
Customers get a 50 per cent increase in disk capacity utilisation by moving from their EVA array to a 3PAR one; a pretty compelling increase. For EVA customers, the EVA arrays have been out in the cold and Peer Motion federation brings them back into the HP corral with somewhere to go. It will be a relief. The new Peer Motion will also ship in November.
We also hear rumblings of a 3PAR MiniMe product, but little else than the name. It may be – although this is pure speculation – 3PAR array controller software implemented as a virtual SAN appliance, like a LeftHand VSA, but that means no performance-boosting 3PAR ASIC. MiniMe also sounds small and personal.
When asked about these matters, an HP spokesperson said: "We do not comment product futures, including whether or not details are potentially in our roadmap."
Over in Oracle-land, a new Pillar Axiom array may be coming, together with a ZFS announcement. An Oracle spokesperson couldn't comment about either of these notions. ®
Re: Too late
EVA is done, being replaced with 3PAR at an HP pace. NetApp has had its dedupe day in the sun. The big players have closed the gap and surpassed them in some areas. There are better, less costly mid-range offerings, IMO.
Re: 3PAR VSA unlikely?
I too see the likelihood of a 3PAR VSA unlikely. The F200 is not a demo toy, it is a dual controller system, the only difference between it and the F400 is it is a 2-controller vs 4-controller(customers can of course run only 2 controllers on the F400). Any new low end box would also be 2-controller, since it'd be much cheaper obviously. The main downside to the F-class in my opinion is the number of PCI-X (not even PCIe) slots available for expansion, there isn't much. I'm less interested in driving tons of throughput as much as I am using more direct attach, which applies pretty much just as much to the F400 as it does to the F200. Of course any two-node 3PAR system can't benefit from Persistent Cache which really is a nice feature to have available.
The only systems that have more than one ASIC per controller is the V class. Everything else in their history has one ASIC per controller. I suspect any VSA that may come out would be for non production use, mainly for training/testing purposes, test out things on the CLI or GUI and simulate things. But I think even that is a stretch and may be a while (years) before anything resembling something like that sees public light.
Redundancy is provided at the controller level, to a (much) lesser extent at the ASIC level on the V class. While the failure of one of the ASICs in the V class could result in the controller continuing to operate I can't help but think the number of scenarios where it fails in the right way is small enough to not make much of a difference, and the customer should not put much weight in the redundancy with the two ASICs, they are there for performance rather than availability.
3PAR also has always been anal about high availability, not allowing customers to buy single controller systems, always having to buy disks to provide redundancy across at least two drive cages.
About damn time
We were looking to purchase 3PAR, but they couldn't tell us anything about future products, I.E. what they were going to do about the fact FC disks are stopping production shortly. Even with an NDA! Bought from another vendor in the end :(
It's a shame because I still think the 3PAR premise is awesome, way better than the system we bought in the end.