Will Hitachi's high-end refresh be more of the same?
Two years and counting
Comment Substantial refreshes of high-end storage boxes by the big storage beasts are rare, but they do happen. With EMC having revamped Symmetrix, attention is turning to the others. What is Hitachi Data Systems going to do?
The five big storage hardware players are EMC, HDS, HP, IBM and NetApp. HP resells HDS' USP-V as its XP, and until that OEM arrangement stops we can cut our list to four. The pressure to refresh comes from developing user needs on the one hand, and meeting competitive produces on the other. The users want to store more data and store it efficiently, to grow their capacity cost-effectively, manage their storage more easily, with a background of data centres getting virtual servers, and getting virtualised themselves, and they want to respond to consolidated networking, aka Ethernet.
The competitive pressures come from the other storage array-supplying giants, and from next-generation SAN suppliers 3PAR, Compellent, and Pillar, characterised by thin provisioning, settable quality of service (QOS), and automated block-level data movement across tiers respectively. Of course they do much, much more, but these three attributes stand out.
IBM has its DS8000 which some say is long in the tooth. It is a monolithic array like HDS' USP-V and EMC's Symmetrix, and IBM has been steadily issuing minor upgrades to it, such as flash solid state drive (SSD) support and thin provisioning, but hasn't signalled that a major refresh is coming. Symmetrix now has tiered storage, thin provisioning, much more scale, with federated controllers for example, and is getting FAST automated data movement across its tiers at the LUN level soon and the sub-LUN level next year. It has responded to user needs and to 3PAR, Compellent and Pillar Data.
NetApp has just released its converged clustering and non-clustering array operating system, ONTAP 8, and says a high-end FAS 6000 refresh is on its roadmap but has no details available. NetApp EMEA marketeer John Rollason says it's coming in a year or two but he can't talk about it.
ONTAP 8 enables multiple heads on a FAS array, with an Ethernet interconnect between them, but there is no switch-like fabric between them and the storage shelves, as there is with competitors' top-of-the-range enterprise storage arrays - not that NetApp sees this as a missing item.
HDS' USP-V is an array with a virtualising and scalable controller, which can virtualise third-party arrays and include them in its single logical pool. This extended controller functionality is something that IBM provides by combining its SAN Volume Controller (SVC) with a DS8000 or other arrays in its range such as the XIV. EMC could do it with an Invista/Symmetrix combination but Invista appears to have been de-focussed. NetApp can do to with a separate V Series controller linked to a FAS array. Only HDS has multi-array virtualistation and array controller combined in one product.
Hu Yoshida, HDS' chief technical officer, said: "Within the next two years we'll see a refresh of the USP-V. The basic elements are sound - separate control and data paths, the global cache, the switch system. This basic core architecture will stay but the manifestation of it will differ."
"The competition put a modular array on a switch and scale out," alluding to IBM and XIV. "A scale-out means each modular unit can't do anymore work itself. We're scaling up and out."
"Modular arrays are not really high-availability, with (just) two controllers. With (our) monolithic USP-V array you have 128 processors. You may see more."
Is it going to adopt Intel X86 processors? "Today, the USP's processors are not Intel." That seems like a careful answer.
USP-V 2 could see the adoption of Intel's Xeon 5500 processing engine with, guesstimating, up to 256 controllers or controller engines, interconnected with a higher-speed switched fabric, and supporting multiple tiers of storage ranging from SSD through 15K rpm drives to 7,200rpm SATA bulk storage disks. Volumes will be thinly provisioned and there will probably be some form of automated movement of data blocks between tiers. We might also expect multi-tenancy features, the ability to group users into an entity and give them defined qualities of storage service concerning performance, data protection and security.
HDS is now doing this with its HCP archive platform and it would seem reasonable to extend it to raw block access and the USP-V. USP-V 2 could or should also receive Fibre Channel over Ethernet (FCoE) support.
HDS has good architectural differentiation from other enterprise array supplies and this has served it well, well enough for the refresh principle to be "more of the same". We know a NetApp high-end refresh is coming. HDS has told us a little of its intentions.
What about IBM? The XIV array looks to not to be a DS8000 replacement, and the previously rumoured DS8700 or whatever a revamped DS8000 will be called may be closer than we think. ®