HP converges on itself with bulked up storage offering
EVA Clusters, mid-range dedupe and iSCSI box
HP has upgraded its storage offerings as part of a Converged Infrastructure (CI) announcement, making EVAs clusterable, enhancing mid-range dedupe and introducing a new LeftHand iSCSI box.
It's announced a raft of new servers as well as storage.
CI is HP's integrated IT stack, its response to Cisco entering the server market and building integrated server stacks with EMC and VMware. Since David Donatelli joined HP from EMC to run its ESS servers, storage and networking business, the company has moved to making big quarterly announcements like this rather than streaming them out continuously.
First up is the ability to have pre-built EVA mid-range storage array clusters.
HP's new EVA Cluster is a pre-configured and tested factory system using EVA 6400/8400 arrays, with two Fibre Channel (FC) switches and an Ethernet switch included, as well as a Data Path Module (DPM) pair. HP says the DPMs "are intelligent switch-like devices that perform the virtualisation of data at switch-like speeds".
They are sourced from LSI (pdf), which acquired them when it bought StoreAge a few years ago, and contain a Xeon processor and 2GB of RAM. They interconnect host servers with the EVA cluster and virtualise its storage.
A basic 2-node cluster can scale to 624 disk drives and 16 solid state drives (SSD). The 6-node maximum EVA Cluster Domain can have up to 1.9PB of data stored on 1,872 1TB FATA (Fibre Channel Attached, Technology Adapted) drives and 48 SSDs. All the storage is configured into a single virtual pool with a single point of management (Command View SVSP).
Thin provisioning is supported as is LUN migration between the cluster's arrays. The cluster software provides a volume manager, optional SVSP (SAN Virtualisation Services Platform) Business Copy (local replication using snapshots), and SVSP Continuous Access for remote replication. You can expand beyond six nodes by adding another DPM pair and increasing the FC switch configuration with more ports. It's possible to connect selected third-party arrays from IBM, EMC, Sun and SGI as well; not NetApp though, nor HDS.
You can get quick cluster specs here (PDF).
The HP StorageWorks EVA Cluster FIO Starter Kit pricing starts from €59,000 with two EVA nodes.
P4800 BladeSystem SAN
HP has a new P4800 BladeSystem SAN which holds up to 63TB of data across four storage blades with 40 disk drives each. It is intended for virtual desktop applications and works with both VMware View and Hyper-V with Citrix XenDesktop.
The P4800 BladeSystem SAN is a component in HP's CI client virtualisation reference architecture, the first such pre-sized and pre-tested component or template, which is intended to make purchase and deployment fast and simple. HP says it supports up to 1,600 desktop users and claims it does so "at 50 per cent less cost and requires 60 per cent less space than traditional client virtualisation implementations".
The P4800 BladeSystem SAN starts from €190,000.
Next page: Deduplication
Sure, the EVA Cluster is about re-using their existing hardware and software - but what's wrong with that if it introduces new configs? In this case it seems - based on its spec page - they are extending SAN/iQ portfolio, used on Lefthand aka P4xxx-series devices, to the EVA line with storage clustering added.
I don't really see it as a problem as long as they get priced accordingly (Lefthand is waaay overpriced, it's rather a joke) and finally they are willing to come out with 2TB or at least 1.5TB SATA errr, I mean FATA drives for these storage systems...
Sorry, maybe it's just a hangover from yet another contentless 100-slide vendor session, but this isn't really much new news! The "new" EVA Cluster is just the existing SVSP hardware, software, and a pair of EVAs, all bundled into one order code. Whilst the SVSP stuff is good, the new "Cluster" is hardly groundbreaking. TBH, I was hoping the EVA Cluster was going to be a bit more inventive.
De-dupe? Everyone's got de-dupe now. It does look mildly interesting, but I have yet to have any vendor actually come onsite and meet their brochure figures for either compression or savings in backup time. I hope hp have been conservative in their performance claims - they usually are - as sometimes it's embarassing when you have to explain to a boardmember that the brochure figures and vendor benchmarks he swallowed were "best case", i.e., "never going to happen in the real World"!
The new VC card is interesting, but mainly as it should allow us to use just one mezz slot for IP and/or SAN, which means we have more mezz slots free for other cards (important for clustering or webserving, where I want as many IP ports as I can get onto the card but still have redundant links for SAN). But the FCoE upgrade possibilities? FCoE is currently hamstrung by the fact that FCoE is a one-hop solution until they sort out the inter-switch protocol standards. So that means you can upgrade and go FCoE from the VC card/module to the top-of-the-rack switch (well, in this case to the blade chassis switch module), but then you have to split it out into seperate SAN and IP cabling again. So still virtually zero reason to actually have an FCoE capability in the first place, other than so hp can say "yeah, we can do what CISCO can, if customers ever want to".
/Off to the pub!