3.014: so much 'ado about . . .
Wow – what a bustling couple of weeks!
So many competitive storage announcements, you'd practically think they were all scheduled to maximize their disruptive impact on Q4 storage spend.
When you're the market leader, as EMC has been for the past 2 decades-plus, you learn to expect this almost annual frenzy. It comes with the target that leaders have tattooed on their backs.
This year the wanna-bee followers seem particularly agitated, though. Hitachi invested heavily in marketing sizzle for the first time since Mr. T was their chosen spokesperson – and with good reason, I'll admit: by my observations of IDC Storage Tracker data, Hitachi's delay in refreshing the aged USP-V (coupled with the loss of Sun as a reseller) has driven 5 straight quarters of USP-V market share declines vs. VMAX and the newly retired IBM DS8700.
Hitachi obviously had to try something different, even if it meant moving to yet-another new processor base. But unable to change their architecture to fully leverage industry-standard open components, their "rush" to market was slowed by the need to create FOUR proprietary ASICs. And those ASICs further handcuffed the move to the Intel platform. With the unavoidably long lead-times of ASIC development, Hitachi was locked into implementing with the PCIe Gen1-based infrastructure and processors, even as Intel is delivering the second-generation of PCIe Gen2 CPUs and interconnect. The net result? Using the same Intel processor as the 19-month old VMAX, the new VSP can't even double the performance of the USP-V that it doesn't quite replace.
That leaves VMAX at the top of the performance heap, having more than doubled the performance of the DMX4 when it was introduced in April 2009.
As a further testament to the insignificance of the VSP, I'll also note that HP has chosen to use a totally different name for the product in their lineup. Not only has my old boss shunned the brand, he and his new head of storage outright told the world that Hitachi remained in the product lineup only to support HP's mainframe customers and to fill the void above 3PAR until such time as it grows up. That must have thrilled HP EVA and XP customers alike, both groups who now find themselves sitting on dead-end kit with no defined escape path.
But that's a story for another day…
Yep, beneath all the new-found bravado and marketing spin, the simple reality is that Hitachi & HDS are still attempting to follow EMC's lead, and they're falling further and further behind.
Over at IBM, someone over there is apparently tired of getting left behind in the storage space. That someone must have an MBA as well, since they've apparently figured out that bundling storage with servers at no extra charge is not a long-term profitable strategy – especially when so much of the server base is shifting to Intel-based compute clusters rather than proprietary hardware. Hard times lead to big gambles, and the industry analysts apparently think that IBM has gone all-in on this one.
Like Hitachi, IBM has sort of replaced its high-end DS8700 with the new DS8800. I say "sort of" because the DS8800 doesn't yet support all the same features that the DS8700 does, so IBM is keeping the old array on life support for those that need esoteric stuff like:
The following functions are currently not available on DS8800:
- Quick initialization and thin provisioning support
- Remote Pair FlashCopy support
- Easy Tier support
- Multiple Global Mirror session support
- z/HPF extended distance capability support
- z/OS distributed data backup support
- IBM Disk full page protection support
- 16 TB LUN size is not available
(from the IBM DS8800 announcement materials)
IBM made lots of noise about how the DS8800 reused 95% of the DS8700 code, but I'm skeptical of that claim – I find it hard to believe that thin provisioning, Easy Tiering, Global Mirror and the rest represent 5% of unique code. TonyP asserted that the above features were "forked out" of the main development tree to be reunited sometime in 1H'2011 – I read that as admitting the DS8800 was rushed to hit a date for marketing value rather than customer value. With all their hype around "green,"
I find it particularly interesting that thin provisioning and Easy Tier were deferred by the powers at IBM, since the very purpose of both is to improve utilization (reduce wasted/unused capacity), reduce footprint and lower power and cooling requirements (fewer drives), and to improve overall performance and efficiency (thanks to more IOPS/drive with SSDs and more GB/drive with SATA). Seems like the IBM sales teams are going to have to speak out of both sides of their mouths until the rest of the DS8800 software is designed, developed and debugged.
Finally, I haven't forgotten the newly-minted V7000 array that pretty much my took pal @BarryW off-line since the beginning of 2009. First – hats off to you sir…you've been lobbying to turn that software into a storage array for far too long. And thankfully, you appear to have overcome the insistence of the publicly-anointed father of storage, relegating his XIV abomination to merely a footnote in the announcement (in fact, the only reference to XIV was to call out that the new SVC GUI was modeled after the pooled simplicity of XIV's UI).
But like the DS8800, the new V7000 clearly isn't quite ready for the market either…sure, the code base is stable – it's been out there quite some time now. In fact, you might say it's beginning to show its age, being as it still lacks the true scale-out clustering and distributed active/active cache coherency that fuels EMC's VPLEX. I suspect the newly formed V-team found it a bit more challenging to handle all the enclosure and spindle management that was originally envisioned for the product – the result being a dual-controller mid-tier (barely) array that can support only a tiny fraction of the capacity supported by EMC's Unified platform. It's a start, I guess, but it clearly isn't going to be the CLARiiON-killer IBM was boasting about yesterday – at least not until it scales up to a real processor and something better than 240 drives (the max drives the V7000 supports out the gate is only 120 according to the announcement materials)
I gather from the blatant exclusion of XIV from IBM's obviously coordinated and planned Major Storage Announcement Event that someone has finally figured out that the XIV architecture's inherent fallibility was too much of a liability. Given the rate that VMAX and CLARiiON are replacing XIV arrays that haven't lived up to the hype, coupled with the snub at this week's event, I figure Moshe left IBM for much the same reason Randy Moss left the Patriots.
Nobody wants to work where they aren't appreciated, right?
That's all for now, but I'll undoubtedly have more to say about all this "ado" in the coming days and weeks.