I am recovering from the wonderful meal prepared by my wife, son, and kin last night for our Turkey Day festivity and working on a piece about disaster recovery for Microsoft Exchange, SQL Server and SharePoint.
The meal didn’t cause me any indigestion — mainly because I didn’t eat very much (too busy serving as host for the massive event). What has twisted my guts a bit today are two things. First, the Frenchman sent over his latest issue of Storage Newsletter.com (which I read religiously). His lead story is about Fujitsu’s latest Eterna-line storage array announcements. Read the whole story here.
The quote that jumped out at me:
Dr Helmut Beck, Vice President Storage Business at Fujitsu Technology Solutions, says: “All ETERNUS DX systems offer high-end features and set new standards in innovation. The systems provide multi-dimensional scalability and the capability to maximize the pooling of storage resources. Eco-mode functionality places ETERNUS DX400/DX8000 series among the most energy-efficient systems in their class, and virtualization capabilities including Thin Provisioning make the systems ideal building blocks for Dynamic Infrastructures.”
While I have no axe to grind with Fujitsu or anyone else in this industry per se, I do wonder why it is worth bragging that a vendor has added “thin provisioning” and others of a long list of “innovative features” to their boxes. We haven’t, it seems, learned anything from the first decade of the 21st Century. We still see the vendors pushing more junk functionality onto their array controllers — functionality that all customers neither need nor use, but must pay for nonetheless. Moreover, I sense that many of these enhancements are being added, not because of any demand from consumers, but just to seem ”competitive” with other vendors who have them — like IBM or EMC or…
Part of this phenom may reflect the current operational reality in many IT shops. Many layoffs, fewer personnel to manage storage, storage capacity growth = traction for marketing messages about doing more with less (aka autonomic storage capacity “management” functions like on-array thin provisioning). The problems, as I see them (and maybe I ate too much Thai Curry or Oyster Dressing last night), are that on-array thin provisioning approaches
(1) use proprietary forecasting algorithms to anticipate storage capacity demand — forecasting engines that are fundamentally flawed by virtue of the fact that storage demand may not be as readily predictable as the vendor suggests, opening the door to a disaster if an application ever demands all of the space it thinks that it owns in a sudden “margin call”
and
(2) discourage companies from doing anything whatsoever to address the real problem of the storage junk drawer – the data glut that sees so much undifferentiated and mostly junk data occupying spindle space forever, which in turn requires adding more expensive “thin provisioned” storage to the pool. This plays into the hand of the vendor who charges customers, to quote FalconStor CEO Huai, $60 per GB for disk space that costs him 16 cents per GB. (See Huai’s interview at the C4Project.org for more insights from this fascinating fellow.)
The second thing driving vendors to add “on-array thin provisioning” and other “innovative features” to their wares is, of course, the financial analyst. The too-often-storage-knowledge-impaired expert tech analysts of the stock market world have made it clear that they view any vendor who isn’t adding junk functions to their array controllers as yesterday’s news — they are just not keeping up with the pack. Every six months or so, an EMC, IBM or E-I-E-I-O brand vendor must announce sexy new functionality, or he appears to be standing still. That is how bullshit becomes institutionalized, not to mention how things like market meltdowns begin.
Why don’t these vendors put their sexy new functionality on an optional appliance, like a virtualization controller on some generic software hosting platform? Oh wait. Then they would be like DataCore or FalconStor. Plus, they would lose all of that chewy proprietary goodness that locks customers into their box of spinning rust and locks out competitors’ boxes of spindles…even though every storage vendor is selling the same box of overpriced disks.
The second contributor to today’s antacid addiction is a post at StorageRap. I don’t know these folks from Adam, but their blog earlier this week cited my tale of woe about long rebuild times on a RAID 5 set to promote, of all things, 3PAR thin provisioned storage!
The title of the post was telling: “The Cult Leader of Cheap Storage Guzzles His Own Medicine.” I read the piece because I wanted to know who the cult leader of cheap storage was (wanted to send him or her my best wishes, in fact). Turns out he was referring to me!
Moi?! Cult leader? Wow, what a handle — and given that I have no cult of followers I am aware of, that is an interesting notion. I have no Palin-esque Facebook page or Fox News Channel/conservative talk radio show bully pulpit. With the exception, perhaps, of my tongue-in-cheek ALT.SNW event in Orlando a couple of years back, I haven’t staged any storage Tea Parties or 9/12 Rallies lately.
I have a few firearms, but my house hardly comprises a cultist compound populated with starry-eyed followers.
I write a lot, but hardly expect to change views of consumers who still take their cues from the likes of Gartner, IDC, ESG, or the other folks deriving their revenues from a business model that pushes “insightful reviews of and commentary about technology” as dictated by the vendors who pay them — whether or not they actually provide value to the consumers.
As for guzzling my own medicine, the writer left out two important facts.
First, the Promise VTRAK array in question pre-dated the proliferation of PMR technology (high capacity) SATA drives. It did not support RAID 1 or ”10.” We had pulled it from our test bed after several years of service, where it performed admirably as an iSCSI target with lower capacity drives — rivaling Dell EquaLogic and even pre-HP LeftHand gear in their day. Having decommissioned the array at the shop, and I adopted it to store some of my millions of PowerPoint decks and art files and video clips in my home office.
I threw some high capacity drives into the array because I had them laying around. Given the vendor advertised failure rates of current generation SATA drives (five years), I figured that I had somewhere in the neighborhood of 24 months before I needed to worry about RAID 5 rebuild times.
Would I have gone this path in designing a production system in a business setting? Do I look new?
The second thing that the post ignored was that thin provisioned systems do nothing to resolve this problem — at least nothing related to thin provisioning. 3PARs RAID rebuild technology has nothing to do with thin provisioning.
Frankly, I would prefer to replace RAID altogether with something like RAGS, which Xiotech-channeling-Seagate-Advanced-Technology-Division has neatly added to its ISE technology arrays. RAGS dramatically shortens rebuilds by mirroring spot rights on different platters as a copy-on-write strategy I like a lot. They are currently only doing this with FC and LCFC drives, not SATA — and certainly not with the consumer grade SATA drives I was using in my home office. Now, ISE is what I WOULD USE to build a world class, business class storage infrastructure.
One last thing: the Promise folks are bending over backwards to get my problems resolved, which may have to do with a controller issue, not a RAID rebuild problem as originally thought. I like those folks and think that they make some respectable gear. Low cost to be sure, but solid for the three or four years that we used it for iSCSI testing.
Happy post-Turkey Day to my cult followers.
You must log in to post a comment.