StorageMojo @EMC World

by Robin Harris on Wednesday, 5 May, 2010

Leaving Sunday for beautiful downtown Boston. Really: much nicer since they finished the Big Dig.

Hoping to learn more about Atmos. Did you know it is available as an online service?

And I want to talk to some of the RSA researchers working on storage, but I’m not sure they are part of EMC world. Maybe I can squeeze in a visit to their Cambridge lab.

Any thoughts on stuff I should see? If you’ll be there and would like to talk please comment or email.

The StorageMojo take
Heard that of the 15,000 or so folks expected, some 5,000 will be carrying an EMC badge. Nice break from cubicle heaven.

Shades of DECWorld. But back in the 80s the minicomputer business was getting killed by PCs. The threat to EMC’s business is commodity-based scale-out storage, something they are wisely embracing with Atmos.

DEC founder and president Ken Olsen grew DEC for 30+ years, a record I don’t believe has been equaled by any other computer industry entrepreneur, though Bill Gates comes close. But even Ken didn’t know how to navigate the shift from vertical integration to assembling horizontal commodities. Nor did he see how commoditizing DEC’s proprietary technologies, such as DECnet, VMS and VAXclusters, could have ensured the company’s long-term survival.

That’s why Atmos is important to EMC. As our data continues to cool, commodity-based scale-out – or specialized hardware with commodity prices – storage is critical. Tucci seems to get that.

Courteous comments welcome, of course. Looking forward to seeing springtime Boston – if the weather cooperates.

{ 1 comment… read it below or add one }

John May 14, 2010 at 5:12 am

I’m possibly not the only one who would privately place DEC’s demise squarely at the feet of Palmer’s Microsoft love-in and the conflagration of Cutler’s departure, allegedly with copious 132 char – wide green and white paper, Hustvedt’s accident, and the fairly scary thought that VMS was really very complete even in the early 80s, with lock-up in some businesses it continues to enjoy for good reason. Amusingly, maybe as a psychic legacy, to this day VMS manages to be utterly schitzoid regarding TCP/IP stacks. And am I imagining things, or did Palmer enjoy a lucrative sinecure visiting Redmond for occasional meetings, about a whole statute of limitations later?

Anyhow, IIRC ca. ’82 DEC was the 800lb gorilla, eating IBM’s lunch, on the numbers, and i see a bit more than co-incidence and simple misdirection in their precipitious fall. More here than PCs and “open systems”, a bit more on which below.

Since, however, this is a EMC / big tin* / storage discussion, i recall SUN making overtures (not successfully in this direction) as to how ODS FS was inherently slow. Nothing to do with having proper integrity checks, cluster locking, things which kept our business in healthy sleep patterns. I seem to remember also this was the time SUN was boasting (very early 90s?) that LUL – London Underground Limited – had just replaced its signalling comptroller vaxen with U10’s or 20’s, being soooo much faster. That was about the time i started walking everywhere in London, rather than get stuck when the signals failed.

We’ve the same problem today. I’m harping on about this theme, i know. What i see is just massive overhead to achieve anything interesting, whether pure cluster, scale out, scale up, scale “yer Mom”. I don’t believe cycles are cheap, because they tend to need, well, some tending to. I now have to consider whether a switch really has “wire speed”, what vlans, bonding, L3 table size i’m dealing with . . . it’s on a par with running a modest sized ISP to connect some of these so-called solutions.

I think, harking back to that SUN vs VMS pitch, that we’re blind-sided by aggregated bandwidth whch is going in no particular direction, unless you’ve built your apps ground – up, (which case i contend you can do very well without buying packaged products), and ludicrous – i mean wonderful – IOPS. I’ll demure on IOPS, they’re really very very nice to have, but i contend they’re also a distraction, as all nirvanas are, even calling “c.p.” as to wear levelling and all that low – level bother you have to actually consider when purchasing.

Before i loose the plot: it is fairly obvious that big numbers do not a game win. But they are just as capable of disrupting well-formed technology today as they were when DEC realized, that for all their engineering chops and great attitude, they couldn’t market their way out of a wet paper bag.

Atmos is decidedly interesting. Like much or most of what’s out there, for this value of “interesting” i mean, you’re fascinated, because it requires true grit to find out what it is for your scenario.** But i disagree with the idea that data is simply “cooling”, or that Atmos, or much of EMC counts as “commodity-based scale-out – or specialized hardware with commodity prices” . I think that (“cooling”) merely means someone can’t make the call as to when data truly is archive, on its eventual limitations – timed way to the land-fill, and that for the rest, a big and nebulous “rest”, all those big performance numbers will lead before very long into real desire for better than near – line access. I think this will mimic the value curve of the internet, where ease of access to historical records, even of no obvious historiographical interest, has been a key component to positive “network economics”, or at least the fascination of the many who will, or might, pay. However the curve – ball thrown with vigor at my contention comes with full – twist along SOX – style retention and privacy axes.

Attempting to wrap, and get to some actual work, “open systems” to my mind all too often translate as “you can plug anything in” and so, well, “anything” just does get plugged in. Corollary to that, if i could attach any number of dumb disks to a vms cluster, and have an api / fs export that really worked, i think i’d happily forgo a smorgasbord of wire – level real – time network splits, block – level routers, wan – optimisers and all the gooey glue, because, well, i could. Yeah, HP, you could actually sock to Oracle whilst they bury ZFS under their stack!

For all my comments agin’ SUN, they’re to do with (it feels ancient) FUD over VMS, not to do with storage. Oh, OK, i have to say i think ZFS is freaking wonderful, but i also believe Oracle is working on how to use all those cores, about which i commented, append to a later date post of Robin’s, so i believe, whilst ZFS will not be submarined, we’ll not see it any longer as a potential panacea. Oh, the true cost of StorageTek!

Yes, maybe “shades of DECWorld” or rather any mature tech conference worried as to disruption. All pressing flesh, all beady eyes and confidence, but little looking under your own feet.

Regards to all & must say – whether i’ve added a jot or not to the discussion, very much enjoying responding to this site, so thanks again.

– john

* I say “tin” because “iron” suggests to me “wrought”, as in forged, a more wholesome and appealing visual metaphor than Campbells tins.

** I love being cold called. “What are you selling and how much is it?” covers all sins, and may leave your interlocutor in paroxysms of contorted verbeage which are fun to replay to the office.

Leave a Comment

Previous post:

Next post: