Storage Soup

Jun 25 2008   12:17PM GMT

The enterprise and open-source storage

Beth Pariseau Beth Pariseau Profile: Beth Pariseau

Jesse at SanGod wrote an interesting post the other day entitled “Enterprise storage vs….not.”

I have a cousin. Very well-to-do man, owns a company that does something with storing and providing stock data to other users. I don’t pretent do know the details of the business, but what I do know is that it’s storage and bandwidth intensive.

He’s building his infrastructure on a home-grown storage solution – Tyan motherboards, Areca SATA controllers, infiniband back-end, etc. Probably screaming fast but I don’t have any hard-numbers on what kind of performance he’s getting.

Now I understand people like me not wanting to invest a quarter-mil on “enterprise-class” storage, but why would someone who’se complete and total livelihood depends on their storage infrastructure rely on an open-source, unsupported architecture?

Jesse goes on to point out the resiliency and services benefits of shelling out enterprise bucks. His post sparked a conversation between me and an end user I know well whose shops (in the two jobs I’ve followed him through) are as enterprise as they come. This guy also knows his way around a Symmetrix, and Clariion, and NetApp filers, and when it comes to the secondary disk storage and media servers he’s building for his beefy Symantec NetBackup environment…he’s going with Sun’s Thumper-based open-source storage.

Obviously it’s a little different from cobbling together the raw parts, and Sun offers support on this, so it’s kind of apples and oranges compared with what Jesse’s talking about. But I’ve also heard similar withering talk about Sun’s open storage in particular, and can only imagine Sun’s open-source push is making this topic timely.

This is the second person I’ve talked to from a big, booming enterprise shop who picked Thumper to support NetBackup.  The first, who had the idea more than a year ago, was a backup admin from a major telco I met at Symantec Vision.

Obviously it’s not mission-critical storage in the sense that Symmetrix or XP or USP are, but I’d venture to guess that for a backup admin, his “complete and total livelihood” does depend on this open-source storage. As for the reasons to deploy it instead of a NetApp SATA array or EMC Disk Library or Diligent VTL? Both users cited cost, and the one I talked to more recently had some pointed things to say about what enterprise-class support often really means (see also the Compellent customer I talked with last week, who found that the dollars he spent made him less appreciative of the support he got from EMC).

This ties in with a recent conversation I  had with StorageMojo’s Robin Harris. He compares what’s happening in storage to the relationship between massively parallel systems and the PC in the era of the minicomputer.  When the PC arrived, the workstation market was dominated by makers of minicomputers, the most famous being Digital. Minicomputers were proprietary, expensive and vertically integrated with apps by vendors, much like today’s storage subsystems. Just as the PC introduced a low-cost, industry-standard workstation and the concept of a standardized OS, Harris predicts clustered NAS products built on lower cost, industry-standard components will bring about a similar paradigm shift in enterprise storage.

While there will obviously remain use cases for all kinds of storage (after all, people still run mainframes), I suspect people are starting to think differently about what they’re willing to pay for storage subsystems in the enterprise, regardless of the support or capabilities they’d get for the extra cash. And I do think that on several fronts, whether open-source storage or clustered NAS, it is looking, as Harris put it, like the beginnings of a paradigm shift similar to those that have already happened with PCs and servers.

That’s not to say I think Sun will win out, though. For all Sun’s talk about the brave new world of open-source storage, I haven’t heard much emphasis placed on the secondary-storage use case for it. And that so far is the only type of enterprise deployment for Thumper I’ve come across in the real world.

4  Comments on this Post

There was an error processing your information. Please try again later.
Thanks. We'll let you know when a new response is added.
Send me notifications when other members comment.
  • Carter George
    Sun's Thumper is a perfectly good feature-rich Unix file system. There are others like it, but combined with Sun's strong and reasonably-priced Thumper hardware, it's an all around good middle-of-the-road storage solution. What's amazing to me is the positive spin on ZFS - it's just a well-featured single node Unix file system. Better than the free ones you get with Linux, but it's not clustered, not distributed, does not have a global namespace to stitch together "bricks" of commodity storage in to a unified fault tolerant whole. It's none of those things. What will be far more challenging to the existing enterprise status quo is that the big boys are all bringing out scalable commodity storage offeringsof their own, ones that will undermine the margins of their own high-end Enterprise storage platforms. EMC has announced Hulk and Maui, and is even shipping Hulk, if not the more elusive Maui. IBM is readying XIV. HP has announced their CFS-based Extreme Data Storage built on Proliant blades. There are your top three technology vendors, period, launching scalable, commodity based storage platforms. It's one thing for Ibrix, Isilon, Panasas, and PolyServe to bring those things to market - it's another when it's HP, IBM, and EMC. So the landscape is changing, and the relentless drive to commodity price points that wiped out the Enterprise Server 5 years ago is indeed about to start on storage.
    0 pointsBadges:
  • Karl Lewis
    SUN's acquisition of Lustre has the potential to make them the risin star (pun intended) of scalable, commodity storage. Combining Lustre's ability to scale performance across nodes with ZFS' immense scalability and rich feature set adds up to a rich product set. I've heard people refer to this mythical combination as "the death-knell of enterprise storage vendors". Being able to use cost-effective Thumpers as storage bricks and combine them in interesting, automated ways will certainly make users think twice about their next storage purchase. But, who knows if this product combination will ever exist?
    0 pointsBadges:
  • Steve Chalmers
    Sitting in an obscure corner of the Gates building on the Stanford campus is a computer motherboard and some disks in an enclosure made of Legos. It's protected in a display case, of course. That was Google's first computer. Up 101 in Mountain View in the computer museum is a rack, with flimsy sheet metal trays to which computer motherboards (and disks) are attached. That's Google's first cluster. I work in the storage business, and have never set foot at Google. Best we can tell, nearly all of Google's business today is run off of huge clusters of client motherboards + client disks, attached to trays descended from what's in the computer museum. If a tray fails, the software works around it by rerouting work to one of the trays with a duplicate of its data. Compared to that solution, the Tyan motherboard, external SATA controller, InfiniBand, and probably premium SATA disks would seem gold plated. And most importantly in financial services where milliseconds matter, those enterprise array controllers don't add their millisecond (or yes, these days, even a tenth of a millisecond) to the latency of an I/O. Oh, and the Thumper would cost far more than the Tyan homebrew. Still vastly less than a DMX or competitor. From the point of view of preserving my job, those kinds of solutions scare me. But the software cost of developing them is quite high, and having a person on site with both skills and parts to repair is also expensive. So for now I sleep at night.
    0 pointsBadges:
  • Jesse
    Centerra - largely a "cheap" storage solution supported by a comprehensive software package. It's a great option for archival and would probably be suitable for a high-bandwidth archiving application if it was capable of...well, higher bandwidth. Support and resiliency - that's a great way to put it. You might be able to put together a solution that is resilient through multiple Raid levels. (ID Software Raid-5 striped across hardware Raid-5 or something equally complex) but you're still the one getting the page at 4am when the system jumps the shark. Personally I'd pay big money *NOT* to be the one getting the page at 4am. Especially if it's someone else's money. :) I prefer the casual email the following morning from EMC stating - "A drive went bad, we fixed it, no host impact."
    0 pointsBadges:

Forgot Password

No problem! Submit your e-mail address below. We'll send you an e-mail containing your password.

Your password has been sent to:

Share this item with your network: