In-Depth
Just NBOD It
How much would you spend for 2 TB?
A week or two ago, a spokesperson for Network Appliance, quoted in the trade press speaking about his new storage platform (the FAS 3040), stated that it provided the merits of enterprise-class storage to midsized companies for the affordable entry price of $83,000 for two terabytes. After RAID and snapshot features were applied, the resulting available capacity would be significantly less than 2 TB, but the platform, he noted, was extensible to more than 250 disk drives and over 126 TB of total capacity.
There were several unanswered questions in the press accounts. First, how did the vendor calculate that 2 TB of storage could possibly be worth $83,000?
They referred to the platform as a universal architecture, a description that was presumably meant to communicate value over time. The consumer could grow infrastructure along the lines of the architectural model offered by NetApp for the foreseeable future, meaning one less issue to trouble over: the important issue of how to build storage.
Second, the FAS 3040 offered all of the value-add features of its larger cousins in NetApp’s product family, from back-end connectivity to either Fibre Channel or iSCSI storage, to the ability to make disaster-proofing snapshots of data a routine that would bulletproof your data storage (assuming that you had another 3040 containing another set of the data somewhere outside the disaster zone). The 3040 MSRP is lower than the more pricy FAS3070, introduced to the market only a few months ago.
Third, purchasing the product bought you the NetApp brand name. In these days of burgeoning recession, and the conservative buying practices sure to follow, established brands should provide a sense of comfort to buyers.
However, even after considering all of these factors, $83K for two terabytes of disk still struck me as exorbitant. Two terabytes of SATA disk can be purchased today for less than $750 from any number of vendors. I can add software for cloning and taking snapshots from Acronis, Symantec, DataCore Software, or any several other commercial or open-source solutions for next to nothing.
Moreover, I have doubts about the performance of a storage array with a full load of 126 TB of capacity behind a single head—even a NetApp head, which is one of the best in the business. Add to that Sunnyvale’s tendency to take products off support prior to the completion of a service/warranty agreement, and its historical predilection to refresh its own product family without full support for backward compatibility with earlier models, and I have to wonder about the strategic architecture messaging the company is using today. By the time a company needs to grow its storage, its universal architecture might not be so universal.
Now, don’t get me wrong. There is nothing questionable about the Network Appliance platform that has reached my attention as of this writing. I just wonder whether it is worth the outlay, especially when you can buy a SNAP Server from Adaptec for about an eighth of the cost and which outperforms the NetApp box hands down.
Enter NBOD
Add to that an apples-to-oranges comparison to Zetera’s just-announced NBOD architecture. NBOD (networked bunch of disks) is the next stage of the continuing story of Zetera. Having made it possible to connect to disks using IP and UDP, Zetera says why not make the creation of volumes from a bunch of network-connected drives simply a function of the protocol?
Using UDP unlocks the ability of IP to multicast—to make selected subscribers to a multicast a collective target for block data. There is no rival to this technology in the market today, and despite the claims from some critics that it is not an enterprise-class storage alternative, I think the handwriting is on the wall.
For example, NBOD would provide a nifty solution for e-mail hosting from Microsoft. Redmond could complete its faulty “shared nothing” Exchange Server clustering story readily using NBODs. They could also enable the fielding of extensive object-oriented storage behind their Windows Storage Server R2 by connecting NBODs to the server, then sharing the storage out to applications via CIFS/SMB or as a block storage pass-through.
This begs the question, of course, of whether Microsoft servers and apps are “enterprise class.” Depending on your viewpoint, you might decide one way or another. However, current numbers suggest that Redmond operating systems are dominant in business, installed on the preponderance of both clients and servers hosting meaningful business data today—by a factor of 4 to 1.
With NBOD, Zetera continues to put pressure on Big-Iron guys in a not so “in your face” manner—mainly because of the focus of the company’s channel partners, including NetGear and Bell Micro, to deliver Zetera-based solutions to home users and smaller firms. Going forward, whether through these channels or newer relationships—or possibly by building its own gear—Zetera technology is bound to move up the food chain into the enterprise.
If we have the ability to disconnect the RAID and virtualization component from the array with Zetera, take a look at NEC’s just-announced HYDRAstor technology for the rest of the story.
HYDRAstor is basically a storage services software family, in some respects similar to Windows Storage Server R2, but targeted at traditionally UNIX-based application support. With HYDRAstor, it may well be possible to place all value-add functionality traditionally sourced from brand-name storage vendors into a network and to create huge grids out of whoever’s RAID, JBOD (Just a Bunch of Disk), or NBOD disk products you care to buy.
In the process of evaluating these options, ask yourself how any vendor providing commodity spinning rust in a branded cabinet can legitimize the huge price tag for what they are selling. Let me know what I’m missing here: jtoigo@toigopartners.com.