CAS at a Crossroads

When it comes to content-addressable storage (CAS), the word on the street is software, and a smattering of companies are beginning to pay attention.

Let's start at the top: CAS creates an object-oriented, searchable archive of so-called "fixed-content" data, such as medical images, that can be used for recovery after a failure. While EMC's Centera is the best-known CAS box, products from the likes of Archivas, Bycast, Nexsan, Permabit, Sun, and Veritas (now Symantec) have emerged in the last two years. (See StorageTek Rolls Its Own CAS, Veritas Archives Another Startup, and Nexsan Targets CAS Startup.)

Here's the problem: Most solutions require a vendor's own hardware or appliance. Additionally, getting some applications to work with a CAS system requires proprietary APIs (application programming interfaces). (See CAS Conundrum.)

But a couple of startups, including Archivas, Canadian startup Bycast, and stealth-mode Caringo (no Website yet), are leading the charge toward software-based wares that ultimately could prove more efficient -- maybe even cheaper than what's now available.

Of these, Bycast is farthest along. Based in Toronto, the vendor offers software called StorageGRID that works with a range of hardware platforms across multiple sites at the same time. Using a browser-based, Linux-based interface that runs under Linux, Windows, or Unix, StorageGRID parses metadata from CIFS, NFS, or HTTP-based files and creates objects from the metadata. It then uses rules to act on the objects. A file on a SAN, for instance, can be moved to tape after 60 days.Importantly, Bycast's StorageGRID also works automatically with medical imaging files based on the Digital Imaging and Communications in Medicine (DICOM) spec for Picture Archiving and Communication Systems (PACS), Part 10. That's a long way of saying StorageGRID supports the systems most hospitals use to store images like mammograms and other scans. And, since medical imaging is a primary market for CAS, that's significant.

StorageGRID's compatibility with medical systems and indifference to underlying hardware have led both HP and IBM to OEM it -- HP for its HP Medical Archiving Solution (MAS) and IBM for the IBM Grid Medical Archive Solution (GMAS).

HP insists that OEMing Bycast isn't competing against its other CAS system, the Reference Information Storage System (RISS), which it acquired with Persist Technologies in 2003. (See HP Adds Archiving Apps and HP Buys Archive Guys.) While RISS is a general-purpose ILM product, the Bycast-based MAS is better earmarked for healthcare customers, say HP spokespeople.

Bycast's software-only approach, as well as its medical expertise, appear to be putting it on the map. At least one user picked Bycast-plus-HP over EMC. Tim McFarlan, director of technology management at Phoenix-based Banner Health, says the Bycast solution was the only one not tied to a vendor's SAN that he could find back in 2004.

Banner Health is a system comprising 19 hospitals, several long-term care facilities, and numerous clinics throughout the Western U.S. McFarlan needed to store medical images across five hospital locations in a way that would guarantee the return of information after a failure."We were focused on a regional archive solution for PACS," he writes in an email. "We started with a solution proposed by Fuji, our PACS vendor, [which had] a partnership with EMC." However, EMC's synchronously mirrored SAN-based solution, which McFarlan installed in one hospital, didn't cut the mustard when there was a system failure. "We quickly became aware of the limitations of this configuration as far as providing fault tolerance at the hospital in the event of loss of our WAN or PACS application failure. There were also concerns about the lack of HSM management/administration by our IT folks that we were going to end up with dozens of SANs that were going to be very challenging to administer."

While Bycast won't say what it charges vendors like HP and IBM for its software, a spokeswoman says the package costs roughly $1 per Tbyte. Considering that CAS hardware/software packages can sell for $1,000 per Tbyte and more, it's clear there's room for some savings, despite the variable hardware costs.

Bycast isn't without its drawbacks. The company claims "dozens" of customers have glommed onto its wares. But the firm, founded in 2000 with funding in the range of $16 million, doesn't plan to extend beyond the medical vertical anytime soon. What's more, it still requires customization to import data in HTML and other languages that use tags.

But HP and IBM's adoption of Bycast signals the usefulness and adaptability of its software-based approach, which appears to be catching on: Archivas VP of marketing Asim Zaheer says his firm is bent on a software-only model that uses hardware from distributor/resellers like Dell and Source Code Corporation, the vendor's main hardware partner.

Archivas isn't focused on hardware margins at all, Zaheer says, even though about 40 percent of the cost of an Archivas CAS solution can be hardware. "We don't mark up hardware at all, only enough to cover costs." Archivas charges $15 to $20 per Gbyte for its solutions, and costs drop with the size of the overall solution. Prices can go as low as $10 per Gbyte for systems of 100 Tbytes or more, Zaheer says.Archivas's system supports NFS, CIFS, and HTTP data and will be compatible with DICOM Part 10 within the next 60 to 90 days, Zaheer says.

On the downside, Archivas doesn't have APIs, and if customers want to set some kinds of policies (beyond retention) to control the flow of data in the system, they must modify the applications some other way.

What of other vendors? A spokesman from Permabit says its software will run on other platforms, and the vendor is adopting that strategy, but no timeframe has been set. Nexsan seems intent on its own hardware. (See Yad Vashem Holocaust Museum.) As for EMC, the company has nothing to say or announce on this topic, according to a spokesman.

So far, it's too soon to tell whether the CAS tide will turn. One thing to watch for is Symantec's anticipated release of the CAS product for which Veritas acquired DataCenter Technologies last spring. (See Veritas Archives Another Startup.) Since Symantec has a huge market presence, isn't tied to a particular hardware platform, and isn't bent on a specific ILM niche, it has the chance to innovate. So far, though, no date has been set for delivery.

Mary Jander, Site Editor, Byte and SwitchOrganizations mentioned in this article: