Sistina Makes Inalienable Writes

Having multiple servers simultaneously writing to the same file has tended to be a lot like a room full of people talking about different things at the same time -- the end result is likely to be an incomprehensible mess.

Now clustered file system startup Sistina Software Inc. says it has found a way to turn the loud, babbling hum into separate, intelligible flows of information. The company claims that the new version of its global clustering file system allows up to 256 nodes to concurrently write to the same file without corrupting the data (see Sistina Enhances GFS).

The Minneapolis company, which has been developing its global clustering file system software for Linux for the past five years, announced the availability of its GFS (Global File System) version 5.2 this week, touting better locking capabilities and a new data journaling feature for enhanced performance.

Distributed file systems like Sistina's GFS allow organizations to consolidate server and storage resources into a single management domain, linking thousands of different data storage repositories into a single SAN, which can be based on either Fibre Channel or IP. Essentially, each server shares a single file store and can serve files to other systems on the network via Common Internet File System (CIFS) and Network File System (NFS) protocols.

Like previous versions, GFS 5.2 is built to handle "locks," which is what prevents two machines from writing to the same area on a disk and thus corrupting the data. But in addition to that, the company says, the new software offers more flexible locking mechanisms that allow users to partition a single file into multiple secure parts. By enabling simultaneous writes to the same file, Sistina says it offers a tremendous scaleability and performance boost to enterprise applications like high-end distributed databases.The feature is a useful one, according to Dan Kusnetzky, VP of system software research at IDC. "They're attempting to go back in time to do what Digital Equipment [now a part of HP] did with VAX clusters, where all of the files on all of the systems were shared," he says. "That's always been the high-water mark for clustering technology, but it's still something many systems can't do today."

In addition, the GFS 5.2 software offers prepackaged OmniLock modules aimed specifically at quickly configuring high-availability, high-performance environments. This shortens the time needed to deploy mission-critical applications, the company says. Like previous versions, the new software offers meta-data journaling, which is a mechanism used to track meta-data changes like altered file-names. The new version extends this capability to include pure data journaling, which is useful for tracking changes to the data itself in small files like email.

Other features carried over from previous versions of the software include a dynamic multipathing capability, which ensures that there are two separate paths from the server to the storage device. If one path goes down, the system picks up automatically, Sistina says.

Sistina, of course, is not the only player on the clustered file system block. Companies like Advanced Digital Information Corp. (Nasdaq: ADIC), IBM Corp. (NYSE: IBM), PolyServe Inc., Qlusters, and Sanbolic Inc. also have offerings in this space, and Spinnaker Networks Inc. sells a NAS hardware system that incorporates a clustered file system (see PolyServe Musters Reliable Clusters).

Sistina, however, has been around for longer than most, and boasts more than 50 customers, including Deutsche Telekom AG (NYSE: DT). Joaquin Ruiz, Sistina's VP of marketing and product management, also says that the company's other product line, the logical volume manager (LVM), its open-source contribution to the Linux kernel, has about 50,000 installations. The company says it's constantly expanding its customer base, especially through its partnerships with Hewlett-Packard Co. (NYSE: HPQ) and Fujitsu Ltd. (OTC: FJTSY).As for the 5.2 version, Ruiz says Sistina has had a number of customers beta testing the product over the past few months, and that it already has 12 prepaid customers. "Obviously we're excited," he says. "This is the first time that we've presold so much. There was a lot of pent-up demand."

Despite its progress, however, Kusnetzky cautions that Sistina is bound to run into more competition going forward. "I expect we will see a lot of competition emerging that isn't visible today," he says. "Some of these people have a lot of clever offerings."

In addition, Sistina's decision to offer Linux-based software could cause a slower uptake of its products, he says: "One of the issues is that Linux has not yet reached mainstream status in all marketplaces... Like other vendors offering software on Linux, they face the fact that an education process is needed."

Ruiz, for his part, insists that Sistina's software is far superior to other offerings on the market, claiming that the GFS 5.2 is the only clustered file server that allows for the clustering of up to 256 nodes. (By contrast, PolyServe says its latest software can cluster up to 16 nodes.) Sistina says it already has one customer operating 150 nodes, but wouldn't tell us who that is.

Sistina, which employs 40 people and has received $20 million in funding to date, expects its new software will bring it closer to its goal of reaching the break-even point in the first half of next year (see Sistina Seeds Growth). If the company does need any more funding on its road to profitability, says Ruiz, it will come from strategic partners, not venture capitalists. "We're funded to cover our roadmap," he says.The GFS 5.2 software is currently available directly from Sistina, and costs $1,500 per node. Customers of previous versions are eligible for an upgrade deal at a lower price, the company says.

Eugénie Larson, Reporter, Byte and Switch