Startup Touts Massively Scaleable File Storage

A tiny Cupertino, Calif., startup is launching network file virtualization on a grand scale with open-source software and standard x86 hardware.

Parascale boasts three beta customers for a file system called the Virtual Storage Network (VSN), which uses Linux-based servers to create a massive "file grid" comprising one virtual file system -- or more, depending on the application.

A typical VSN, execs say, comprises 16 to 32 storage nodes managed by one control node. Nodes consist of individual Linux file servers and SATA storage from the likes of Dell, HP, or IBM. Parascale VSN "file grids" can support multiple Pbytes, the vendor says. Management software ensures load balancing, automatic file migration, and automatic replication among all nodes.

Parascale's VSN is set for general availability in November. Pricing will be on a subscription basis and range from about $300 to $700 per spindle annually, depending on options.

The value proposition is clear: reliable high-volume storage without specialized hardware, something Parascale says is particularly in demand among companies with a lot of rich media content. Indeed, Parascale's gunning for customers who'd otherwise use clustered NAS la Isilon and others. It's also pitching against high-end clusters and emerging services from the likes of Amazon and emerging storage service providers.There's lots of competition there, but Parascale is ready with an answer for everything. The startup says it's more scaleable than clustering software and does not require client agents or software. It is cheaper than NAS for users with lots of enormous files requiring a single namespace. And as for competing distributed storage services, Parascale says it doesn't restrict file size and supports standards like NFS, HTTP, and FTP, instead of relying on proprietary protocols.

"Everyone loves his first NAS. But to increase capacity you need to forklift-replace it or bring in another file system," asserts CEO Bill Evans. "We are not a new file system... We write software that aggregates existing Linux file systems to create a large storage pool."

At least one of Parascale's early adopters is pleased so far. "I like how I can use any off-the-shelf hardware for the Parascale product. It allows me to choose my own levels of cost and reliability and scales easily," writes Jonathan Bensen, director of IT at WAN optimization supplier Blue Coat Systems, in an email today.

Bensen and his group have been beta-testing the Parascale software since January 2006 for "cheap, fast storage" in applications that don't necessarily require the reliability of more costly and road-tested NAS. "We are using the system for backup [and] staging," he notes. "We use it with [Symantec] Netbackup 5.1 and the built-in disk-stage feature. This allows us to stream backups to disk, instead of slow tape, reducing our backup window. The backups are then flushed to tape during the day."

Bensen reports that Parascale is "extremely reliable. We backup about 3 Tbytes... a week, and have not had to touch it since the installation."Another beta-tester is attracted by Parascale's potential to speed up NAS functions across large installations thanks to its distributed architecture of "loosely coupled" storage nodes coordinated by a single control node.

"For our uses we don't see much value in a NAS vendor's ability to expand a device to store 50TB on a single NAS device if they still require all clients to funnel through a bottleneck of just a few or potentially only a single GigE port," asserts Nathaniel Burger, director of IT at Scientific Drilling in Paso Robles, Calif. "The Parascale solution appears to give us the ability to both easily configure the total amount of available storage as well as tune the total available bandwidth that clients have to access that storage for a much lower total cost."

At least one analyst thinks Parascale will appeal to the need for fast, scaleable storage by organizations like Burger's, but it won't displace NAS easily. "You could argue that if systems like these catch fire, the NAS market would get blown apart and margins go to hell in a handbasket," says Arun Taneja of the Taneja Group. "That won't happen. NetApp has huge bundles of software for its NAS hardware... These are nontrivial technologies... But a lot of users, particularly in oil and gas, have a requirement for this kind of approach."

But Parascale claims it's been hard at work for 30 months engineering nontrivial technology of its own. "This is a long-gestation product," Evans asserts. "It takes five to six years to produce a mature file system." He points out that Parascale, which has about a dozen employees now, has a number of patents pending, including ones for distributed clustered storage, a parallel virtual file system, distributed "cache coherency algorithms," "multiprotocol file routing," and other intriguing functions.

At least some of these features are reminiscent of Scale8, a firm in which Cameron Bahar, who co-founded Parascale in 2004 with Rasoul M. Oskouy (ex-Juniper, Sun), was a former executive. Ill-fated Scale 8 was also involved in large-scale distributed storage and blew through $55 million in funding before collapsing in 2003. (See Scale8 Flatlines.) That startup tried unsuccessfully to peddle its intellectual property before going bust.Parascale denies any link, however, between Scale8 and Parascale.

Before Scale8, Bahar worked at HP, Teradata, and Locus Computing. Evans (ex-Gateway, Blue Martini Software) joined about 18 months ago.

The volatile storage market is tough to predict, but Parascale appears to have its marketing plan firmly in place. It also has undisclosed seed funding from Oskouy's investment firm and is working toward a series A round, which execs indicate is imminent.Have a comment on this story? Please click "Discuss" below. If you'd like to contact Byte and Switch's editors directly, send us a message.