Startup Pillar Data Systems Inc. unveiled its Axiom storage array this week. But its technology -- already replacing...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
established vendors -- has been somewhat drowned out by the noise over the company's heritage.
Pillar is backed by Larry Ellison, founder and CEO of Oracle Corp., and as a result has wads of cash behind it. To date, the company has raised $150 million, and there's plenty more where that came from.
But what's more interesting is how Pillar has chosen to spend Ellison's dollars in the four years since it was founded. Its first box, available in July, will support NAS only, and comes with dual-ported 400 GB SATA drives. It scales from 2-3 terabytes (TB) up to 160 TB and is priced at $50,000 to $500,000. A SAN version of the system is in beta and will be available in a month.
Despite missing some key features, Pillar has found some early adopters to check out its technology.
Law firm Thatcher Proffit & Wood, LLP in New York is using the Axiom system at its disaster recovery (DR) site to replace an EMC Corp. Clariion 4700. The company still has 6 TB on an EMC Corp. CX 600 at its main site. "Pillar is unproven so we won't be swapping out this one just yet … and until they have asynchronous replication resolved in a serious way, we won't replace EMC," said Christopher Hill, associate director of information systems at the firm. Thatcher Profit is also waiting for a signoff from Pillar on its support of VMware.
Had Hill not found Pillar's system, he would have installed a small Hewlett-Packard Co. (HP) SAN at the company's DR site. A second EMC Clariion was out of the question, however. "Any technology at the secondary site is not benefiting the company unless there's a disaster, so we don't want to waste money putting expensive equipment down there."
Users say there is a four-fold price increase in the price of an EMC Clariion over Pillar's box. A 12 TB Pillar configuration costs $100,000, compared with a 6 TB Clariion for $400,000. "And there isn't a four-fold penalty in performance," Hill added.
Chris Butler, chief technology officer of I/PRO Corp., a Web site auditing company, expects to replace several old Sun Microsystems Inc. arrays as well as two large filers from Network Appliance Inc. (NetApp) with Pillar's Axiom, currently in test. His beef with NetApp in particular is the "single use model" of its filers. "You have to buy a specific model for the type of data you want to store but sometimes you don't know how you are going to grow … With Pillar, I can go up and down the different tiers within the same system and I don't have to have expertise on lots of different boxes."
Butler said he hopes the performance of the system keeps up as he adds more storage. "The proof will be in their ability to service it and how the ATA spindles hold up," he said.
Both these users said they wouldn't have looked at this technology had it not been backed by Ellison.
Nuts and boltsThere are three pieces to Pillar's Axiom system: The Slammer, which is the storage controller, can be SAN or NAS but will work together seamlessly; Bricks, which are the storage enclosures housing the disk drives and RAID controllers; and the Pilot, which is the management system.
The Slammer has an active-active architecture supporting four processors. By way of comparison, EMC's Clariion uses active-passive controllers. The active-active setup means that both controllers are functioning all the time and actively processing data I/O operations. This improves performance in addition to the increased fault tolerance. In EMC's Clariion, the second passive controller only kicks in when a failure occurs.
The Axiom can scale to support four Slammers or 16 processors, and the Bricks also have their own processing power. A single Slammer supports up to 24 GB of cache memory, compared with midrange systems from HP, EMC and IBM that support 4-8 GB of cache. The more cache, the better the application performance will be.
The system uses a virtualization technique to stripe data across hundreds of disks, versus 16 disks in the Clariion, and it localizes rebuilds on each Brick instead of trying to rebuild an entire 250 GB drive, which can take a whole day.
The Axiom NAS offering supports a file system that has been tested to 48 TB and includes a global namespace that isn't global, but can present a single file system view across multiple Axiom Slammer nodes in a single location. Pillar claims it has simplified provisioning down to a few clicks and a set of templates will set up resources automatically, according to the application being provisioned.
"Defining a LUN isn't simple: Take an EMC system, they sell a program for $50,000 just to help you tune," Workman said.ILM and disk short stroking Pillar's big claim to fame is the way in which it uses disk short stroking, which is not a new technique but has been adapted by the company for tiered storage. Instead of simply placing the most important data on the outer edges of the disk, or on more expensive drives where it is more easily accessible, Pillar's technology goes a step further and tiers the data within the disk itself.
In other words, the least important data resides in the middle of the disk, taking longer to fetch, while critical data lives on the outer edges where the robotic arm can reach it more easily. "With EMC, important data can go on the edge, but the rest of the drive is wasted as you can't put any more data on that disk," said Tony Asaro, senior analyst with Enterprise Strategy Group.
A spokesman for EMC argues that it doesn't matter where the data is stored if you have smart algorithms that can prefetch it. "We can figure out what data is going to be called for next based on what's just been called for and bring it in ahead of time."
Another startup working on optimizing storage for the most appropriate data is 3Par Data Inc. It recently introduced a dynamic optimization capability on its inServ array that enables administrators to convert a volume from one service level to another in seconds, the company claims.
3Par's service levels are mapped to four different criteria -- not just geographic data placement on the disks – which 3Par calls radial placement. It can also convert a volume from RAID-10 and RAID-5 as needed, choose between massively parallel or finely restricted use of ports, cache, processors, loops and spindles, as well as move data between high-performance FC drives or more economical SATA drives.
"We've always being doing radial placement of data," quipped David Scott, CEO of 3Par. "It's no big deal."
Click here for more of today's news.