Pillar's midrange array gets by on charm

Pillar Data users say the startup is aggressive in its pricing and customer service, but the technical merits of its Axiom box still leave a bit to be desired.

Almost six months after Pillar Data Systems Inc. announced its midrange array, the Axiom, early adopters say the price is right and the startup's extra attention to customer service make it a good buy, but that they continue to wait for technical improvements to be made.

The Axiom consists of three main elements: the Slammer, or storage controller; Bricks, which house the disk drives and RAID controllers; and the Pilot, which is the management system. The product's announcement in June extolled its technical virtues, including disk short stroking, which tiers data within a disk.

The company also has the backing of Larry Ellison, founder and CEO of Oracle Corp, and as such, has gotten off to a far stronger start than most startups. As of the announcement of the Axiom, the company had raised $150 million in funding. Currently, that number stands at $180 million and Pillar officials said they expect it to top $200 million by the end of the year.

Related articles

Little guys with big chops

Ellison startup taunts EMC, NetApp  

Midrange arrays get a makeover

But Randy Bucking, senior network manager for Swift Trade, Inc., a Canadian securities company, said those talking points had little to do with why his company selected the Axiom. For one thing, the company doesn't have a single Oracle database installed -- its Windows servers are running MySQL and its Sun infrastructure runs Sybase exclusively.

For another, they haven't actually seen any benefit from the disk short stroking yet.

This is because Swift Trade is still waiting for Pillar to sort out a firmware upgrade necessary for full compatibility with the Cisco switches Swift Trade uses exclusively in its shop -- "front to back," said Bucking.

Right now, Swift Trade's Cisco Catalyst 6500 switches are running a protocol called LACP, which stands for Link Aggregation Control Protocol. Similar to EtherChannel, the protocol combines Ethernet links for redundancy and to increase their speed. The protocol relies on two main algorithms, one called Round Robin and another hashing algorithm. Bucking said the Axiom's Pilot management system has so far proven incompatible with the hashing algorithm.

This incompatibility doesn't mean the Axiom can't work with the switches at all -- it just means the company's transfer of data will be slow. Therein lies the problem. As a company processing millions of transactions a day, Swift Trade's most critical data has no tolerance for latency.

Ironically, this intolerance for latency was what had made Axiom's disk short stroking feature appealing for Swift Trade. But without the LACP compatibility, any reduction in latency gained by that process is negated, according to Bucking. As a result, Swift Trade has yet to put its Axiom systems into production and can't until the issues are resolved.

"It's something they can't seem to duplicate in their lab. They haven't come up with a solution just yet," Bucking said.

But, he added, he wholeheartedly believes they will.

Why? "They've shown us great customer service so far. They've shown a great responsiveness and willingness to work on our problems," he said. "A lot of times we felt like we're kind of in the dark up here in Canada. Other companies haven't been aggressive enough with their service to continue to win our business -- Pillar has."

Bucking said Pillar had hired an engineer stationed in Canada specifically to service Swift Trade's shop, a move that impressed him.

Meanwhile, Bucking said Swift Trade is still waiting for Axiom to achieve compatibility with its iSCSI infrastructure and 10 Gigabit Ethernet (GigE) LAN; they plan a migration of their data from their outdated Network Appliance Inc. (NetApp) FAS 250s over to a temporary CIFS and NFS system until the Axiom can work with iSCSI and support 10 GigE) NAS ports. Bucking said Pillar has promised the iSCSI and 10 GigE compatibility in the first quarter of 2006.

Given then, that most of what Bucking stated was his technical attraction to the product has yet to come to fruition, what's the real difference? Cost.

"Basically, the Pillar system was beyond cost effective," Bucking said. He wouldn't say exactly what that price was, but it was 20% lower than NetApp's price for new filers and also significantly below the Dell Inc., Hewlett-Packard Co. and EMC Corp. products the company also considered.

"Any of those products probably would have worked," Bucking admitted. "But Pillar was much more aggressive in cost."

Ethan Erchinger, operations manager for Plaxo Inc., an online information management company based in Mountain View, Calif., has a similar story. He raves about Pillar's cost-effectiveness and, like Bucking, uses the word "aggressive" to describe its customer support.

Erchinger praised Pillar for sending SAN designers to work with Plaxo's network administrators to meet the company's admittedly unusual needs. Like Swift Trade, Plaxo has no tolerance for latency, but is using Fibre Channel network protocols and a Brocade-switched SAN fabric, and so far has not encountered any compatibility issues.

But Plaxo's deployment of the Axiom has still not been without its own technical hitches.

When first installed, the Pilot's reporting tool had bugs, reporting incorrect information and performance statistics off by a decimal point. The software also sent "spurious" error messages about back-end maintenance issues to network administrators who could do nothing about them.

The company has also had to work extensively with Pillar to work out a SAN design that will work with an idiosyncrasy of its MySQL environment. "Our hot backups in MySQL require us to back up straight to disk, all at once," Erchinger said. "It's not like Oracle, where you can stream data over the network."

Erchinger said the reporting tool bugs were quickly resolved, however, and the SAN design issues were solved with the purchase of four Axiom boxes from Pillar, which were then set up in a complex failover system that ensures that primary and secondary copies of each of its databases are never stored on the same box.

Here, Erchinger said, Pillar's cost advantages proved crucial. "Buying four boxes instead of two as we originally planned was not a huge cost differential for us," Erchinger said. "And with more boxes, we can more easily scale capacity and throughput." Erchinger also said his shop liked the disk short stroking feature, but was still hoping Pillar would make improvements so they could move LUNs on each disk on the fly.

"Right now they have a certain quality of service model on each disk for each LUN and it stays that way," Erchinger said. "I'd like to see this be something we can change, so that if we need to fail over to backups stored in a secondary portion of the disk, they can be given primary bandwidth."

Still, Erchinger said, he's been pleased overall with the Axiom and with Pillar as a company. "They don't have tens of thousands of customers to deal with," he said. "We've been able to escalate issues fairly quickly."

Dig deeper on Disk arrays

Pro+

Features

Enjoy the benefits of Pro+ membership, learn more and join.

0 comments

Oldest 

Forgot Password?

No problem! Submit your e-mail address below. We'll send you an email containing your password.

Your password has been sent to:

-ADS BY GOOGLE

SearchSolidStateStorage

SearchVirtualStorage

SearchCloudStorage

SearchDisasterRecovery

SearchDataBackup

Close