Lee Campbell, network engineer at the San Diego-based Salk Institute, knew that once Network Appliance Inc. (NetApp) acquired Spinnaker Networks Solutions Inc. in November 2003, his days using the startup's high-performance NAS file system were numbered.
"It was a dead product once NetApp got hold of it, and we knew that," he said. But he was unprepared for NetApp's lack of support for the product, which was still under maintenance and the company's disregard for the problems he encountered using the technology.
For the most part Campbell was impressed with Spinnaker's product -- a clustered file system that supported NFS and CIFS and up to 512 servers clustered together to provide 11,000 terabytes (TB) (or 11 petabytes) in a single global file system. Dubbed the SpinServer, its target was companies with large compute clusters requiring high-performance access to file-system data.
"There was nothing else like it on the market," Campbell said. He bought SpinServer for the computational neuroscience lab at the Institute where researchers, including himself, a Ph.D in neuroscience, were simulating brain processes and storing the images online.
Campbell deployed a 5 TB SpinServer in July 2003 and immediately found a problem with the high-availability (HA) failover detection system. A failure could be bad enough that clients could not access data but not bad enough to alert the system. "It has an inadequate detection for failover," he said.
Contacting NetApp about the problem, Campbell discovered that the company has a category of failures it refers to internally as "cosmic failures," which are problems that it can't, or won't, fix. The failover detection fault the Salk Institute experienced was deemed a "cosmic failure," and Campbell never heard from NetApp on the issue again.
He was surprised at the company's lack of concern. "I'm afraid of going forward with them … I wasn't impressed with their support … They said we're not reinventing this for you, and yet it was sold as [supporting] HA."
According to Campbell, a second problem on the SpinServer involved an error correction code failure on a memory chip. The company sent out the wrong replacement part resulting in five days of disruption. The final straw broke when NetApp pitched Campbell a new 40 TB filer for $1 million dollars. "I was so insulted, I threw them out," he said. "They came back a couple of weeks later with a totally different quote."
Campbell was so soured by his experience that he's reluctant to look at anything from NetApp, including the upcoming Data OnTap GX. This is the product that NetApp has spent more than two years developing, merging Spinnaker's OS with its own NAS file system.
"These features are hard to make [work] without trying to integrate them," Campbell said.
Instead, he is evaluating EMC Corp., Apple Computer Inc., BlueArc Corp., Sun Microsystems Inc. and Isilon Systems Inc. file systems as possible replacements for Spinnaker. "I'm not happy with anything NetApp has shown me," he said. "We'll keep the Spinnaker and forklift [upgrade] when the time comes."
Data OnTap GX late
NetApp declined to comment on the Salk Institute's problems but separately it has acknowledged that the integration process with Spinnaker has been hard. During an earnings call last November, NetApp's CEO Dan Warmenhoven said that the launch of OnTap GX would be delayed by a quarter, "due to the complexity of integrating millions of lines of code." According to industry analysts, the first product to support the new OS is code-named Excelsior and will be demonstrated at the company's analyst day on March 14. It is a new high-end box positioned above the FAS 980, scaling up to 64 TB with double the performance.
Going forward, all new hardware from NetApp will support its existing OS, Data OnTap 7G and Data OnTap GX. Customers will require more complex hardware configurations to run OnTap GX as it supports four-node, eight-node and eventually 16-node failover. This requires different cabinet structures, interconnects and internal networking between nodes not necessary on NetApp's existing two-node failover systems. The first release of OnTap GX will not support all NetApp's applications, forcing users to make a choice of whether they want all those applications or some from OnTap GX and some new features, like the four-node failover capability.
"This is a major [product] transition for NetApp," said John Webster, founder and analyst with the Data Mobility Group. He says the company has to be "very wary" of its users looking at alternative suppliers as it goes through the transition. He drew an analogy with IBM's effort to kill its virtual machine/virtual storage engine OS and move people over to AS400. "The customer base looked at alternatives … HP [Hewlett-Packard Co.] got wind of it and put together a program to win over VSE users -- it was very successful."
More turmoilAnother sign of the turmoil around OnTap GX is the recent departures from the group developing this product. George Vaiser and Jeff Whitney in sales and marketing recently quit NetApp to join Maxxan Systems Inc., according to a press release issued by Maxxan on Jan. 31. Both were at Spinnaker when NetApp acquired the company. Furthermore, Jeff Hornung, vice president and general manager of enterprise file services at NetApp, formerly vice president of marketing at Spinnaker, is expected to leave soon. According to sources close to NetApp, all three were frustrated with the delays and internal politics besieging the OnTap GX rollout.
Analysts say its unfortunate NetApp let the Salk Institute slip away as it was a ready-made OnTap GX user. "EMC is chomping at the bit waiting for this product to come out," said Arun Taneja, founder and consulting analyst with the Taneja Group.