Riding a 10G gale at SNW Fall

Riding a 10G gale at SNW Fall

Despite some annoyances caused by Hurricane Wilma, the 2005 SNW Fall repeated the success of previous shows, albeit with perhaps just a little more spice. In fact, unexpected and off-stage news, namely HP finalizing the acquisition of AppIQ, and IBM planning -- together with other major storage vendors -- to extend open source software to the new frontier of storage management seemed to hit attendees just as much as Wilma's best efforts.

However important these two announcements were, one wonders if they really have anything to do with SNW. Probably not, but they certainly added a few decibels to the show's already noisy buzz.

The increased noise, however, didn't prevent me from noticing the drum rolls surrounding companies such as iVivity and Neterion. These vendors and their partners were promoting 10Gbps transfer rates for storage.

"Ours is the only storage network processor chip that can deliver sustained full 10-gig throughput," says Zulfiqar Qazilbash co-founder of and chief strategy officer at iVivity.

Qazilbash adds that the company's iDiSX2000 chip has been used in several OEM products offering lower-speed front-end connectivity, for example, in devices with multiple 1Gb iSCSI ports.

"However, our chip can also master the FC [Fibre Channel] protocol, so we decided to complete that equation with a new companion chip," Qazilbash explains.

In fact, iVivity used SNW Fall as the stage to announce its iFCC4210, a new chip that, like its iSCSI brethren, can push bits indifferently through a single 10Gbps port or multiple lower-speed FC connections.

The iFCC4210's FC features don't mean iVivity is ignoring iSCSI. Far from it. The company, in cooperation with partner Vativ Technologies, demonstrated storage access at 10Gbps delivered over plain vanilla copper wires. It's doubtful that many end-users will ever buy directly from either of these two companies, but expect to find their components in upcoming FC or IP-based 10Gb products.

"Our chip is not expensive," Qazilbash says. "We should be able to sell it in OEM prices close to $25 per board. Ours is a programmable platform that facilitates building any kind of storage."

Neterion is another name you should memorize for your high-speed networking needs, in part because the company just signed an OEM agreement with IBM to supply Xframe II Ethernet adapters for xSeries servers.

According to Philippe Levy, director of marketing for Neterion, the Xframe II takes advantage of the fast PCI-X 2.0 bus to reach transfer rates close to 10Gbps. Previous versions based on PCI-X 1.0 maxed out at 7Gbps in similar conditions.

The IBM deal was announced only after the show, but at SNW I saw Neterion's Xframe II actually used in 10GbE (Gigabit Ethernet) products from at least two of its partners, Intransa and String Bean Software.

Blame my hectic schedule, but I couldn't spend much time at the Intransa booth, where the vendor was showing its SAN attached to blade servers and also a prototype of a 10GbE SAN.

As for String Bean Software, I wasn't surprise to see its WinTarget apps shown at the Microsoft Pavilion, because the application scored some interesting high-performance results on Windows Server 2003 machines mounting XFrame II cards.

In case you forgot, WinTarget can turn any Windows server into an iSCSI target. Install free iSCSI drivers from Microsoft to your application servers, and you have what's probably the easiest-to-implement and most affordable IP SAN on the market.

WinTarget is obviously a tempting option for small Windows shops still fiddling with DAS (direct attached storage), but the obvious question is "Will it scale?"

"The 10GbE/WinTarget results we published [at SNW] will certainly advance the discussion of iSCSI scalability in the market," says Mickey McIntire, CEO of String Bean Software.

In fact, one of the benchmark data points shows a transfer rate close to 9Gbps. Check out those results here, but before assuming that you could easily do the same on your servers, bear in mind that the tests were run on RAM disks, essentially using memory as the target, rather than disk drives. This eliminates the latency caused, for example, by seek and rotation delay.

Of course, this doesn't invalidate the benchmark, but to attain similar results in a real world setting you may need a massive battery of disk drives and controllers.

"We've certainly contributed to eliminating the myth that the 'plumbing' won't support the kind of performance/scalability needed for more advanced iSCSI-based solutions," McIntire says.

This is exactly what vendors like iVivity, Neterion, and others are trying to prove.

Follow Us

Join the newsletter!


Sign up to gain exclusive access to email subscriptions, event invitations, competitions, giveaways, and much more.

Membership is free, and your security and privacy remain protected. View our privacy policy before signing up.

Error: Please check your email address.


Show Comments