Brocade's Silkworm 12000 Fibre Channel switch has taken some hits for falling short of the high- availability standard set by McData and Inrange directors. Derek Granath, Brocade's director of product marketing, core switch platform, says a firmware upgrade to version 4.1 later this year will narrow that gap.
The upgrade adds hot code activation capability to the existing hot code load capability. That will eliminate rebooting to activate new code - about 20 seconds, plus additional time if servers need to re-initialize paths. OEMs will qualify the new firmware toward the end of 2002. Brocade expects to ship products in the first quarter of 2003.
Will hot code activation make the 12000 a high availability switch?
Arun Taneja, senior analyst, Enterprise Storage Group, says today's 12000 is really a pseudo-director, but thinks "it is designed to be a director and will get it there in time."
But Greg Schultz, director of storage networking at Inrange, points to the common chassis and controller shared by the 12000's left and right halves as an example of a single point of failure that separates the switch from a true director.
Yet experts agree that most switch failures are the result of human error, primarily incorrect changes in configuration. The only corrective for that is to maintain dual redundant switches.
McData's Jeff Nelson, engineering fellow, says that "Redundant fabrics eliminate some of the problem, but you're going to force some external software to identify a problem and make the switch. And if it's a transient problem, can the external software (such as EMC PowerPath) identify that, or will it ping pong?"
Taneja says that while dual director configurations are popular now, "with larger directors like the 256-port versions, we will see single configurations, especially after these products are deemed to be five nines."