The debate is over. Ten gig Fibre Channel will not step outside of the inter-switch link (ISL) space unless it be for applications demanding exceptional SAN bandwidth. This advantage will last until about 2011 when 16Gb/sec Fibre Channel - 16GFC in QLogic parlance - should arrive and negate 10GFC's speed advantage.
There is a whole new Fibre Channel idea here. It is a roadmap and, courtesy of Skip Jones, QLogic's director for planning and technology, we were shown a sight of it. First of all two types of Fibre Channel link are defined: Base10, meaning numbered in tens and starting with 10Gb/sec; Base2, starting with 1Gb/sec (1GFC) and simply doubling: 2GFC; 4GFC; 8GFC; etc.
We have 2GFC installed today with older 1GFC links in place as well. The beauty of 2GFC is that devices are backwards compatible with 1GFC; you can plug a 1GFC link into a 2GFC port and it will work, at 1GFC speed of course. You can't plug a 1GFC line into a 10GFC port. It won't work. Nor will a 2GFC line, nor a 4GFC line. In fact, no Base2 Fibre Channel link will work with a Base10 port.
Each Base2 FC device after 2GFC will maintain backwards compatibility for two previous generations. Thus either a 1GFC line or a 2GFC line can be plugged into a 4GFC port and it will work at the line speed. This lack of backwards compatibility is killing the general use of 10GFC ports to connect HBAs to switches and switches to target devices; that and the lack of any general need for 10GFC bandwidth. Jones says: "The Base10 migration path is dead."
Ten gig is okay for ISLs because, Jones asserts: "10 gig ISLs don't have to be backwards-compatible with anything. Copper is cheap. Because we're talking to ourselves we don't need inter-operability with everything." Expect suppliers to provide 10 gig copper ISLs between their directors and switches.
But there is a perceived need for 4GFC bandwidth and vendors are ramping up their offerings very quickly. Jones again: "Everything (enterprise) is going 4 gig. SMBs will stop at 2 gig, and be charged less." The irony is that customers don't actually need it yet: "We expect rapid deployment but not everyone needs it. Suppliers have it as a competitive checkmark. It's a vendor push. The speed isn't actually needed." The price premium for 4 gig is negligible so customers may as well have it instead of 2 gig, and then have the bandwidth headroom for when their SAN traffic really needs 4 gig.
The Base2 Fibre Channel roadmap looks nice, regular and logical but there needs to be a storage networking health warning about the expected dates. Transitions from a speed to the next level could be technically hard to achieve and delays might happen. Jones has a comment here: "Eight to 16 will be tough. Fancy stuff is needed. Maybe dual encoders for backwards compatibility."
The roadmap is now largely agreed by FC suppliers. It wasn't all plain sailing according to Jones: "Both the FC and iSCSI communities have learned their lesson on 10 gig. There is massive industry support for 8 gig. Cisco and Brocade fought against it. Now they are on board. It was a unanimous vote for 8 gig."
1GFC has a throughput of 200MB/sec.
2GFC has a throughput of 400MB/sec.
4GFC has a thoughput of 800MB/sec. Its T11 spec was technically completed in 2003 and product is available now, 2005.
8GFC has a throughput of 1600MB/sec. Its spec should be completed in 2006 and product available from 2008.
16GFC has a throughput of 3.2GB/sec. Its spec should be completed in 2009 and product available from 2011.
32GFC has a throughput of 6.4GB/sec. Its spec should be completed in 2012. Market demand will drive availability.
64GFC has a throughput of 12.8GB/sec. Its spec should be complete in 2016. Market demand will drive availability.
128GFC has a throughput of 25.6GB/sec. Its spec shpould be complete in 2020. Market demand will drive availability.
All the future dates are estimated.
This is quite a roadmap. If the industry can stick to it then SAN planners will have a good framework to use in their planning - a first for the SAN industry.