The transition from 1 Gigabit per second (1-Gig or 1G) to 2-Gig Fibre Channel (FC) devices is nearing completion. In 2001, only about ten percent of the FC market consisted of 2-Gig FC devices. Last year, it was up to ninety percent. So, as with any aspect of computing technology, it was the perfect time to announce the next speed upgrade.
QLogic Corporation announced last April that it was going to start producing 4 Gigabit per second (4-Gig or 4G) Fibre Channel chips, host bus adapters (HBA), and fabric switches. In June the Fibre Channel Industry Association lent its support to this standard. Several manufacturers – including Hitachi Data Systems, Agilent, Seagate, and JNI – committed to developing 4-Gig products. In September, Infineon and PMC-Sierra released the first 4-Gig components (chips and transceivers).
It’s important to understand, though, that at this point the change is being dictated by the manufacturers, rather than in response to consumer demand.
“There was not pent up user demand for 4 Gbps FC, nor were the major storage OEMs asking for the boost in performance,” says Richard Villars, vice president, storage systems for Framingham, Mass.-based IDC. “The move to 4 Gbps FC is being driven by silicon development patterns.”
Because of advances in the manufacture of silicon and components, 4G FC will deliver twice the performance for about the same cost as 1G/2G. And, since it is fully backwards compatible with the lower speed components, the 4-Gig units can just be sandwiched in with the older ones, rather than having to perform a technology migration.
Page 2: Scenarios for 4-Gig FC Adoption
Continued from Page 1
Scenarios for 4-Gig FC Adoption
This gives three different scenarios for adoption of 4-Gig FC:
- In most cases, companies will just move existing operations from 2G to 4G as they add capacity or replace existing equipment. Within roughly the next two years, 4G will be more reliable and cheaper than 2G. It would be like buying a new PC. Even if you could get by with an 800 MHz CPU, it is likely that you will get one in the 3 GHz range because that is what is on the market currently. Since 4G FC will be no more expensive than 2G, there will be no reason to go with the slower systems, even if they are available.
- Tape backup systems are one part of a network that can benefit from the higher speed, as it allows them to perform faster backups. These will likely go to 4G, though it is not yet certain. StorageTek, for example, had not committed to 4G as of early January. Rather, it is considering going straight to the non-backward-compatible 10 Gbps (10G) Fibre Channel spec.
- Companies that have deployed highly customized applications that they can’t risk degrading while optimized drivers are developed will stay with 1G/2G FC, even though they may have to pay a premium for this older technology.
The other option is going with 10-Gig FC, but this will have limited use. In most cases there is no reason to upgrade to 10G, especially considering that it costs five times as much per port and is not backwards compatible. It will be useful, however, for interswitch links within a SAN. Certain high performance computing (HPC) sites that need extremely high data throughput could also see a benefit from upgrading to 10G.
But for most companies, there is no compelling need to upgrade, as the FC bit rate isn’t typically the bottleneck limiting the performance of their systems. In other words, it won’t improve service and performance levels if all it does is flood the server host bus faster than the processor can handle the data.
“For most companies, the decision to purchase FC products will have nothing to do with a need for more speed,” says Villars. “Rather it will be driven by cost and product availability.”
Those products will start becoming available toward the end of this year, with all the FC vendors shipping these products by the middle of 2005. After that, it will follow a similar growth curve as 2G FC, reaching ninety percent penetration by mid 2007. By that point, the data rate will probably have doubled once again, and we will see 8G FC, which would be backwards compatible with 1G, 2G, and 4G. But that decision will be made by the component manufacturers, not the SAN OEMs.
“The next generation is always going to be cheaper than the current one, even if it has more capabilities,” says Villars. “As the storage industry becomes more dependent upon third-party component manufacturers to build lower cost hardware, storage OEMs will have less and less control over the timing of new technology developments.”
Feature courtesy of EITPlanet.
Back to Enterprise Storage Forum