Neil, thanks for the article.
<<<For example, Gigabit Ethernet, despite its appropriateness as an IP networking technology, is a poor channel technology. Gigabit Ethernet's small 1.5KB packet size induces a large processor load because the processor has to service so many more interrupts.>>
<<<Flow control is another case in point; using packet loss as a trigger for flow control is a waste of resources.
Fibre Channel does not lose frames, due to its built-in, standard, flow-control mechanism.>>>
<<< In addition, there is the issue of SCSI support. Fibre Channel has it; Gigabit Ethernet doesn't. In fact, ask any RAID, disk drive or tape vendor about its plans for Gigabit Ethernet--they don't exist. Without SCSI support, Gigabit Ethernet cannot circumvent SCSI's scalability and performance limits. Meanwhile, vendors and server companies have announced SCSI-over-Fibre Channel support.>>>
This shows how shallow PC-Week is. The writer didn't bother to ask anyone for these pertinent questions. I didn't expect him to know the answers to all these, but he should have asked around instead of putting his foot in his mouth.
I am glad Dr. Thomas Raeuchle, VP of Engineering at Ancor, took issue with him and clear things up.
KJ |
