In my May 5, 2008, story package on Cisco's 4900M 10 Gigabit Ethernet switch, I listed some of the drivers that I thought would make this high-bandwidth, direct-connect networking necessary. Among the things I listed were application consolidation, large format media files and compute-intensive applications.
May 14, Network Instruments will release survey results that looks at survey results for 10GbE network adoption, talking about switch-to-switch (as compared to the switch-to-server focus of my review and story package). The survey basically shows that 10GbE networks are being implemented around the world and that one quarter of respondents said they will have some 10GbE networking in place by the end of 2008.
For IT managers, the other factor driving 10GbE implementation is the price per port. 10GbE port prices are driving downward. This week I'm set to spend a day with the engineers at Arastra in preparation for a review of the Arastra 7148S, a one-rack-unit, L2/L3/L4 48-port 10GbE SFP switch with 800G bps of bandwidth. Arastra officials have been consistently cagey in telling me how much the unit costs. However, they say that the price per port is going to be around $400. That is significantly lower than the price I computed for the Cisco 4900M. There are significant differences in the switches, but so that we don't miss the forest for the trees, the price per port is coming down. It likely will not be much longer before the cost for 10GbE is only two or three times the cost of 1Gb.
There are all sorts of ramifications for the advent of 10GbE networking. Fibre Channel is under heavy pressure. I continue to wonder about heat dissipation for servers operating at such a rate that they need 10GbE connectivity. And there is a question about where the bottleneck moves to when a 10GbE pipe removes bandwidth constraints ... at least for the present.