40Gb and 100Gb Networks

While it has become a fixture in virtually every home and business, Ethernet will not dominate multi-gigabit networks to the extent it has dominated sub-gigabit networks, according to Freesky Research‘s soon-to-be-released report: 40 and 100 Gigabit Networks: Technologies, Markets, Applications.

As LAN speeds increased to a gigabit, the most cost efficient networks embraced Ethernet as their primary data link protocol. But as LAN, SAN, and MAN speeds surpass a gigabit, the most cost efficient networks are embracing multiple data link protocols. Freesky Research’s analysis has found these diverse topologies are not the result of sunk costs in Fibre Channel or Infiniband, but rather the ability to combine low cost, inter-processor links in I/O and storage with low cost, inter-switch links in LANs and data centers. Theoretical concerns about multi-protocol complexity are being shot down by practical concerns about capital budget simplicity.

With MACs and ASICs shipping in exceptionally high volumes, Ethernet port prices declined at more than twice the rate of ATM and FDDI competitors in the late 90s. However, as link layer technologies head up to 40 and 100 Gigabits per second, they are sharing many similar FPGA, SerDes, and encoding technologies, which is making it much more difficult for any one protocol to gain a cost advantage through volume production.

"The defining economic characteristic of sub-gigabit networks was framing, while the defining economic characteristic of multi-gigabit networks is clocking," said David Gross, author of the report. "Therefore, in 40 and 100 Gigabit networks, Ethernet will frequently interconnect with Fibre Channel, Infiniband, even SONET, and will not be able to kill off those protocols the way it decimated Token Ring, FDDI, and ATM."

The study will include forecasts for 40 and 100 Gigabit technologies, and will evaluate deployment trends in individual industries, including financial services, telecommunications, and higher education.