A very intriguing news story broke the other day that may have escaped readers’ attention amid all the hullabaloo surrounding the HMRC’s missing discs: Google is making its own network switches using optical fibre connections.
Nyquist Capital analyst Andrew Schmitt was the first to reveal this latest twist in the ongoing evolution of the world’s number one search firm. Through talking to a number of sources in the optical networking industry, Schmitt discovered that Google is developing switches based on comms chipset vendor Broadcom’s 10GbE silicon. The switches apparently use the small form-factor pluggable plus (SFP+) optical transceiver module. Schmitt wrote up the story on his blog under the heading “Google’s secret 10GbE switch”.
So it would seem that Google, which already builds its own servers, has seen another opportunity to save some money by doing a spot of DIY. Quocirca analyst Rob Bamforth believes this is another case of an IT-savvy business taking a cold, hard look at all the snazzy products being pushed its way and realising that it is being asked to pay for a lot of stuff it doesn’t need. What Google is doing, according to Bamforth, is following the Kiss concept: keep it simple, stupid.
Bamforth said this kind of thing happens all the time in the IT industry: an
organisation suddenly kicks through a blockage and triggers off a stampede.
So has Google shown the way to the promised land of cheap 10GbE interconnects?
Well, in his blog entry Schmitt pointed out that “this non-standard and very low
cost optical format should prove just as attractive to other datacentre
customers”, and it is clear that it is datacentre customers who will drive the
10GbE optical interconnect market.
Hang on a minute, I hear you say, aren’t we forgetting about the IEEE 10GBase-T standard using copper cabling for the network interconnects? Doesn’t copper still have a future? Well, the trouble with 10GBase-T is that its power-per-port requirements are simply too hefty for what is an increasingly energy-conscious enterprise market. Currently, it’s 10W per port, although 5W systems are due next year.
This is one of the reasons why some industry experts are now speculating that glass, not copper, will soon be carrying the majority of network traffic around the datacentre. Of course, the demise of copper has been predicted before, but as demand for faster network throughput grows, with 40Gbit/s and 100Gbit/s standards being planned, can copper really be made to perform at these speeds, and over the standard 100m reach?
It seems that the sands of time may be about to bury copper.
