Why Optical Switches Can't Fail
It comes down to money – or more precisely, the value of the traffic passing through tomorrow’s optical switches, which are likely to handle at least 1,024 wavelengths each capable of carrying 10 Gbit/s of traffic.
If such a switch was running at full capacity, carrying telephony traffic worth 1 cent a minute, then the carrier would be clocking up a cool $26,667 per second, according to Denny K. Miu, IMMI’s president and CEO.
Of course, the switch is unlikely to run completely full, and the days of getting 1 cent a minute for telephony traffic are probably numbered. All the same, the message is clear: carriers stand to lose an absolute fortune if one of tomorrow's optical switches actually stopped working.
This huge potential loss has dictated the design of the MEMS (micro-electro-mechnical systems) used in IMMI’s switch, according to Miu (see Switch Startup Raises MEMS Questions).
Miu points out that reliability is far less of an issue with other MEMS applications. A large screen based on a similar number of mirrors would generate revenues of about $2,500 an hour if an audience of 500 paid $10 each to watch a two hour movie, he says. That’s close to 40,000 times less dollars per second than a telecom switch.
Miu also draws attention to some hidden costs that carriers will face when they get around to installing 1000x1000 port optical switches in the future.
Reliability will be so important that they’ll have to install two switches in each location – an operational one and a backup. That means they’ll need 1,000 tiny 1x2 switches on at least one side, and possibly both sides of this setup, to send traffic to both switches and collect it again on the other side.
That’s probably one reason why IMMI’s first products will be arrays of multiple 2x1 switches on a single piece of silicon.
By Peter Heywood, International Editor, Light Reading, http://www.lightreading.com