Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Newbridge Networks -- Ignore unavailable to you. Want to Upgrade?

To: fumble who wrote (14887)	11/22/1999 8:01:00 AM
From: gbh	Read Replies (2) \| Respond to of 18016

fumble, without going into too much detail, here's a brief description of two standard chipsets I've looked at. IBM Prizma -- a native 6Gb switch-on-a-chip. It scales with with no latency degradation to 25Gb/s using four of the chips. To get to 50Gb/s uses 8 devices connected via short reach (12-24 inches) coax, inside the box. They advertise the chipset as "scalable", but although it is scalable in terms of port (uninteresting, as beyond 16 ports is ), and speed, the speed expansion increases latency through the switch beyond an exceptable level. The chipset, in reality, is a 25Gb/s chipset, with "marketing" scalability to 50 and 100Gb. This is why I have "reservations" about NN use of this chipset at 50Gb, unless IBM did something specific for NN. This could be possible given the delays in getting this product to market. A big issue at the time, was this chipset had no support for hardware failover, a critical feature for any carrier. Also, this chipset had no capability to scale to 2.4Gb/s ports, and native port speeds were only up to 800Mb/s, with scalability to 1.6Gb/s. MMC Networks - 5-20Gb/s Switch Fabric. A "cheaper", easier to design to, solution than IBM, although at the time I evaluated, the 20Gb wasn't available. A relatively straightforward shared memory based switch, useful for edge applications but not even close, for next generation core apps. Basically, there were no off the shelf products that could support core switches above 20-25Gb/s. And this isn't surprising, given the volume of these switches sold doesn't justify a semiconductor company's development of one. This is why I found it surprising that a company like NN would use something other than an internally developed fabric for its core switch. Its certainly conceivable that the year long delay in delivering a 50Gb class switch is because they chose to develop their own fabric. Or perhaps IBM developed something superior than the Prizma for NN. We'll find out soon enough, I hope. Gary