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 : Juniper Networks - JNPR -- Ignore unavailable to you. Want to Upgrade?

To: Teflon who wrote (1817)	10/20/2000 7:29:26 PM
From: The Phoenix	Respond to of 3350

The Corvis switch is switching lambda traffic flows... the entire lambda at once. It is not processing the traffic within the flow and "re-routing" it to other lambdas. That is, all the traffic in a flow gets routed all at once...there is no-per-packet processing going on. The real benefit for a switch like the Corvis switch is simplified provisioning of new circuits (lambda's) or the re-routing of lambda's. This doesnt' mean that switching is taking place throughout the day...just that flows are set up from a central management system somewhere in the sp network without requiring technical support at each POP. So, the Corvis switch is an optical cross connect - not a packet switch or optical per-packet router. There is much confusion on this point. Don't worry - Gilder doesn't get this either. OG

To: Teflon who wrote (1817)	10/20/2000 8:01:36 PM
From: dwayanu	Read Replies (1) \| Respond to of 3350

Hi Teflon: What am I missing? The difference between 'switching' optical channels (lambda's in Gilder-speak, wire-pairs in ancient AT&T-speak) and 'routing' individual packets of data based on eg the IP address header in each packet. Huber was not talking about packet routing. Terminology gets sloppy in casual conversations. An optical 'channel' is a single given wavelength of light, one of many stuffed into a given fiber by eg DWDM. As the total many-channel optical signal passes along a long length of fiber, it gets fuzzy and occasionally (like every 100 miles) has to be 'regenerated'. Old style regeneration consisted of optically splitting the signal into each of its channels, converting each channel's data into electronic data including cleanup, reconverting that clean data back into a lambda (single wavelength) signal, and recombining all those lambda's back into a single signal to be fed back into the fiber for the next 100 mile leg. Gets really messy when a modern fiber carries several hundred lambda's at the same time. As I understand, Corvis handles this regeneration without converting the optical signal to electronic data, using in-line regeneration amplifiers to clean up and strengthen the signal. Thus Corvis can send a signal a few thousand miles without having electronics in between. 'Switching' in this Corvis context refers to redirecting a given lambda from among the many arriving on a fiber. Used to add lambdas to and drop lambdas from a given fiber's signal. Nowadays optical switching can be done using eg Bragg gratings, AWG filters, or more recently MEMS (microscopic mirrors) or ink jet bubbles (no joke <g>). Again, like the regen, to avoid converting the optical signal to electronic just to separate out some particular channel of the signal. 'Routing', in total contrast, occurs out at the edges of the network, to receive a stream of data (like that contained in a single lambda) and then to direct individual packets of data from within that stream to their various target machines. This requires that the target address (eg a PC's IP address) of each packet be read and interpreted. Nobody has figured out yet how to do this at light speeds, because target address interpretation requires computer-style electronic logic. BTW, because routing at optical speeds is so difficult, that is the basis of the long running Gilder argument that the network will move from packet-switched traffic back to switched circuits (lambda's), because once a circuit is set up (half second or so), the traffic can be steered from any point A to any point B at light speeds without electronics. All you have to do is be able to stuff a few tens of thousands of lambdas into each fiber (and manage to untangle them at the other end). All the above 'switching' and 'routing' is quite different from the local network 'switches' being sold by eg Juniper and Cisco and Extreme Networks, which route IP packets from machine to machine <VBG>. Hope that helped (and hope I got it at least mostly right <g>) - Dway