What I mean here by delays, is that while there is a delay at that interface, it is miniscule. Look it up.
Sattellite delay on the other hand is huge. You measure it in fractions of a second.
I find it humorous, also, that people go on about the delays involved with detecting a signal at the EO interface, but totally miss the huge delays inherent in the packet protocols that are being run over them.
Real basic network design. Any time you have a packet net, you have a series of store and forward "buffers" at all nodes in the network. What these nodes are is really just a computer that receives a packet in, and sends a packet on to the next location. Now, what does that mean? That means that the absolute minimum store and forward delay is the amount of time it takes to receive a complete packet of size N.
What does this mean for the performance, as opposed to the bandwidth of the "circuit"? Let's make it really simple. If you have a link A-B-C, where B is the store and forward node, the performance of that link from A->C is the same as a non-packet (i.e. circuit switched) link from A->C with a bit rate of precisely 1/2 that of the A-B-C link circuits.
(In this example, all internode links have the same bandwidth.)
That is the best case. 1/2. But the throughput is the same.
The reality is that there are further delays at each node. (i.e in in any packet net, like Cisco's.) Now, expand on this to a large net where a message will traverse many switched circuits between store and forward packet nodes.
I have again and again explained this. Most recently to an operations department who were mystified as to why, when they upgraded part of a long distance link in a packet net they only say a small improvement in performance.
Compared to all of that? Those delays at the EO interface are nothing at all. The Ciena execs are quite correct. |
