| Title: | FDDI - The Next Generation |
| Moderator: | NETCAD::STEFANI |
| Created: | Thu Apr 27 1989 |
| Last Modified: | Thu Jun 05 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 2259 |
| Total number of notes: | 8590 |
Hi, a customer of mine states that in order to obtain a throughput of one Mb/s on FDDI, you need 1 Mips of CPU. His assertion is based on the observation of his FDDI Network using TCP/IP on diverse systems (SUN, DEC, IBM ...) but is not the result of a thorough performance study. For him, it is a rule of thumb for determining the performance he can expect between two systems. I obtained between 20 and 21.5 Mb/s on a DS5000-200 (21.5 MIPS) with TTCP in UDP, which looks like what the customer states but of course this is not a proof. Can anyone validate/invalidate this "rule" of 1 Mips <-> 1 Mb/s ? By the way, the customer is CNET (French Telecom Research Lab) and seems to know a lot about TCP/IP Networking. Regards, Thierry
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 511.1 | MIPSBX::thomas | The Code Warrior | Fri Mar 20 1992 10:38 | 2 | |
Depends on the system. I've gotten 12Mb/s on a 3 MIP system. | |||||
| 511.2 | KONING::KONING | Paul Koning, NI1D | Fri Mar 20 1992 16:58 | 14 | |
Well, as you probably know, all generalizations are false. :-) For implementations of modest efficiency running on modern processors and operating systems, it's probably a pretty reasonable first guess. For older processors and systems, the results tend to be better (due to less system overhead and proportionally less of a processor performance hit due to interrupts). For well optimized implementations, and especially for implementations in bounded-function boxes, the performance is going to be MUCH better than predicted by your "rule". (Consider the LANbridge-100, which handles 10 Mb/s with a wimpy little 68000.) paul | |||||