[Prev] [Next] [Top]

High Performance TCP/IP communications with HPs and IBMs connected around an Ancor Fibre Channel Fabric

I. Introduction


This note is aimed at providing communication times for transferring data from memory to memory, between IBM RS/6000 and HP/9000 workstations, via a 16 port Ancor Fibre Channel Fabric, using TCP/IP protocols, when running High Performance TCP/IP on machine's kernels. High Performance TCP/IP is nothing else than the common name for RFC 1323. Today, most workstations (IBM, HP, DEC, SGI at least) support this potentiality, but to our knowledge, none implements it by default. This is the user's responsibility to switch it on (see for instance appendix for switching RFC 1323 on under AIX Operating System).

Figure 1 shows the experimental hardware configuration used (all the Fibre Channel products are currently running Fibre Channel version 3.9).



Figure 1: Experimental hardware configuration.

The two HP/9000 machines are 700 series (735) with a PA7100 processor running at 99 MHz. The IBM RS/6000 workstation pool is composed of a 590 model with a Power2 processor running at 66 MHz and a 250 model with a PowerPC 601 at 66 MHz. Two kinds of adapters, depending on the machine in which they are intended to be located, can be used to provide access to the Fibre Channel world (see [3], [4], [5], [6] for more detail):

Some communication times have been previously reported in papers [1] and [2]: they were performed strictly with the same hardware, but the RFC 1323 was not, at that time, turned on. The comparison of the communication times presented in this note and in the previous notes ([1] and [2]) will allow to quantify the improvement brought by running High Performance TCP/IP on workstations, for TCP/IP communications.


RD11 - 23 NOV 94
[Prev] [Next] [Top]

Generated with WebMaker