paul@phoenix.Princeton.EDU (Paul Lansky) (09/18/88)
I've done some more investigation of disk throughput under ultrix 2.0 and higher on MicroVaxII and am beginning to wonder if the problem is not in the Generic File System overhead. I earlier was led to understand that there were some additional timeouts in the 2.0 kernel which provoked an allergic reaction in the Emulex QD32 controller. But I saw some tests on RA81s under Ultrix 3.0 which were significantly poorer than what I get to fujis (with Emulex QD32) under 1.1. I first assumed that this could not be GFS's fault since I have a data-acquisition driver which calls the disk strategy routine directly with a block list, and assumed that this was circumventing GFS overhead. But perhaps I'm wrong. What I fear is that DEC has decided that future CPUs and busses will catch up with GFS overhead (throughput on our 8700 is fine) and that MicrovaxIIs will have to suffer for the time being. I'd appreciate any views, experiences, or opinions in this area, as well as experiences with 4.3 on Microvaxen. Thanks Paul Lansky paul@phoenix.princeton.edu Music Department Princeton University