darrell@sdcsvax.UUCP (02/14/87)
I'm looking for: 1) definitions of operating system overhead 2) estimations or measurements of overhead for real systems, preferablly mainframes or supercomputers 3) arguments for or against using overhead as a way to compare operating systems. Any leads would be appreciated. Marty
darrell@sdcsvax.UUCP (02/18/87)
Does anyone have any good ideas on how to quantify the performance of a fileserver? We have an Ethernet with 40 MG1 workstations using NFS to mount filestores from three servers. We need a way of measuring the quality of a server's performance, so that we can make an informed choice of what hardware to use when upgrading the servers. The problems appear to be: 1) Relativity Server load depends on what the clients ask for, so you need to know what all the clients are doing as well. 2) Interference We can't dedicate any of our hardware exclusively to fileserving so the "background load" needs to be taken into consideration. 3) Uh, What's the Question? We don't know what units our answer should be given in - (mean time until service) per client? This is "Benchmarkers Stone" stuff and so I don't expect complete answers, but does anyone have any suggestions at all? -- William Roberts ARPA: liam@cs.qmc.ac.uk (gw: cs.ucl.edu) Queen Mary College UUCP: liam@qmc-cs.UUCP LONDON, UK Tel: 01-980 4811 ext 3900