dsr@stl.stc.co.uk (D.S.Riches) (12/10/90)
We've recently hit a problem with running Ingres over a network. Our database sits on a Sun 4/280 and runs ingres/bin/lockmgr Using NFS we have a number of 3/60 workstations on a subnetwork who use this database. In the past few weeks we've been experiencing performance problems when accessing Ingres. Looking at the Sun4 we can see that when a connection is made to the backend the folloing statistics are applicable :- %vmstat -S 5 procs memory page faults cpu r b w avm fre si so pi po fr de sr x0 x1 x2 x3 in sy cs us sy id 3 2 0 0 10720 40 44 2 0 0 0 0 0 0 1 0 18 129 46 4 5 91 2 0 0 0 10520 0 0 0 0 0 0 0 0 0 0 0 661703 480 51 49 0 4 0 0 0 10456 0 0 0 0 0 0 0 7 4 0 0 482076 565 44 56 0 1 0 0 0 11456 0 0 0 0 0 0 0 2 1 0 0 581413 394 61 36 3 Generally the cs column has been in the 500s and the sy column in the 2000s. I don't believe we're doing anything out of the normal with Ingres requests (we're using ESQL). Has anyone seen this type of behaviour before and do you have any remedies? Dave Riches PSS: David.S.Riches@stl.stc.co.uk (or dsr@stl.stc.co.uk) ARPA: dsr%stl.stc.co.uk@earn-relay.ac.uk Smail: Systems Methods, (Dept. 607, T2 West), STC Technology Ltd., London Road, Harlow, Essex. CM17 9NA. England Phone: +44 (0)279-29531 x2496
mac@sequent.UUCP (Cormac Burke) (12/11/90)
In article <3844@stl.stc.co.uk> "D.S.Riches" <dsr@stl.stc.co.uk> writes: > >We've recently hit a problem with running Ingres over a network. >Our database sits on a Sun 4/280 and runs ingres/bin/lockmgr Have you considered running the lockdriver instead of the lockmgr? Don't know if this is the issue with context switching, but it should be easy enough to test. It's pretty extensively documented that the driver works much more efficiently than the lockmgr. Good luck! Cormac