sean@ukma.UUCP (Sean Casey) (11/23/85)
I've noticed that our Vax 11/750 with 5 megs running 4.2 BSD almost never uses more than half of the physical memory. Does anyone know of a simple way to retune the vm scheduler? I've considered raising the virtual paging point by fiddling with some of the constants, but I'm not sure of all the consequences. Is there anyone out there that can give me some help? Any ideas at all would be helpful. Thanks, Sean Casey -- ------------------------------------------------------------------------------- Sean Casey UUCP: sean@ukma.UUCP or 915 Patterson Office Tower {cbosgd,anlams,hasmed}!ukma!sean University of Kentucky ARPA: ukma!sean@ANL-MCS.ARPA Lexington, Ky. 40506-0027 BITNET: sean@UKMA.BITNET
friesen@psivax.UUCP (Stanley Friesen) (11/28/85)
In article <2391@ukma.UUCP> sean@ukma.UUCP (Sean Casey) writes: >I've noticed that our Vax 11/750 with 5 megs running 4.2 BSD almost never >uses more than half of the physical memory. Does anyone know of a simple >way to retune the vm scheduler? I've considered raising the virtual paging >point by fiddling with some of the constants, but I'm not sure of all the >consequences. Is there anyone out there that can give me some help? Any >ideas at all would be helpful. > Actually it is simpler than you think. The scheduler in 4.x is controlled by three *variables*(not constants). If the variables are initialized to zero(the default) the boot procedure calculates standard values, if they are initialized to other values the boot procedure leaves them alone. This means even a binary site can tune the scheduler, by patching the values in vmunix. The variables are: lotsfree - high water mark where paging is turned off desfree - desired minimum: turn page daemon on minfree - serious minimum: page daemon to overdrive The default values are: lotsfree = 1/4 of available memory(excl. kernel) desfree = max(0x190, 1/8 available memory) minfree = max(0x80, 1/2 of desfree) (Where the number specified is the number of core allocation units, which is either pages or klusters, I don't have the time to figure out which) Decreasing these values will slow down paging and lead to greater memory utilization. I would not recomend reducing the values of desfree and minfree below the stated minimums. -- Sarima (Stanley Friesen) UUCP: {ttidca|ihnp4|sdcrdcf|quad1|nrcvax|bellcore|logico}!psivax!friesen ARPA: ttidca!psivax!friesen@rand-unix.arpa