SAS User <sas@SDAC.HARVARD.EDU> replied: > We have two new servers and this crashed on both. We only have about > 60 people spread across about 10 machines, but here's the skinny on the > new, fancy ones: davinci and verdi: > > No user restrictions. > > Both are running Solaris 8. Davinci - 32 bit Verdi - 64 bit > > Davinci is a Sun Enterprise 450 > Verdi is a Sun Fire V880 > > Davinci > > > > 4 CPUs 480 MHz > > 4.1 Gigs of Memory > > 7.8 Gigs of Virtual Memory (Swap) > > > > Verdi > > > > 8 CPUs 750 MHz > > 32 Gigs of Memory > > 67 Gigs of Virtual Memory (Swap)

Both of these should be able to handle your PROC MIXED. But there are two possibilities.

[1] It is possible to mis-specify your PROC MIXED statements so you chew up unbelievable amounts of real estate. If you specify a model that will create p columns in the X matrix and g columns in the Z matrix, then you are going to need something like 40*(p*p + g*g) + 32*(p+g)**2 bytes to hold everything. Look at the "Model Information" table to see how many columns your user is creating in the X and Z matrices. Look at the model and see if it is properly specified.

[2] As I pointed out to someone else only this morning, sometimes the install of SAS on SlowLaris has the nasty problem of putting the SASWORK directories in the /tmp partition, which can cause major slowdowns and horrible choking clogs. Check this on both machines, and either point SASWORK elsewhere or get your sysadmin to fix the install as soon as feasible.

HTH, David -- David Cassell, CSC Cassell.David@epa.gov Senior computing specialist mathematical statistician