Device count alone is not sufficient to size file systems. You need to take into account images that will be stored, number of config revisions, syslog message collected, job data retained, log size, etc.
Figure on a solution server, NOT including each of the things above, each device will cost about 2 MB. The more syslogs, configs, and change records you collect, the more you will need to increase /opt/CSCOpx. The more software images and job data you retain, the more you need to increase /var/adm/CSCOpx.
You can assume an average syslog message is 100 bytes, and the average image size is 20 MB. Job data depends on the number of devices per job, and the type of job.
As for tftpboot, this only needs to be large enough to hold the largest SWIM job.
If possible, use logical volumes which can easily be resized as needed. You absolutely do not want to run out of space under /opt/CSCOpx as that will corrupt databases.
Just as an example, I am managing 100 devices in my lab. I have a SunFire V210 running Solaris 9 running LMS 2.6. The machine shipped with two 72 GB drives. I dedicated one drive entirely to /opt and swap (64 GB for /opt and 3 GB for swap). My / is 2 GB, my /usr is 27 GB, and my /export/home is 27 GB. The rest I allocated to /var (11 GB) and swap.
/opt/CSCOpx takes 4.5 GB, /opt/CSCOcwbS takes 253 MB, CSCOipm is 108 MB, and CSCOipmClient is 44 MB. /var/adm/CSCOpx is currently taking 3.3 GB with 2.5 GB of that being logs (since it is a TAC lab).
