I am interested in hearing from anyone who has tried increasing the I/O
throughput for AFS. With our backup software we have found a big
discrepancy with the I/O throughput between AFS and UNIX file systems.
I am looking for any suggestions on tuning AFS so the I/O throughput
starts approaching that of standard unix. I understand
there is a reasonable amount of overhead associated with AFS but the
performance we are experiencing is abysmal at times (6 KB / sec).
We have seen some small improvements in throughput by increasing the chunksize
from 64KB to 1MB on the eight servers running AFS 3.3A and Solaris 2.3. We
also increased the disk cache from 30MB to 125MB. We saw another improvement
when we switched from 125MB disk cache to a 50MB mem cache. None of these
changes have so far produced any significant increase.
Dale Wolf
AG Communication Systems
Phoenix AZ
[EMAIL PROTECTED]
