Suggestions for planned bonnie++ benchmark

Dear all,

I've posted some time ago and reported problems with latency on our SAN.

We've come to the conclusion that we've to run a benchmark, we've
suggested bonnie++ and our Datacenter has welcomed that idea.

We mainly host DB2 Databases on the SAN and have an application which
writes (and reads and then deletes) a lot of tiny to small files.

We'd like to run several benchmarks in parallel to identify the
bottelnecks (and to bypass the filesstem cache).

HW Specs:
SAN Box (HDS Thunder 9570V, 2GB, 60x146GB HDD shared with other apps)
Server (SUN E12k, 10CPU, 48GB, Dual FC, UFS)

Planned benchmark switches:
# bonnie++ -b -x 5 -s 8192:32k -n 0 -r 4096
Eight bonnies running in parallel on the data/idx/tmp/log partitions
of two databases (4x2 DBs, 32k is the main prefetch size of our DBs)

plus

# bonnie++ -x 5 -s 10240:32k -n 300:48000:77 -r 5120
for the partition hosting the flatfile based application

....making it a grand total of 9 bonnies (which should hopefully resemble
our environment pretty well)

What would you change in the benchmark setup?
Any comments are appreciated.

Cheers,

Jochen

P.S.
I ran the two benchmarks in seqential order on a standard PC, the first
took about 2.5h, the latter about 20h.
As expected the block throughput was about ~35-40MB/s, about half that
for Chr based transactions. The file creation benchmark really showed
the limits of my PC at approx. 330/320 writes, 70/60 reads and 400/150
deletes per sequential/random action.
(PC, 1.8GHz, 512MB, 160GB ATA, Lin2.6, Reiser)