Hi,
We've been looking at buying some new disks for our HP 9000 K410 Unix
server. The server is basically used for a data warehouse on IDS (7.23, soon
to be 7.30). As we need to buy a new IO bus to fit the disks (via fast/wide
scsi's) we've been testing out the current system. By:
1. Creating multiple copies of a 50 Mbyte table on various disks on our
current EMC raid box,
2. Running the same, simple, but long running query on these tables
we believe we can see that a HP-PB (HP Precision Bus - maximum throughput 32
Mbits /s) becomes the limiting factor on I/O throughput.
Two questions. Firstly does anyone know what is the throughput on an HP-HSC
IO Bus? Secondly, does anyone have any experience of handling the situation
where your disk arrays can provide data faster than the IO bus can handle?
As we use parallelism as much as we can, we seem to be getting to the stage
where database tuning is going to be pointless as the IO subsystem is
rapidly flooded by almost any large query.
Thanks for any comments
Martyn Hodgson
martyn.hodgson@eaglestar.co.uk