agent bufferpool hit ratio

[DB2 News Group]

Using the new snapshot function in version 8 I have come across something
interesting.

When computing the bufferpool hit ratio and tablespace buffer pool hit
ratio, both of these show up as very low (< 2%) for a large table (MDC)
(400+ million rows) upon which full table scans are performed.

When running a query against the table which performs a full table scan (get
a count of all MDC cells by partition), I see that the bp hit ratio is very
high for the agent.

I can think of only 1 possible explanation for this. It would be that the
prefetchers (which are working quite well as async read requests are 95%)
are doing the I/O and that the agent servicing the query does not have to do
this I/O.

Can anyone validate/contradict or otherwise shed light on this. Any help
would be appreciated.

Thanks.
Locke