Odd issue here. One of our apps reported slightly increased latency on a
heavily used database which started Monday evening and has been consistent
since. This is not large from a database perspective but the increase of
average response time from 1 to 3 microseconds has had a noticeable impact on
their performance.
No execution plan changes, a slightly heavier load at peak times (up about 10%
from last week) but nothing that I would expect to have such an impact. We
still see the increased latency when the server is 90% idle and the load
average is 5 (32 cpus, 16 cores).
In doing an AWR report comparison for comparable times one major difference I
saw was that 2 frequently run queries were suddenly doing a lot of physical
i/o. For a comparable 2 hour period they went from 1.5 million to 1.8 million
executions but physical reads increased from 0 to 1.2 million. I sampled a few
other random times and this was consistent. The queries are both doing index
access. One is an index range scan and the other a unique scan against the
primary key.
I checked with the app group and they have no explanation for why the app might
suddenly be querying blocks that aren't in cache whereas they weren't last week.
I am reluctantly considering adding the tables to the keep cache even though
they are huge (4-5G@). They are frequently accessed so I don't see any real
downside to this other than the huge amount of memory I would be reserving for
them.
We are going to test this in a non-production environment tomorrow but my two
questions are:
1. Is anyone aware of any problems this might cause? We have not used the
keep cache in the past.
2. Is there a way to preload the data once the db_keep_cache_size has
been set, the sga increased to accommodate it, and the table assigned to the
keep cache? A full table scan would usually bypass the buffer cache so I'm not
sure of the best approach here.
3. Any other ideas on what might cause this?
TIA!
Jay Miller
Sr. Oracle DBA
