Hi,

We're running Sybase 11.9.2 and recently increased our User Log Cache
(ULC) to 16384 from 2048. Sysmon was showing that well over 20% of ULC
flushes were due to a full ULC (sometimes over 50%). Since the
increase this figure has reduced to less than 1% and the new cache
size is usually fully utilised (on a 1 minute sample)

However, we have also recently experienced what we think are disk i/o
problems. We have occasional high disk activity (high Average service
time (residency)) and "Avg Time per Normal Chkpt" sometimes in excess
of 5 seconds. This can happen for a period of 2 hours roughly twice
per day and causes our apps to run too slowly. There is no known
interference from other processes.

The network guys say that the disks (SAN) are not to blame - and that
we should consider reducing the ULC back to 2048. I'd like a second
opinion before trying this.

Thanks in advance..
Chad.

Server Version: Adaptive Server Enterprise/11.9.2.3/1159/P/SWR 9218
ESD

Hi,

In order to reduce the time of a checkpoint I would consider
decreasing the "recovery interval", which will lead to more and
shorter checkpoints. In addition I would suggest to increase the
"housekeeper free write percent" (for example from the default of 1 to
2), in order to distribute I/O more evenly.

Regards,
Ulrike

chadjackson25@hotmail.com (Chad) wrote in message news:<7ea25765.0311240301.25215ae9@posting.google. com>...
> Hi,
>
> We're running Sybase 11.9.2 and recently increased our User Log Cache
> (ULC) to 16384 from 2048. Sysmon was showing that well over 20% of ULC
> flushes were due to a full ULC (sometimes over 50%). Since the
> increase this figure has reduced to less than 1% and the new cache
> size is usually fully utilised (on a 1 minute sample)
>
> However, we have also recently experienced what we think are disk i/o
> problems. We have occasional high disk activity (high Average service
> time (residency)) and "Avg Time per Normal Chkpt" sometimes in excess
> of 5 seconds. This can happen for a period of 2 hours roughly twice
> per day and causes our apps to run too slowly. There is no known
> interference from other processes.
>
> The network guys say that the disks (SAN) are not to blame - and that
> we should consider reducing the ULC back to 2048. I'd like a second
> opinion before trying this.
>
> Thanks in advance..
> Chad.
>
> Server Version: Adaptive Server Enterprise/11.9.2.3/1159/P/SWR 9218
> ESD