Long time for flush IDES

I notice, sometimes, spikes in transaction time on production database. Usual
transaction time is 30 ms, but sometimes its go up to 100-400 ms. When I look
into backup of log files (I use Log Explorer), to see whats happened during
that transactions, I see what most of time was spend for "flush IDES"
operation. The operation do some changes in sysindexes table - what it change
Log Explorer not show. I thinked it may be autoupdate of statistic. So I
turned off autoupdate statistics option, create statistic option, and torn
page detection, on test server. Next I called SPs to see whats writing to
log. No changes - there is still some spikes, but very small, caused by
"flush IDES". What it is and how to decrease the influence?
Possibly previous message is hard ti understand. Many of transactions with
long time of execution execute during time when was no other transactions, so
no locks.
"andsm" wrote:
[vbcol=seagreen]
> Check locks of what? What was locked? And how to decrease influence of this
> "flush IDES"?
> "Aleksandar Grbic" wrote: