--- In ib-support@y..., "csswa" <csswa@y...> wrote:

> --- In ib-support@y..., "alex_vnru" <ded@h...> wrote:
> That you have a db 20 times bigger
> and far more garbage to be collected than my example -- and it takes
> just 2 minutes ??!

Andrew, I for a rather long time did'nt pay attention for this
issues, but IIRC garbage collection is performed when last connection
is finished too and when last of my user disconnects his gds_inet
process on server live about 5-7 minutes after client application is
closed. Add this time to sweep time.

> In your opinion, do you feel that the huge difference in
> sweep performance between these two examples can be attributed to

the

> linux vs NT and RAID/2Xeon vs my less powerful system (Duron 850,

512

> ram, Quantum Fireball 20 gig, fb1 superserver)?

Just sweeped the same db created from gbk on PII 450 SCSI 2 QF 10Gb
1Gb RAM (last the same as on production server) and for 3 days used by
2 developers in regular debugging of application (it will grow for
some more years). 4 minutes. There shouldn't be as much garbage as on
real one, though. So I think you should search reason for this
difference in FB CS/SS difference, transactions gap or M$ features,
file cache for example. Maybe more experienced in this subject person
can give more light, I never used neither FB SS nor M$ as server. And
I have a hazy recollection that when this PII was production server
for IB4 CS on SCO Unix and db was about 1-1.5 Gb and was'nt sweeped
for monthes :) sweep duration was about 30-40 min.

Best regards, Alexander V.Nevsky.