| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 4878.1 | | NOVA::SMITHI | Don't understate or underestimate Rdb! | Mon Jan 06 1997 10:30 | 11 |
| 4878.2 | | NOVA::MCGEE | Oracle Rdb Mission Critical Engineering | Mon Jan 06 1997 12:21 | 5 |
| 4878.3 | | NOVA::SMITHI | Don't understate or underestimate Rdb! | Mon Jan 06 1997 12:41 | 6 |
| 4878.4 | adjusting working-set doesn't help? | UKVMS3::SHISCOCK | stand and deliver | Mon Feb 10 1997 08:08 | 49 |
|
Well I'm not sure how to narrow that gap between the performance
for 4.2 and 7.0. I've increased working-set and shaved off 3k faults
but the elapsed cpu time ($sho proc/acc) remains around 9 seconds.
In this test case 4.2 is almost twice as fast.
I've tried BIND_BUFFERS, BIND_WORK_VM and query outlines but the
impact is negligible.
The stratergy is the same in both cases. V7.0 does slightly fewer
dio but always 50% more page faulting. I've incremented WS repeatedly
and compared the two runs and always seen a reduction in pagefaults but
that is the only impact made.
Anything else I can try or experiment with?
cheers,
Steve
Firstn Get Retrieval by index of relation DATE_FILTER
Index name DATE_FILTER_HASH_IDX [7:7]
Sort
Leaf#01 BgrOnly HOLIDAYS Card=110484
BgrNdx1 H5_IDX [5:5] Bool Fan=20
Get Retrieval by index of relation RT_GROUP
Index name RT_HASH_IDX [1:1] Direct lookup
Get Retrieval by index of relation AC_GROUP
Index name AC_HASH_IDX [1:1] Direct lookup
Process Quotas:
Account name:
CPU limit: Infinite Direct I/O limit: 100
Buffered I/O byte count quota: 49872 Buffered I/O limit: 100
Timer queue entry quota: 10 Open file quota: 1000
Paging file quota: 248696 Subprocess quota: 20
Default page fault cluster: 64 AST quota: 108
Enqueue quota: 3000 Shared file limit: 0
Max detached processes: 0 Max active jobs: 0
$ sho work
Working Set /Limit= 8000 /Quota= 20000 /Extent= 25000
Adjustment enabled Authorized Quota= 20000 Authorized Extent=
25000
|
| 4878.5 | | DUCATI::LASTOVICA | Is it possible to be totally partial? | Tue Feb 11 1997 10:39 | 12 |
| Firstn Get Retrieval by index of relation DATE_FILTER
Index name DATE_FILTER_HASH_IDX [7:7]
> Sort
Leaf#01 BgrOnly HOLIDAYS Card=110484
BgrNdx1 H5_IDX [5:5] Bool Fan=20
Get Retrieval by index of relation RT_GROUP
Index name RT_HASH_IDX [1:1] Direct lookup
Get Retrieval by index of relation AC_GROUP
Index name AC_HASH_IDX [1:1] Direct lookup
Since you are doing a sort, make sure that WSQUOTA and WSEXTENT
are close to each other.
|
| 4878.6 | didn't help | UKVMS3::SHISCOCK | stand and deliver | Tue Feb 11 1997 10:47 | 2 |
|
Thanks Norm. I tried with the same values and it made no difference.
|