Performance-Question
Date: 1995/09/29
Message-ID: <DFons9.L2_at_merlin.unterland.de>#1/1
Hello,
I have a problem with a Oracle-Database on a HP 9000/K200 UNIX-System.
We did a number of optimization at the Unix-Level (Striping the Disks, distributing the database-tablespaces over different groups of disks and so on) but the effect is very strange:
What the hell is going on? IMHO both cases require a full table-scan. Why is the second one so much slower than the first one? In both cases the Space occupied on disks is nearly the same.
I checked the wait-statistics at OS-Level for those processes and discovered, that in the second case, the process spends most time waiting for a semaphore, but I dont know which one.
OK and now the technical stuff:
The Unix-Box is a HP 9000/K200 with HP-UX 10.0. We have 15 Disks, 2 Gigabytes each and 12 of them are organized in 4 volume-groups. Every volume-group is dividen into 3 logical volumes, every volume ist striped over all three disks. Combined together, we have 12 filesystems each filesystem with a size of approximately 2 gigabytes.
Oracle version 7.1.4 is used.
The tablespace consists of 4 data-files, eah with a size of 1.75 gigabyte. The primary extent is 1749 MB, other extents are 874 MB megabytes. PCTINCREASE is 0.
PCTFREE is 10 % and PCTUSED is 80 %.
The data is 7.500.000 rows with approx. 200 bytes of size for each row. The tablespace holds only the first extent.
If more informations are necessary, please e-mail.
Thank you very much
H. Becker
-- ---------------------------------------------------------------------- Harald (Harry) Becker harry_at_merlin.unterland.de 74080 Heilbronn Want to turn your Pentium-PC into a gameboy? type WINReceived on Fri Sep 29 1995 - 00:00:00 CET