Query takes twice as long for identical database on two seperate systems.

Question

Hi, I am looking for some help with a performance issue on a particular database.
I am executing the query
'for $o in collection("rx.dbxml")/rx[dbxml:metadata("janusId")="12345"] order by $o/@dos descending return <item id="{ replace(dbxml:metadata("dbxml:name", $o),"&#34;", "_") }"> {$o} </item>'
on two different systems through a python xmlrpclib ServerProxy object. This is the my test query but the performance of other queries exhibits the same issue.
The data on both systems is identical and the time required to execute this query is the same for both systems across
every database, except for one. For this particular database, the time required to execute the query on System B is double
that of System A.
This is only the case for one particular database which makes me think that it is a data issue?
However the data is identical on both systems.
Is anyone aware of any data related issues that would cause such a performance hit based on
the different setups listed below? What should be my next step in diagnosing this problem?
System A
- 1 x Quad core Intel(R) Xeon(R) CPU E5450 @ 3.00GHz
- 32 GB RAM
- Red Hat Enterprise Linux Server 5.1 (Tikanga)
- Berkeley dbxml 2.4.13 patched to 2.4.16
- bsddb3 4.5.0
- Python 2.4.4
- Twisted 8.2.0
- SOAPpy 0.12.0
- Filesystem Size Used Avail Use% Mounted on
/dev/sda5 39G 33G 4.3G 89% /
System B
- 2 x Dual core Intel(R) Xeon(TM) CPU @ 2.80GHz
- 12 GB RAM
- Red Hat Enterprise Linux Server 5.3 (Tikanga)
- Berkeley dbxml 2.4.16
- bsddb3 4.7.5
- Python 2.4.4
- Twisted 8.2.0
- SOAPpy 0.12.0
- Filesystem Size Used Avail Use% Mounted on
/dev/mapper/VolGroup00-LogVol00 384G 88G 277G 25% /
Thanks for any help.

Answer 1

Try looking at db_stat output for each system for the query itself.
1. clear stats:
db_stat -Z -m
2. run the query
3. look at stats:
db_stat -m
Note any suspicious differences. You can/should also monitor disk and IO during the queries. If the plans are truly the same and the data is the same that points at system differences which would mostly be IO. At least that's the first place to look.
Regards,
George