| Weird client performance problem [message #284910] |
Sun, 02 December 2007 18:18  |
dhock194
Messages: 7
Registered: December 2007
|
Junior Member |
|
|
I have a strange performance issue, and was wondering if anyone else has seen something similar......
Environment 1 (Linux app -> Instantclient -> Single DB install)
- Oracle 10gR2 single Oracle server (not RAC), single database with 12 tables
- Custom application running on app server (Linux 64bit) using Oracle instant client
Environment 2 (Windows app -> Std client -> RAC)
- Oracle 10gR2 single Oracle server (not RAC), single database with 12 tables
- Same Custom application running on app server (but on Windows 32bit) using full Oracle client install (not instant client)
Environment 1 (Linux app -> Instantclient -> RAC)
- Oracle 10gR2 Linux (RHEL4) 2 node RAC cluster, single database with 12 tables
- Custom application running on app server (Linux 64bit) using Oracle instant client
We have been performing performance testing on each of these scenarios, and the results are somewhat confusing. Testing is done by generating inbound application transactions to the apps server, which in turn generates all required Oracle calls. Most of the calls are done as stored procedures, and most of the code uses OCI directly.
For Environment 1 above (Linux app -> Instantclient -> Single DB install), we get good performance. Same for Environment 2 (Windows app -> Std client -> RAC), with good performance. However, when running Environment 3 (Linux app -> Instantclient -> RAC) we get 40% less performance of the other two.
The Windows and Linux apps are nearly the same, except for minimal OS considerations, so they should about run the same. Since both apps can achieve the higher performance, as can RAC, the only variable that would seem to make sense is that possibly InstantClient does not do well with RAC – does this make any sense, and has anyone seen anything like this?
|
|
|
|
| Re: Weird client performance problem [message #284914 is a reply to message #284910] |
Sun, 02 December 2007 19:59  |
rleishman
Messages: 3728
Registered: October 2005
Location: Melbourne, Australia
|
Senior Member |
|
|
For a 40% difference, isn't it much more likely that something is happening at the server? Have you tried tracing the jobs or running Statspack and comparing for each environment?
Ross Leishman
|
|
|
|
] 
Blog
Search
Help
Members
Register
Login
Home
