I'm evaluating dotConnect for Oracle as a replacement for our current ODP.net solution. I created a simple test scenario with two queries repeated 10,000 times. The first was a simple 'select *' from a small table the second a select from a larger table of 16,000 records using a parameter to return appx. 10 rows via a DataReader.
In OCI mode the dotConnect version ran a hair slower with the first query and a good bit faster in the second over ODP.net. What concerned me was when in Direct Connect mode the performance took a significant hit. The second query was roughly 30% slower than ODP.net. The second test set was also running in 64bit mode so I was surprised with the 32bit OCI still beating it out.
I'm wondering if anybody has had similar experiences with Direct Connect. Is it understood to be slower than OCI/ODP.net or are there tricks / tweaks to get at least comparable performance?
Any thoughts or insight would be appreciated.
Thanks!
Sieg Nagel
Direct Connect Performance Issues
1. dotConnect for Oracle (OCI) vs ODP.NET.
The performance in this case should be approximately equal. The difference may be caused by using insufficient data types, etc. Please send us your test project with the DDL/DML script and the exact results you have got. We will try to reproduce the issue in our environment. Also tell us your current versions of your dotConnect for Oracle (x.xx.xx, via the Tools | Oracle | About menu of VS), Oracle server (xx.x.x.x.x), Oracle client (xx.x.x.x) and ODP.NET (x.xxx.x.x). Did you perform your tests locally or via local network?
2. dotConnect for Oracle (Direct) x64 vs ODP.NET x64.
You have received expected results. The direct mode of our provider uses the features of the 7th and 8th OCI protocols. The latest versions of ODP.NET take advantage of later versions of the OCI protocol. The main advantage of the Direct mode is that dotConnect for Oracle talks to Oracle server in person, without calling any third party libraries. It allows your application to work with Oracle directly through TCP/IP protocol without involving Oracle client software.
3. dotConnect for Oracle (Direct) x64 vs ODP.NET x32.
The results should be equal to the previous comparison. Probably, you will get some performance gain with the x64 mode only in case of BLOB and string datatypes.
The performance in this case should be approximately equal. The difference may be caused by using insufficient data types, etc. Please send us your test project with the DDL/DML script and the exact results you have got. We will try to reproduce the issue in our environment. Also tell us your current versions of your dotConnect for Oracle (x.xx.xx, via the Tools | Oracle | About menu of VS), Oracle server (xx.x.x.x.x), Oracle client (xx.x.x.x) and ODP.NET (x.xxx.x.x). Did you perform your tests locally or via local network?
2. dotConnect for Oracle (Direct) x64 vs ODP.NET x64.
You have received expected results. The direct mode of our provider uses the features of the 7th and 8th OCI protocols. The latest versions of ODP.NET take advantage of later versions of the OCI protocol. The main advantage of the Direct mode is that dotConnect for Oracle talks to Oracle server in person, without calling any third party libraries. It allows your application to work with Oracle directly through TCP/IP protocol without involving Oracle client software.
3. dotConnect for Oracle (Direct) x64 vs ODP.NET x32.
The results should be equal to the previous comparison. Probably, you will get some performance gain with the x64 mode only in case of BLOB and string datatypes.