on 05-05-2008 9:25 PM
Hi All,
We are currently experiencing some major performance problems with one of our JDBC connections in production.
Our scenario is this, we are using the J2EE engine of our XI box as a gateway for BI UD Connect.
We are extracting from a simple Informix DB table some 600,000+ records which is a full extract, but this is taking like 2.5 hrs to do. This is the worst table but there are others that are really slow also.
If we extract directly using the same JDBC drivers but outside of XI it takes 1.5minutes.
Can anyone offer any assistance?
Are there any parameters we could change that would affect the performance?
Thanks in advance.
Craig
Use Stored Procedures
and check batchmode, i assume you are reading
all records into one messages.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Craig,
Could you give us more details about your implementation ? You say you do not use XI, I'm not sure to understand this statement
When you extract the same amount of records, do you use the exact same query ? When extract records thru XI JDBC adapters, extracted recordsets need to be parsed to an XML document, and depending on the complexity of a row (nr of fields, length etc), the SAX parser can take a long time and need some resources, things that do not happen when querying using any jdbc client, I guess ...
Chris
Christophe,
We are merely using the J2EE stack connecting it to the BW ABAP stack via RFC.
Here is an extract from the SAP document "How To - Using UDI to Connect to Non-SAP Database"
UD Connect utilizes the J2EE engine of the SAP Web Application server to access external data. This Howto
will deal with the set-up and configuration of the relational JDBC connector.
This guide will detail the steps necessary to connect the SAP BW ABAP stack with the J2EE stack via RFC
as well as configuring the JDBC connector to use the database vendor provided JDBC driver.
Hope this helps.
Basically XI's J2EE Engine is being used as a gateway to BW nothing more.
Thanks
Craig
Craig,
I'd first suspect some resource bottlenecks at the J2EE engine, but it will require some actions to prove that ... Have you tried increasing allocate resources for your JAVA stack (memory etc) ? I know this is quite generic (dummy) answer, but without additional hints, I do not know what could cause such situation
Is the "full" process (jdbc extraction and transfer back to the caller) taking such a long time ? Or only the select itself ?
Chris
Thanks to everyones input.
I think I have found the cause and it turns out to be BI_UDI component issuing a Group By clause at the DB for every field.
I have raised a customer message with SAP to try to get this resolved since it is fixed for BI_UDI for NW04s but no mention if it's fixed for BI_UDI 350 which is what we are using.
Thanks again.
Craig
User | Count |
---|---|
98 | |
11 | |
11 | |
10 | |
10 | |
8 | |
6 | |
5 | |
4 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.