Unanswered: Sybase ASE, Replicaiton Server and BATCH jobs
Please forgive me if this is not the appropriate place to ask a question that regards both Sybase, Replication and BATCH jobs.
Has anyone had any experience in running night BATCH jobs in a repliacted environment? For example, we are using the "Warm Standby" Replication Model for Sybase ASE and Sybase Replication Server. This model has only one Sybase Replication Server, which is hosted on the replica / warm standby. Our goal is to run night BATCH jobs, but were concerned if this would be too overwhelming for Replication Server and the Warm Standby.
What are your thoughts? Does replication have to be turned off while a BATCH job runs, then later re-sync the servers? Or, can we run the nightly BATCH job and have faith that the Replication Server can handle such a volume of data / transactions?
Thank you for your help. Look forward to your feedback and ideas.
This is certainly one of the areas where there needs to be a marriage between the Application and Replication.
Since you didnt give much details I will start off with a few basic questions and answers.
1. What kind of batch jobs? bulk inserts / updates / deletes
2. Suggest trying run these jobs on a test environment and start benchmarking the latency and see if its an acceptable one.
3. If its not an acceptable latency
A) Check where there is bottleneck. Replication Agent, DSI etc.
B) Fine tune Replication Server as much as possible.
eg., scan batch size, sqt max cache etc.
After fine-tuning whatever is possible on the Replication Side,
Application Level - Some examples
A) Mass deletes can be replaced by stored procedure replication
B) There could be a lot of temporary work tables within your databases, which you dont need to replicate, as its just used to populate data temporarily and gets nuked. Also, other tables which you do not actually need replicated.
In short, once you have a latency issue, you need to find the actual process which is performing the job and see what all you can do to improve the performance in the Rep Server or use workarounds on the Application Side.