Vertica Performance troubleshooting

davds

Hi,
We just got a new cluster, which is much better hardware spec than current one. but when we run the performance test, on the same job, a new one was slower than old one.
Has any document or troubleshooting tips on this kind of issues?

thanks
David

ChuckB

Assuming you don't have access to any tools to accelerate the process, I'd start by doing a PROFILE on a simple query, for example:
PROFILE select date,count(*) from table group by date;
We have a number of articles about this, but the short version is to use the session/transaction and statement number to query dc_query_executions. This will help you identify if the time increase is in the client or server, query planner, network, or executor, that sort of thing. If it is in the execution, then you can look at execution_engine_profiles to see where the time went, setup, individual operations, etc.

It is worth running and reviewing the Vertica performance tool output. This may tell you if the CPUs aren't coming up to speed quickly, or something like that, and you might be able to tune your hardware settings.

qinchaofeng

You can look at the explain ,then confirm resource pool and execute a few more times to compare

davds

I narrowed down the issue, even I did simple sql it was much slower than old one.
sql like: select date,count(*) from table group by date;
any configuration I need to check?
thanks
David

cdaly1970

Hi,
You comment is that the new hardware spec is better, but how is it better? Are you sure the hardware is configured to run optimally for low latency? If you can provide some more information on the new hardware, it would be helpful.
-Chris