Re: Query Parallelism and Multiple Volumes

[Ulf Wendel]

Well, depends what we're talking about. The index scan will be performed serially but the question is where the index data is:

- in the buffer/main memory
- on the disk

If the information are in the main memory, it's a sequential operation. However, if the question is more about disk IO, it's kind of parallel.

MaxDB spreads the data over all volumes and has dedicated IO worker for each volume which get their tasks from queues. Thus you can expect the queue with the IO requests be filled serially but the IO worker fetching the data asynchronously (and in parallel) into the buffers. Once it's in the buffer, it's not a parallel operation to compute the join result - strictly speaking.

However, there's parallel server for parallel index read in the latest versions...

Depending on what the person who asks wants to know or into which light you want to put MaxDB you can answer the question with "yes" or "no". I guess the question was about IO, therefore: yes, asynchronous fetching by multiple IO worker wll be performed which means parallel activities going on.

Ulf