parallelizing queries

[Niklas Saers]

Hi,
I've got a server setup with one master node and four replicants. We do a lot of analysing, using only the replicants, leaving the master for write queries. My problem is that at the moment only one slave at a time does the analysis.

For the analysis we use proprietary software that we have no control over. It uses single queries to gather all the data it needs, and these queries take up to 15 minutes, not because the data transferred is much, but because of many joins and limits and lots of data to be processed on the server.

I don't believe I can come much further in optimizing the speed on the single replicant, therefore I would like to use the other replicants to split up the processing so that the load is relatively uniformly spread across the nodes, the result gathered in the end and sent to the application. What I'm wondering about is: how do I do this?

I was hoping to use NDB Clustering for this, but since the NDB is for storage it is my understanding that the actual processing will still be done on one SQL node.

I was hoping to use partitioning for this, but as I understand it partitioning is only one one computer with multiple disks.

What can I do to have multiple SQL nodes working on a single query?

Cheers

Nik