Cluster load averages

[Alastair Symon]

We're testing MySQL cluster with 1 management, 2 storage nodes and 1 API/mysqld node (all on different boxes) prior to a roll out to a small web server setup with 4 load balanced apache/php/mysqld boxes.

On the current test set up the two ndbd machines are a twin xeon 3.0G with 4G ram and a single xeon 3.0G with 4G ram on standard linux kernels. They have nothing else running on them in this setup.

My question is what load average should we expect to see on the ndbds? With no sql requests the load average varies greatly up to a maximum of about 1.00 and mostly sits around 0.3/0.4. This seems very high for a cluster 'at rest'. Is this typical, are we doing something wrong, or is there some tuning we should do to the cluster?

Our cluster setup is pretty much plain vanilla with ample ram available to ndbd (66% used at present).