Re: Speed Issue

[Harrison Fisk]

Hi,

Yes, generally communication happens between all of the data nodes when running. This is to synchronize locking, data writes, etc... between the nodes. This is required to make sure that everything is consistent between the different systems.

There is some work to allow MySQL Cluster to pick certain nodes based on what the connection being used is, but that only applies with things like shared memory vs. tcp/ip vs. SCI and such. It isn't always that large of a performance gain either.

MySQL Cluster is designed to run on high performance networks, not slower ones (same as most clustering software). If you have to have a slow network involved, then I would look into replication as it performs well over slow interconnects. As of MySQL 5.1, you can also replicate between seperate clusters as well across slow connections.