Hi,
Cluster is going to run slower than most regular databases on a single system. This is due to the fact that cluster is designed to be run across multiple systems, not a single one, so uses more reliable interprocess communications and signaling systems. If you really want to run it all on a single system, then you should consider setting up the shared memory transport as talked about at
http://dev.mysql.com/doc/refman/5.0/en/mysql-cluster-shm-definition.html to minimize the overhead.
You never told me the results of switching everything to the gigabit setup and more SQL nodes? Did you see a speed increase or not? Did you monitor the network load or load averages of the different systems while it was running?
Also just to manage expectations a bit, writes are going to be slower in general than a local system until you saturate the disk subsystem on the local database. This is because you are writing data to multiple places compared to one local place. That is the price to pay for high availability in this case. If you don't care about high availability, then consider setting NoOfReplicas=1 in the config.ini which would be closer in availability (and performance of writes) to a single system setup.
Check out
http://dev.mysql.com/tech-resources/articles/mysql_clustering_ch5.pdf which is the performance chapter I wrote regarding a lot of this, which might help you understand a bit more what is going on.