The other table (myisam) is updated every 6 hours with a procedure (lets say its name is X: Applications use X1 Insert new data into X2 Send applications to use X2 Delete all from X1 Copy all from x2 to X1
Most Inserts happen at these four times of the day.
generally I wouldn't assume a slowdown is fragmentation (unless the tables have never been defragmented, which is possible)
if it turns out not to be, your next step would probably be to do an analysis of the slow query log and running it through pt-query-digest after a couple days worth of logs
do you know if it's MySQL cluster or percona xtradb cluster? I'm assuming the latter, because MySQL cluster users NDBCLUSTER engine (it's not myisam or innodb).
I've also not really worked with either one, so wouldn't know how to direct you here
Well, my plan is to setup the slow log and analyze it first. In the meantime, I could increase the key and innodb buffer sizes. With 64GB or RAM, having 27 unused seems a waste.
I suppose there is no harm in raising them.
Upgrading will be a suggestion for later. And moving from VPS to a real server.
