On Oct 19, 2006, at 8:26 AM, Scott McCool wrote:
We're looking at migrating to a 5.1 cluster of 4 machines. Our
primary goal
is redundancy, then scalability. I see how to make the ndb & data
nodes
redundant, but I'm not sure what the best practices are for
abstracting
application access. I'll run at least two mysql nodes, but what is
the best
way to make the cluster transparent to application code?
Can I stick the mysql nodes behind a BigIP pool?
Yes, that's always a good idea.
Some sort of DNS round robin?
Some sort of floating/virtual IP configuration?
The virtual IP is also possible, but in my experience a hardware load
balancer is usually a good way to go.
Also, for the best performance match between the NDB nodes and the
MySQL nodes, you may want to have twice as many CPUs on the MySQL
boxes as on the NDB boxes.
JD