Running Read heavy queries on Percona XtraDB cluster

Question

I am creating a new 3 node Percona XtraDB cluster from scratch for a brand new application. Part of the application will generate read heavy queries which we intend to loadbalance across the cluster. Previously in a Master slave setup those read queries used to generate replication lag which effected our application. Since we needed to fetch reads with not more than 3 seconds latency.

Now with Galera cluster it looks as if that problem is solved as replication is managed differently.

My question is, would those read heavy queries would cause galera cluster to cause some problems ?. More importantly would these queries slow down write performance of the cluster ?

Any experience in this area would be helpful.

Answer 1

Yes and no.

Galera-based solutions (such as PXC) do have different characteristics. It depends read vs write, distance between nodes, length of transactions, etc, etc. Also, Galera has a way to deal with "critical reads" ( wsrep_sync_wait ) that Master-Slave did not have.

In any kind of system, you should study read-heavy queries to look for better indexes and/or better query formulation; this will improve any system. Would you care to start another question to tackle such?