Galera cluster latency issues

[atulmv]

I ran into a seemingly well written article today on setting up a Galera cluster with master-master replication on Ubuntu. The questions that appear unanswered to me

a. Suppose I setup a cluster with 5 servers scattered across the four corners of the globe. Will the efforts at keeping them in synch result in a significant latency issue that will be to the detriment of the system as a whole?
b. Am I write in thinking that provided I get Galera up and running properly I can pretty much leave my PHP PDO code unchanged in the blissful assumption that Galera is handling replication so well that I have a cast iron guarantee of being able to access the same data on each server?

[przemek]

So as Galera is called “virtually synchronous replication” you may expect that the latency between the nodes will affect the whole cluster performance. It also applies (however in a bit different extent) for the individual nodes capabilities as the slowest node may again affect whole cluster operations.
Regarding accessing same data on each node (server) - see the causal reads setting if you really need it.
From my experience you will want rather reliable WAN links between the nodes.
Also since Galera 3 there are improvements to handle WAN scenarios better:
http://www.mysqlperformanceblog.com/2013/12/19/automatic-replication-relaying-galera-3/

Some more hopefully interesting lectures for you:
http://www.mysqlperformanceblog.com/2012/01/19/percona-xtradb-cluster-feature-2-multi-master-replication/
http://www.mysqlperformanceblog.com/2013/03/03/investigating-replication-latency-in-percona-xtradb-cluster/
http://www.mysqlperformanceblog.com/2012/06/14/comparing-percona-xtradb-cluster-with-semi-sync-replication-cross-wan/