FYI, we just looked into putting a 75-100GB table into the cluster, but
figured out that we can house the same table by load balancing
traditional InnoDB databases across ~4 servers, rather than across ~16
with the cluster.
Realized that I didn't mention something which is imporant here - this
table, though large, doens't handle more than 6,000 queries / second. If
it needed to handle much more than that, we would probably need to put
it into a cluster since we seem to be pushing our qps limit with it
right now.