I'm running a test Cluster (5.1.47-ndb-7.1.8-cluster) with two SQL
nodes and two NDB nodes. The MGM daemons run on the SQL nodes. The
machines are pretty fast ("m1.large" on AWS, approximately equal to
dual-core Xeon 2.27 GHz, 7.5 GB).

Normally, I had been running with the query cache disabled (query-
cache-type=0). When the cluster is doing nothing (that is, no queries
happening at all), the network traffic on each node, according to
Munin, is around 120 Kbps (in each direction, incoming and outgoing).

I decided to do some testing with "on-demand query caching" (i.e.,
using "SQL SELECT_CACHE .....") and enabled "query-cache-type=2
(DEMAND). After I restarted the cluster, the network traffic to/from
each node almost doubled and stayed that way until I disabled it and
restarted the cluster. Again, this is without any query traffic.

My question is, is this normal? It didn't matter whether I set
ndb_cache_check_time to 0 or to 1500 (1.5 secs), the result was the
same.

By the way, I verified that the on-demand query cache was in fact
working. (I did several queries that would normally take several
seconds and yes, they did finish in sub-second time, after the first
query, when using "sql select_cache".)

Also: I set query-cache-size=10485760 (10 MB) for testing purposes.

Any insights will be much appreciated!
-T.H.