Hi Ian,
Sorry for the late answer, coming back from Thanksgiving holiday week-end!
We're evaluating Sequoia for redundancy and load balancing of our
MySQL database. I've got all kinds of questions, but I'll start with
performance and monitoring.
What do you monitor on Sequoia?
Whatever you are interested in. Most people just monitor the status of
the cluster (up or down), but you can do some performance profiling too.
On a system level, obviously CPU, memory, and network throughput. Is
it possible to monitor number of client connections? Number of
queries per minutes? Response time of database instances? And, if
so, how? I see there's a JMX interface. Is this information
available over JMX?
You can look at the source code of the ShowBackend command of the text
console, it retrieves statistics per backend (a BackendStatistics
object) that contains information such as number of
connections/request/transactions/...
Request response time can be profiled using the SQLMonitoring option in
your virtual database configuration file. You can retrieve this info
through JMX or the console. Another option is to use the Request logger
in log4j.properties and use an appender that suits your needs (e.g., a
socket appender)
Also, I assume Sequoia is generally bound by network throughput. Is
this a correct assumption?
Not necessarily. You can see a high CPU usage if you did not allocate
enough memory to the JVM and the GC has to run all the time. You can
also see significant memory/disk usage if you have a recovery log
database co-located with the controller. Network latency is usually more
of an issue than throughput. Unless your database completely fits in
memory, it is likely that your network throughput will always be greater
than your disk throughput.
The group communication configuration (must provide total order) will
usually not provide any speed close to your maximum network throughput.
Also group communication is only used for write requests so you have to
know how to understand what queries are executed to better undestand
network numbers.
Also be careful about the amount of logging you are using since logging
can be a big disk io consumer.
Thanks for your interest in Sequoia,
Emmanuel
--
Emmanuel Cecchet
FTO @ Frog Thinker
Open Source Development & Consulting
--
Web: http://www.frogthinker.org
email: [EMAIL PROTECTED]
Skype: emmanuel_cecchet
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
