We currently need a way to have performance profiling management solution in place for Neo4j running in Windows Azure cloud. The benefits of course is on going data analysis and performance statistics, not to mention assisting in debugging issues with performance. The best part of it all, is the agent that collects the data just runs as part of the JVM and the data is automatically uploaded to the website where you can view it online.
New Relic Account
The first thing you will need to do is create a new relic account at:
Once, this is done, you can then download the java agent which will contain two files:
The yml file contains the license key and the application name to display on the new relic performance dashboard website. When you use New Relic it has a trial option, so it is easy to test out.
What we do is store these zip files in blob storage and when the worker role is bootstrapping, it will then download the zip file, and then automatically edit the neo4j config files before starting up neo4j.
It is extremely simple to configure the relic agent to run and profile neo4j, all you need to do is edit the neo4j-wrapper.conf file and add this line of code.
We use a relative path, as we store the newrelic jar relative to the neo4j binaries, so all you need to do is store the newrelic.jar file in a location where neo4j can access it from when starting up.
Once this has been deployed to the cloud, we then have performance statistic automatically made available to use via the neo4j JVM on the New Relic web site!
From here, you can actually click a segment on the graph and drill into the method level calls that occurred in the JVM.
Notice that you can get details about the timing of method invocation timings.
What is really cool, is comparing response times between environments, so you can see how fast UAT/Prod/Dev are compared to one another
It is a relatively easy task to get application performance statistics for neo4j running in or out of the cloud and New Relic seems to be a really useful tool with minimal overhead to get up and running, so I would highly recommend using the combination together. This coupled with VisualJM should provide enough performance data and profiling when collecting performance data and compiling reports.