Solved: Is there a capacity planning tool available for On...

Report Inappropriate Content · 05-10-2016 07:55 AM

Is there a capacity planning tool available for On Prem installations?

Would like a tool to help with calculations.

adas

This is a multi-dimensional problem so coming up with a tool for capacity planning is not going to be easy unless you know the customer's API patterns, what those APIs are going to do, which policies are they going to use etc. The capacity planning tool above only deals with runtime components like message processors or routers and that too is an approximation. There's no way you can come up with these numbers for data components like Cassandra because it completely depends on what sort of policies and APIs that they are going to run. The tool also doesn't take into account any scaling needed for Qpidd and Postgres which are vital too, because beyond a certain threshold these components would not be able to handle the incoming load. The only capacity planning you can do is in terms of routers and MPs (sizing, number of instances etc) but that in my opinion doesn't help much because the scenario changes the moment you throw in things like node.js or policies that exercise cassandra extensively. Its not just the policies but also the usage patterns and the amount of data.

View solution in original post

Former Community Member

benchmark.apigeese.net/#/runs

Former Community Member

benchmark.apigeese.net/#/runs

adas

This is a multi-dimensional problem so coming up with a tool for capacity planning is not going to be easy unless you know the customer's API patterns, what those APIs are going to do, which policies are they going to use etc. The capacity planning tool above only deals with runtime components like message processors or routers and that too is an approximation. There's no way you can come up with these numbers for data components like Cassandra because it completely depends on what sort of policies and APIs that they are going to run. The tool also doesn't take into account any scaling needed for Qpidd and Postgres which are vital too, because beyond a certain threshold these components would not be able to handle the incoming load. The only capacity planning you can do is in terms of routers and MPs (sizing, number of instances etc) but that in my opinion doesn't help much because the scenario changes the moment you throw in things like node.js or policies that exercise cassandra extensively. Its not just the policies but also the usage patterns and the amount of data.

mschreuder

There is one other consideration here, for on-premise, the customer has the choice over their hardware, physical or virtual. The actual performance will also be a factor of what type of instances they are using.

Net of all of that, it would take a substantial effort and cost to build any such tool that had even a reasonable chance of giving a reliable capacity projection.

Ideally the Customer should start with a test environment that is grounded in whatever their regular hardware/vm choice is with a topology that is consistent with their desired operations requirements, run some load tests to get a baseline and then tune.

Report Inappropriate Content

It might be useful to provide a few examples (AWS, Physical, Virtual, traffic type, etc.) so people could at least use these as a starting point for their capacity planning estimates.