When to scale Apigee infra ?

[#1]

Hello All,

When do we need to add new Message processor, Message router, or Cassandra database node to our existing setup? Is any documentation available? What is the TPS that each MP or MR will take ?Any benchmark?

[williamspaul-1]

Hey Rajesh,

I sense a couple of questions:

Q1) Why would I need to scale components of Apigee?

A1) You scale Router/MessageProcessors in response to traffic increases, and you scale Cassandra nodes in response to data requirements increases.

Q2) How do I know when I need to scale?

A2) Unfortunately, there's no good universal measure of what constitutes a scale need. The specifics will be more bound to the complexity and traffic needs. A very high traffic event may exceed the capabilities of the TCP stack, or perhaps lower traffic with a very high proxy complexity may cause the need to scale.

Adding to the complexity of the answer is that depending on how you have deployed the platform may affect how it needs to scale -- e.g. RHEL vs. Debian vs. Suse and bare metal vs. VMs vs. Amazon vs. GCP vs. Rackspace etc may all have different kinds of impacts.

My advice: Run performance tests using a real-life traffic profile on production-parity platforms with tools like Grinder, JMeter, LoadRunner, Parasoft, or any other load test tool. A basic statistical regression should give you a very good mathematical model for scaling operations.

Good luck!
~ Paul