How to avoid the Amazon EC2 rip tide?



We all heard about global warming, that a few degrees rise in the environmental temperature causes polar ice to melt, raises the ocean levels by a few feet, in turn causes a havoc with more tornados, flooding and loss of life.

Situation is not so dire on Amazon EC2 yet, but a small activity can change your server’s load impacting its response time for your website. How much depends on when you took its baseline, e.g., performance changes but as compared to what?

To find an answer, we took baseline of a recently provisioned EC2 micro machine, and then ran a wordpress blog on it. This alone promptly changed its performance to Red, which stayed like this for days.

With Default Baseline

However, taking another baseline reading with wordpress running, and then monitoring the server with Meghafind scouts gave a very different picture. What changed is the bottomline from which the new instantaneous performance was measured. The small gyrations you see is not due to our activity, but presumably the noisy neighbors, or other guest VMs, that Amazon has decided to put on the same physical server on which our wordpress is hosted.

With Updated Baseline

Then what should an EC2 user do? We suggest to take several baselines under different loading conditions, and define what’s normal or acceptable for you. Then use it to measure when performance changes and becomes unacceptable.