Concepts

Some jargon and its explanation


Cost-weighted average utilization

Say we have 2 machines A and B:

The un-normalized average utilization would be

( 90% + 80% + 10% + 10% ) / 4
  ~ 47 %

The Cost-Weighted Average Utilization would be:

( 1 $/day * (90% + 80%)/2 + 10 $/day * (10% + 10%)/2 ) / (1$/day + 10$/day)
  =  (0.85 $/day + 1 $/day) / (11 $/day)
  ~ 16 %

The Cost-Weighted Average Utilization in the above example is much lower than the un-normalized average because it gives more weight to the underused and more expensive machine B than it does to the heavily-used and cheaper machine A.

To use isitfit to calculate this, please check the example on the Usage page.

Underused, Overused, Idle, Normal, Burstable

isitfit categorizes instances as:

A finer degree of categorization specifies:

The above categories are currently rule-based, generated from the daily cpu utilization of the last 90 days (fetched from AWS Cloudwatch).

Sizing is simply a rule that says: "If underused, recommend the next smaller instance within the same family. If overused, recommend the next larger one."

The relevant source code is here