Handle pre-aggregated metrics

[mikehale]

In the case of AWS cloudwatch, data has already been collected and aggregated over a 60 second period. Cloudwatch provides access to min, max, avg, sum, and sample count values for this 60 second period. I'd like a way to route these raw data points through l2met without any additional aggregation happening on them. For example, sending the following event to l2met:

$stdout.puts("measure.elb.latency.min=0.000s measure.elb.latency.max=1.660s measure.elb.latency.sum=2715.167s measure.elb.latency.sample_count=67448s source=elb018876.us-east-1d measure.time=1366295100")

Would produce the following metrics:

• measure.elb.latency.min
• measure.elb.latency.max
• measure.elb.latency.sum
• measure.elb.latency.sample_count

The metrics would all have a source of elb018876.us-east-1d and would be timestamped at the value of measure.time (1366295100).

I'm not tied to the exact formatting of the above, I mainly wanted a concrete example to show the desired result of passing pre-aggregated metrics through l2met. This example seeks to use l2met not as an aggregation tool, but simply as a way to route structured log data to a metric storage service in a way that does not require the originating application to be tightly coupled to the metric storage and visualization service. i.e STDOUT plus drains.

[mikehale]

Having this in place would allow me to eliminate m2met.

[ryandotsmith]

@mikehale I think the proposed changes in #69 would handle this case. Correct me if I am wrong.

[mikehale]

@ryandotsmith agree.