A few AWS CloudWatch productivity tips

For those getting into AWS CloudWatch dashboards, I thought I’d share a few productivity tips on changing to your local timezone, using reference lines and comparing averages to peaks.

Tip 1: yes, you can get metrics using your timezone

Is this a relief or what! I spent too much time calculating what my time is compared to UTC so I could figure out when my latency peak is, until I found this little jewel at the top of the dashboard, in time selection. It’s not so easy to spot so here it is:

Open the custom menu in the right hand corner

Select UTC in the top right corner of the custom menu, which will give you the option of Local Timezone. Voilà!

Tip 2: Put some reference (annotation) lines in there

Metrics in your charts change over time, so it’s helpful to set a few reference lines. AWS calls them annotations. You can set them in the Graph Options tab in the metric edit view.

You can choose the value, which axis to put them on, and whether to fill the graph above or below the reference line.

A permanent reference line.

Very helpful when looking at a high-level view of several metrics on a page: is this thing over or under what I consider acceptable?

Tip 3: Look at peaks instead of averages

By default, the metrics you create in the AWS CloudWatch dashboards are averages. This is good information, but what if you need to identify peaks, for example CPU peak times? They don’t show up in the averages at all. Consider this visualisation:

This graph actually has four values in it: peaks for all servers and individual averages for each server.

In this graph, four things are displayed: The Latency Peaks for all three servers, as well as the average latency per server. As you can see (or can hardly see, by the looks of it), average latency is extremely low, sub-second low actually, whereas the servers do have occasional latency peaks, even up to 24 seconds at times! Huge difference, yeah?

So, it’s important to look at both averages, but also consider peaks. Combined with tip #1, you’ll know exactly at what time your peas occur. That’s useful information for example for CPU, disk throughput or in the above case, latency metrics.

This is how the above graph is set up. It’s measuring the latency metric for an Elastic Load Balancer:

Graphed Metrics section, select “Maximum” as the Statistic for the overall metric.

Anyway, that’s all the AWS CloudWatch tips for today. Thanks for the read, please share with your AWS friends!