I am building a simple integration with AWS cloudwatch that basically scrapes metrics from cloudwatch and does the following:<p>1. learns what is normal from past behaviour and alerts on abnormalities. Thus no need to set and maintain alert conditions on hundreds of metrics (at least not on the low level non-SLA metrics).<p>2. auto clusters the above into incidents. Example - if 100 EBS volumes have a unusual write activity alerts, and all are attached to EC2 instances which are within an ELB target group, they will be clubbed together.<p>Is this something you find useful? What other intelligent features you wish cloudwatch had ?
You can try Stackdriver (owned by Google Cloud, but has AWS integration). Stackdriver allows you to aggregate multiple distinct metrics into one timeline, and you can set alerts on the aggregate, etc.