Anomaly metrics

[pavolloffay]

Brainstorm a set of metrics that can lean to anomaly detection. Anomaly detection in a sense for anything which can improve performance, security...

For instance:

• RPC call within the same service
• dangerous SQL queries
• dangerous patterns in URL

[fktkrt]

I also have a few suggestions:

• anomalous spans with suspiciously long latency (compared to similar spans)
• unusual trace structures