Comments for Pierre

[WantingDU]

• Service disruption detection VS Anomaly detection

• Acronym SDD to be added into the terminology part.

• 3.1. Collection: The duration of retention is relatively trivial in the collection phase due to the resource limitation. it's much more important in the long-term storage after the aggregation step. Instead, data granularity and sampling rate, aggregation time bucket during the collection may be more worth mentioning.

• 3.3.4. Machine Learning approaches:
  (Just input, this part might be removed for the first version ACK)
  + discover what people don't see
  - requires data quality (ground truth) and data amount
  - hard to adjust, no much control
  - sometimes can be resource intensive (eg. multiple models per one @vpn)

[pierre-francois]

I'm not sure duration of retention is that trivial :)
I would suggest we go into aggregation time bucketing as well as sampling rate in a next revision?