Uptime Scoring System

[sameh-farouk]

Motivation

Encourages nodes to stay up and penalizes those that are consistently down.

Description:

The idea is to utilize the health indexer to Implement an uptime score system where nodes are penalized or delisted after prolonged downtimes.

• Nodes with consistent uptime gain higher scores, which would increase their chances of being selected for deployment.
• Nodes with poor uptime scores could be delisted for a period of time until their scores improve.

Implementation:

• Use a sliding window technique (e.g., the last 30 days) to continuously update the node's score based on its recent performance. This allows you to account for occasional downtimes while still penalizing nodes that consistently underperform.

• Calculate the uptime score of each node by tracking how often it is online within that time window. Nodes should then be listed or chosen based on their uptime score. Encourage consistent availability by giving nodes that meet expectations more opportunities to be utilized over time. Conversely, nodes with low uptime scores would have fewer chances to be utilized or might be temporarily delisted.

Update:

• An effective scoring implementation may involve utilizing the node's uptime reports from the GraphQL, as we already have accurate uptime data there (from node’s reports).
  We could periodically query the past 30 days' reports and calculate the score, allowing us to accurately score the stand-by nodes, especially if the health indexer treats them as unhealthy.

[xmonader]

For 3.16