Break down retrieval success rate by Deals, Clients, Allocators, MetaAllocators

[patrickwoodhead]

eta: 2024-09-30

Currently, the Spark Retrieval Success Rate (RSR) is aggregated by Storage Provider. This gives us a Spark score for each Storage Provider. However, there are different personas in the FIL+ community who would be interested in seeing the RSR of Filecoin aggregated differently.

Firstly, there is the breakdown by deals. Since Spark is based on payload CID rather than piece CID, there might be multiple retrieval tasks in the same deal. It would be great to see the Spark RSR and the error breakdown by deal.

Secondly, it would be great to see the Spark RSR aggregated by the clients making the deals. Clients have a responsibility in the datacap flow to allocate their allowance as best they can. The allocators would like to be able to see the performance, including the RSR, aggregated by client

Thirdly, moving up a level, meta-allocators and root key holders would like to see the RSR of all deals linked to allocators and meta-allocators.

[bajtos]

Firstly, there is the breakdown by deals. Since Spark is based on payload CID rather than piece CID, there might be multiple retrieval tasks in the same deal. It would be great to see the Spark RSR and the error breakdown by deal.

See:

• Deal Retrievability Score #156

[bajtos]

See also:

• Provide miner RSR per client and per allocator spark#79