Round 1 / Spark Protocol

Prior Art: Previous Filecoin retrieval bots retrieved pieces from Storage Provider by their Piece CID. These requests were made from a set of Protocol Labs controlled servers.

Spark improves on this in the following ways.

• Spark is the first retrieval protocol that samples and measures payload (or file) CID retrievals instead of piece CIDs. This matches what future hot retrieval traffic will look like.

• Spark, leveraging the Station Network, uses a decentralised network of clients to make the requests and rewards them in FIL. This means that Storage Providers are much less able to distinguish between real traffic and sample traffic and so must provide a good service to both. This also keeps the infra costs in the Filecoin ecosystem rather than going to cloud giants.

• Spark uses FVM for managing the payouts to Station Operators running the Spark client.