Overview

When Filecoin Storage Providers accept certain FIL+ Storage deals, there is an agreement to make the data in the deals available for “fast retrieval”. This means as well as storing the sealed copy of the data, another “hot copy” of the data must be stored so that it can be retrieved without having to unseal the sealed copy.

However, this agreement to make a hot copy available is not guaranteed by the core Filecoin protocol. Spark aims to solve this.

Spark is a trustless protocol for verifiying retrievability of certain data from Filecoin Storage Providers. Spark checkers make randomly-sampled retrieval requests to Filecoin Storage Providers and report on the responses. Spark checkers are then rewarded for their contribution to Spark on a regular basis.

The data collected by Spark checkers can then be fed into dashboards, reputation systems or even back on chain.

Spark Protocol Steps

At a high level, the Spark protocol is split into the following steps. More information is given in the pages dedicated to each step.

1. Deal Ingestion: This step creates and maintains an overall list of tasks (CID, SP) to retrieval-test.
2. Task Sampling: Each round (~20 minutes), a random sample of tasks are chosen to be test. Spark checkers can only complete certain tasks from the sample.
3. Retrieval Checking: Spark checkers perform the assigned retrievals from SPs.
4. Measurement Submission: Spark checkers submit their results to Spark API. All measurements are committed to on-chain.
5. Evaluation of Measurements: Spark-Evaluate checks the validity of all measurements. Result are then made available through Spark API and on chain.
6. On-chain Checker Evaluation & Rewards: Spark checker evaluations are posted on chain and they are rewarded for their contribution in the round.

On this page

• Overview
• Spark Protocol Steps