Introduction
Project Bacalhau is a network of compute resources made available to serve computation workloads. Its primary function is to augment the Filecoin & IPFS networks, by allowing to execute arbitrary compute next to the data, but the long term vision is to allow any compute to execute, independent of storage. Bacalhau supports docker containers and WASM binaries and simplifies use of data stored in IPFS and Filecoin by making it appear native to the computational workload. The orchestration of these jobs occurs first, by pushing the compute to where the data is already being stored (by looking up CIDs on the Bacalhau nodes and bidding when it is found) or, failing that, any other node in the network (where the data is automatically downloaded on behalf of the user). This architecture is referred to as Compute Over Data (or COD). The Portuguese word for salted Cod fish is Bacalhau [bakaˈlyaw].
Project Bacalhau focused its first 6 months of development on succeeding in its performance and scale goals. In order to reach scale, we will need to expand beyond just “volunteer” networks, and will need to provide an economic incentive. These will include:
- Incentivizing community members (and Filecoin Service Providers) to run Bacalhau compute nodes as Compute Providers (CPs)
- Incentivizing CPs to both serve (execute) compute workloads and validate the proper execution of other
- Support for Filecoin+ data processing on Bacalhau, and promoting jobs that have been executed from FIL+ deals as “verified results”.
References:
Key questions
Token Incentive Design
- Should Bacalhau issue its own token, operate based on payment of FIL token, or some combination of the two?
- If Bacalhau operates its own token, which blockchain network would be best: a FIL EVM ERC20 token, EVM compatible L1, EVM L2 Rollup (Optimism, Arbitrum, etc.)
- Should Bacalhau support cross-chain incentives natively
- Which opportunities will we have to test/simulate/vet the token incentive design and then iterate or change course? E.g. could the incentive designs be simulated initially, validated in a test network, then rolled out to a mainnet launch? How difficult would it be to change incentive design post launch?
- Should Compute Providers earn block rewards for compute jobs? If so, how is Filecoin protocol updated to support this?
Risk Minimization
- If the incentive design involves a Bacalhau network specific token - which options exist to minimize its legal, regulatory risk? For example, should the incentive design include an inflated token supply so that the token decreases in value over time and is obviously a “utility token” and not an “equity token”?
- Should incentives be designed to enable penalization of malicious user behavior (e.g. bot network attacks)?
Compute Provider (CP) Incentives
- Should the CP be required to stake any portion of a token in order to establish a degree of trust and participate in the Bacalhau network?
- Should the stake be subject to slashing in case of malicious behaviors?
- Should other compute token holders be able to delegate their tokens to CPs for staking risks and rewards (similar to the Cosmos validator and The Graph delegation models).