Provers

Last edit

Contributors

This topic previously appeared in the Starknet Book.

The only Prover in use on Starknet as of this writing is SHARP.

SHARP is like public transportation for proofs on Starknet, aggregating multiple Cairo programs to save costs and boost efficiency. It uses recursive proofs, enabling parallelization and optimization, making it more affordable for all users. Critical services like the gateway, validator, and Prover work together with a stateless design for flexibility. SHARP’s adoption by StarkEx and Starknet highlights its significance and potential for future optimization.

This topic discusses SHARP, how it has evolved to incorporate recursive proofs, and its role in reducing costs and improving efficiency within the Starknet network.

What is SHARP?

SHARP, which stands for Shared Prover, is a mechanism that aggregates multiple Cairo programs from different users, each containing different logic. These Cairo programs are then executed together, generating a single proof common to all the programs. Rather than sending the proof directly to the Solidity Verifier in Ethereum, it is initially sent to a STARK Verifier program written in Cairo. The STARK Verifier generates a new proof to confirm that the initial proofs were verified, which can be sent back into SHARP and the STARK Verifier. Details for this recursive proof process appear below. Ultimately, the last proof in the series is sent to the Solidity Verifier on Ethereum. In other words, there are many proofs generated until we reach Ethereum and the Solidity Verifier.

The primary benefit of SHARP lies in its ability to decrease costs and enhance efficiency within the Starknet network. It achieves this by aggregating multiple Cairo jobs, which are individual sets of computation. This aggregation enables the protocol to leverage the exponential amortization offered by STARK proofs.

Exponential amortization means that as the computational load of the proofs increases, the cost of verifying those proofs rises at a slower logarithmic rate than the computation increase. As a result, the cost of each transaction within the aggregated set is significantly reduced, making the overall process more cost-effective and accessible for users.

In the context of SHARP and Cairo context, a jobs refer to the individual Cairo programs or tasks submitted by different users. These jobs contain specific logic or computations that must be executed on the Starknet network.

Additionally, SHARP enables smaller users with limited computation to benefit from joining other jobs and share the cost of generating the proofs. This collaborative approach is similar to using public transportation instead of a private car, where the cost is distributed among all participants, making it more affordable for everyone.

Recursive proofs in SHARP

One of the most powerful features of SHARP is its use of recursive proofs. Rather than directly sending the generated proofs to the Solidity Verifier, they are first sent to a STARK Verifier program written in Cairo. This Verifier, which is also a Cairo program, receives the proof and creates a new Cairo job that is sent to the Prover. The Prover then generates a new proof to confirm that the initial proofs were verified. These new proofs can be sent back into SHARP and the STARK Verifier, restarting the process.

This process continues recursively, with each new proof being sent to the Cairo Verifier until a trigger is reached. At this point, the last proof in the series is sent to the Solidity Verifier on Ethereum. This approach enables greater parallelization of the computation and reduces the time and cost associated with generating and verifying proofs.

Generated Proofs

STARK Verifier program (in Cairo)

Cairo Job

Prover

New Proof Generated

Repeat Process

Trigger Reached (last proof)

Solidity Verifier

At first glance, recursive proofs may seem more complex and time-consuming. However, there are several benefits to this approach:

  1. Parallelization: Recursive proofs enable work parallelization, reducing user latency and improving SHARP efficiency.

  2. Cheaper onchain costs: Parallelization enables SHARP to create larger proofs, which were previously dependent on the limited availability of large cloud machines. As a result, onchain costs are reduced.

  3. Lower cloud costs: Since each job is shorter, the required memory for processing is reduced, resulting in lower cloud costs.

  4. Optimization: Recursive proofs enable SHARP to optimize for various factors, including latency, onchain costs, and time to proof.

  5. Cairo support: Recursive proofs only require support in Cairo, without the need to add support in the Solidity Verifier.

Latency in Starknet encompasses the time taken for processing, confirming, and including transactions in a block. Latency is affected by factors like network congestion, transaction fees, and system efficiency. Minimizing latency ensures faster transaction processing and user feedback.

Time to proof, however, specifically pertains to the amount of time required to generate and verify cryptographic proofs for transactions or operations.

SHARP backend architecture and data pipeline

SHARP’s backend architecture consists of several services that work together to process Cairo jobs and generate proofs. These services include:

  1. Gateway: Cairo jobs enter SHARP through the gateway.

  2. Job Creator: This service prevents job duplication and ensures that the system operates consistently, regardless of multiple identical requests.

  3. Validator: The validator service runs validation checks on each job, ensuring they meet the requirements and can fit within the Prover machines. Invalid jobs are tagged as such and do not proceed to the Prover.

  4. Scheduler: The scheduler service creates trains that aggregate jobs and send them to the Prover. Recursive jobs are paired and sent to the Prover together.

  5. Cairo Runner: This service runs Cairo for the Prover’s needs. The Cairo Runner service runs Cairo programs, executing the necessary computations and generating the execution trace as an intermediate result. The Prover then uses this execution trace.

  6. Prover: The Prover computes the proofs for each train.

  7. Dispatcher: The Dispatcher serves two functions in the SHARP system.

    • In the case of a recursive proof, the Dispatcher runs the Cairo Verifier program on the proof it has received from the Prover, resulting in a new Cairo job that goes back to the Validator.

    • In the case of a proof that is to be published onchain, the Dispatcher creates packages from the proof, which can then be sent to the Blockchain Writer.

  8. Blockchain Writer: Once the packages have been created by the Dispatcher, they are sent to the Blockchain Writer. The Blockchain Writer is responsible for sending the packages to the appropriate blockchain, such as Ethereum, for verification. This step in the SHARP system ensures that the proofs are properly verified and that the transactions are securely recorded on the blockchain.

  9. Catcher: The Catcher monitors onchain transactions to ensure that they have been accepted. While the Catcher is relevant for internal monitoring purposes, be aware that if a transaction fails, the fact won’t be registered onchain in the Fact Registry. As a result, the soundness of the system is still preserved even without the Catcher.

SHARP is designed to be stateless. That is, each Cairo job is executed in its own context and has no dependency on other jobs, enabling greater flexibility in processing jobs.

Current SHARP users

Currently, the primary users of SHARP include:

  • StarkEx

  • Starknet

Challenges and optimization

Optimizing the Prover involves the numerous challenges and potential projects on which the Starkware team and the community are currently working, including:

  • Exploring more efficient hash functions for Cairo, the Prover, and Solidity.

  • Investigating smaller fields for recursive proof steps could lead to more efficient computations.

  • Adjusting various parameters of the STARK protocol, such as FRI parameters and block factors.

  • Optimizing the Cairo code to make it faster, resulting in a faster recursive Prover.

  • Developing dynamic layouts, which should enable Cairo programs to scale resources as needed.

  • Improving scheduling algorithm. This optimization path is external to the Prover.

Dynamic layouts enable SHARP to determine and scale the required resources for a specific job and adjust the layout accordingly, instead of relying on predefined layouts with fixed resources. Scaling resources can lead to more efficient computation and better resource utilization. This approach can provide tailored solutions for each job, improving overall efficiency.

Conclusion

In conclusion, SHARP is a critical component of Starknet’s architecture, providing a more efficient and cost-effective solution for processing Cairo programs and verifying their proofs. By leveraging the power of STARK technology and incorporating recursive proofs, SHARP plays a vital role in improving the overall performance and scalability of the Starknet network. The stateless nature of SHARP and the reliance on the cryptographic soundness of the STARK proving system make it an innovative and valuable addition to the blockchain ecosystem.