Performance Metrics

We recently conducted a UNITS benchmark using a SN Stack instance and achieved the following results.

• The benchmark did basic transfer transactions. The transactions were suited for parallel processing i.e. they did not touch the same storage slots. Here’s the link to an example transfer transaction.
• We were able to process 7282 transactions in 1s

The UI broke :)

You can verify these results on the explorer where you can see the transactions processed in each block and the time to create (1s).

ZK Proven Transactions

All transactions in our benchmark have been ZK proven with their final state published on Ethereum. You can verify these transactions here:

• Every block on the chain can be ZK proven and verified. We did this for 10 blocks with 7k TPS. It can be verified here.
• State update for block 4149

  • You can see here in the decoded calldata on the page. It contains the state root and the block number.

    

  • You can verify the state root on the explorer too by going to block 4149 here.

    

TPS without Merkleization

Ignoring merkleization (the process of recursively hashing DB entries and finding one root for the entire DB), we were actually able to do 15k TPS as seen in the screenshot below

Decentralized vs. Centralized Approaches

In a decentralized context, merkleization must happen synchronously (although some decentralised chains choose not to merkleize and sacrifice storage proofs, more on storage proofs below). The synchronous requirement exists because:

1. One leader processes transactions
2. The leader merkelizes the data