Alpha-2 Testnet Update

We launched our alpha-2 testnet a bit over two weeks ago, and it has been a very useful one in terms of testing, finding issues, and other takeaways.

We, and you, have utilized this testnet to primarily test decentralized proving and protocol economics. We want to say thank you to all who have helped test—transacting, deploying, node running and proving—and also want to express our apologies for any sub-optimal experience (transaction delays) you may have encountered in helping surface some of the issues we explore below.

Highlights

  • Permissionless proving

    • 126 unique provers

    • 93,146 blocks proven

    • Proof time ranging between 130 and 160 seconds in the past week.

Proving stats dashboard at https://taiko.zkpool.io, thanks to ZKpool.
Proving stats dashboard at https://taiko.zkpool.io, thanks to ZKpool.

Sources to explore further: https://status.a2.taiko.xyz/, https://explorer.a2.taiko.xyz/, https://taiko.zkpool.io.

It is very inspiring to see a decentralized prover network spring up from the community. Thank you to all who are participating.

The distribution of successful provers also shows an expected outcome in the open competition model where the most efficient provers generate the proofs first and most consistently.

  • Monitoring / alerts

    • Successfully tested alerting system when network undergoes stress, latency, or bugs.
  • Bridge

    • Relayer successfully processed 150k+ messages.

    • BLL token, which is designed to fail 50% of the time, let us successfully test bridge message failures, to let users recover their funds on the source chain in the event of an issue.

  • Dapps

    • Our deployed swap dapp and other contracts were identical to the mainnet ones, validating the 1:1 functionality of our network with Ethereum mainnet.
  • Developer Experience

Issues identified

  • Protocol economics (leading to lengthened proposal times as a mitigation)

    • There is a design flaw in our protocol economics. When blocks are proposed with fixed intervals, the prover rewards will monotonically increase, and now it has reached an untenably high value. If we had kept the block time short, the reward amount will eventually have overflown, and the chain would have halted.

    • To prevent that, we had to slow down block proposal (proposal interval of 20 seconds) while working on patching the aforementioned contract flaw/bug. Slowed block times then led to a congested mempool and degradation in user experience that manifested itself in the form of long delays for bridge or pending wallet transactions (Reminder: in alpha-2, we have a single proposer and decentralized provers, the inverse of alpha-1.)

  • Low precision in average calculation

    • The block time and proof time average calculation has a low precision, which ends up with block interval being 1 second and is essentially stuck there. This is an implementation bug. In our big pull request for the alpha-3 testnet, this bug has actually been fixed, but we need to cherry pick it and apply it to the alpha-2 branch.
  • Contracts non-upgradeability

    • Our testnet deployment script did not enable contract upgradability using proxies. This bug prevented us from fixing the above bugs. Of course, upgradability is an important part of a network’s defenses/training wheels, especially in its early days. Not implementing upgradability for this testnet is mostly an oversight, and partially driven by our accepted approach of fast-moving testnet iterations and deprecation.

What’s next

We will keep alpha-2 in a low-throughput mode and hopefully find more issues, together with you. We know transactions lingering in the mempool can be frustrating, but we must continue to throttle block times to avoid the more nefarious overflow. We do not guarantee testnet throughput. If you run into new bugs not mentioned here, please do report it.

Alpha-2 has created good opportunities for us to improve our rollup protocol and also gain hands-on experiences in upgrading/forking the chain. We value the lumps as much as the wins, so again a big thank you to all testers. We do plan on continuing along a path of maximum learning potential, which could see more rapid instantiations and deprecations as we get what we need. We do expect to deprecate alpha-2 and launch alpha-3 in Q2.

Join us 💗

Explore open positions on our job board.

Stay on beat with Taiko 🥁

Subscribe to Taiko Labs
Receive the latest updates directly to your inbox.
Mint this entry as an NFT to add it to your collection.
Verification
This entry has been permanently stored onchain and signed by its creator.