Discontinuing the Deku Project

By Daniel Hines

‍

With sadness we must announce the Tezos Foundation has discontinued the Deku project. The betanets are being shut down, and the repository will be archived. With crypto winter upon us, the Foundation has decided to cut costs, and prioritize rollups instead.

What was Deku?

The Deku project began because we believe there at least two significant blockers to mass adoption of crypto:
- Latency is far higher than what users expect from modern apps.
- Throughput is far too low - if adopted en masse, fees would skyrocket in response to limited block-space.

While appchains and bespoke scaling solutions can help certain applications, these tend to miss the network effects that are core to the value prop of Web3. We need platforms that can support millions of users in an ecosystem of apps interoperating with low latency. Without this, the big players stay away, use-cases for crypto remain niche, and the vision of a decentralized web remains unrealized.

The goal of the Deku project was to tackle this problem with a sidechain for Tezos.
The design was based on a few relatively simple (but largely underexplored) tradeoffs:
- Foregoing Merkelization and of the state and light clients in favor of a flat, in-memory storage with asynchronous state hashing for checkpoints.
- Extensive use of parallelism for networking, deserialization, and signature verification, enabled by OCaml 5's new multicore capabilities.

While not yet equal to our ambition, the Deku betanets running on Tezos Ghostnet validated at least some of our hypotheses: we achieved 50,000 tps and block times under 3 seconds, with many low-hanging optimizations still on the table. Nevertheless we were still not on Mainnet.

Meanwhile, Smart Optimistic Rollups (SORU) will hit Mainnet in the Mumbai upgrade. The Deku team will instead join the other Marigold teams working with the core teams on Tezos' rollup systems, including protocol amendments, Octez upgrades, rollup kernels, and the upcoming Data Availability Layer.

What's next?

We are far from finished in our work to scaling Tezos. Deku was just one bet - rollups are still going strong. We are going to take what we've learned and apply it to the effort we started a long time ago with TORU.

SORU will bring massively higher throughput to Tezos, while also broadening the ecosystem with WebAssembly support. Additionally, members of the (former) Deku team will collaborate with Nomadic Labs and TriliTech on the design of an opt-in low-latency mode for SORU. The first version will target a full EVM-compatible blockchain running as a rollup. This is exciting work - a low-latency rollup would achieve much of the vision for Deku while inheriting the greater security guarantees provided by rollups and first-class support from the Layer 1 protocol (something only offered on Tezos!). More on this project in a follow up blog-post!

Pruning and rebalancing in response to change are core to the value proposition of Tezos. The pruning is always painful, but we are excited and hopeful about the growth it will bring to other areas. Stay tuned - the road ahead looks bright.

‍

‍

If you want to know more about Marigold, please follow us on social media (Twitter, Reddit, Linkedin)!