Giga: the Path to 50x Improvements on the EVM
Over the past few years, we have built up one core thesis - the EVM is here to stay. Nearly every crypto native developer uses the EVM, and the EVM has built a moat through the tooling, mindshare, and community that has built up around it.
Although the EVM is the backbone of existing smart contracts, it wasn’t built with performance in mind. Inefficiency was tolerable a decade ago, when smart contracts were a novel concept. Smart contract activity is increasing at an exponential pace now, and the EVM faces an existential scalability challenge.
Developers are forced to build EVM smart contracts in an environment that only supports around 100 transactions per second [1], which cripples the design space that they’re able to explore. Consequently, the developer experience of building on the EVM is fundamentally inferior to building a web2 application, which doesn’t run into the same restrictions on throughput.
This gap needs to be bridged, and developers should be able to build EVM applications while getting web2 levels of throughput.
Sei’s mission is to scale the EVM
Sei v2, the first parallelized EVM, went live on the mainnet beta in July 2024. Since then, it’s hit several milestones—surpassing $250M of TVL, reaching 400k daily active addresses, onboarding 150 projects on mainnet, and servicing 6.5M total users.
With parallelized execution, fast consensus, and optimized storage, Sei has created the highest performance EVM ever, allowing developers to scale while staying fully compatible with Ethereum’s infrastructure.
Unfortunately, even this isn’t enough to support web2 level scale. Applications such as Google require processing 100,000 complex transactions per second. Currently, no blockchain is able to support that level of traffic. This prevents sophisticated applications from getting built on-chain, and will become the biggest challenge in future growth of on-chain adoption.
The next natural step is to rebuild the EVM from the ground up, with a relentless focus on performance. This will involve creating a new EVM client from scratch, and completely changing the way that consensus and storage work on Sei.
Our next focus is to scale Sei to 5 gigagas per second, creating the most performant EVM blockchain ever.
With these changes, Sei will become the highest throughput EVM chain in the industry, getting 50x greater throughput than anything that is available on mainnet.
Gigagas can be described as the following (Paradigm):
Performance is typically measured by the metric "Transactions Per Second" (TPS). Particularly for Ethereum and other EVM blockchains, a more nuanced and perhaps accurate measure is "gas per second." This metric reflects the amount of computational work that the network can handle every second, where "gas" is a unit that measures the computational effort required to execute operations like transactions or smart contracts.
Standardizing around gas per second as a performance metric allows for a clearer understanding of a blockchain's capacity and efficiency. It also helps in assessing the cost implications on the system, safeguarding against potential Denial of Service (DOS) attacks that could exploit less nuanced measurements. This metric helps compare the performance across different Ethereum Virtual Machine (EVM) compatible chains.
We’re committed to achieving 5 gigagas per second while preserving the core functionality from Sei v2. This will allow developers to build EVM applications that are able to achieve web2 performance.
Roadmap
The Sei Foundation has already directed network stakeholders to organize this performance improvement initiative into three distinct workstreams; Consensus, Storage, and Execution.
Execution Workstream
To achieve 5 gigagas/sec, inefficient operations being performed as part of execution need to be minimized. We aim to create a new EVM client that rebuilds the EVM from the ground up, squeezing out performance at every level of the execution lifecycle.
- Performance-Optimized Client:
Sei will rebuild the EVM, and deliver a new client optimized for speed, pushing the bounds of the EVM. Sei will also incorporate ahead-of-time compilation, allowing the client to ensure high-performance execution of smart contracts, finely tuning memory management and computational tasks.
- Intelligent Parallelization:
Sei will enhance parallel execution by predicting and extracting transaction dependencies through bytecode analysis. This allows non-conflicting transactions to run in parallel while still maintaining optimistic parallelization, boosting throughput while preserving state consistency.
- New Encoding Format:
Sei will introduce a new encoding format designed to minimize processing overhead. Custom binary serialization reduces message sizes and parsing times with fixed-size fields and simplified schemas that accelerate both encoding and decoding.
Consensus Workstream
To hit 5 gigagas/sec, we need a consensus mechanism that optimizes for speed without compromising security. By cutting out the bottlenecks in transaction ordering and execution, Sei can guarantee that the network keeps moving, fast.
- Asynchronous Execution:
Traditionally, blockchains have tight coupling between consensus and execution - consensus cannot be achieved without executing the block and getting a state root. Sei is working to break this by decoupling consensus on the ordering of transactions from execution. Once there is consensus on the ordering of transactions, generating a state root is deterministic - a given ordering can only result in one state root via execution. As a result, finalization of a block does not wait for validators to execute transactions, allowing for higher throughput and reduced latency.
- Multiple Concurrent Proposers
Blockchains typically have one block producer per voting round. Sei will introduce multiple concurrent proposers, eventually allowing any proposer in the validator network to propose transactions that are included in the current slot. This makes it much harder to censor transactions, and helps with overall network liveness, since one proposer going down will have a much lower impact on the overall network.
- New Data Propagation Layer:
Sei will break down and distribute transaction data more efficiently across the network to reduce latency and increase throughput. This streamlining of data flow speeds up block finalization and improves the network's ability to handle higher transaction volumes.
Storage Workstream
Handling a massive amount of transactions means we have to continuously optimize storage from the ground up. Sei’s improvements make sure we can handle all that data without missing a beat.
- Asynchronous State Root Generation:
Sei will introduce asynchronous state root generation to resolve performance bottlenecks caused by synchronous Merkle tree updates. Offloading updates to parallel threads allows transaction processing to continue uninterrupted while state commitments are updated in the background and overall throughput improves.
- Asynchronous Disk Writes:
Building on the novel asynchronous disk writes from Sei v2, Sei will further reduce execution time with in-memory caching and batch disk operations. Write ahead logs (WALs) ensure data integrity, while asynchronous disk flushes prevent I/O wait times from slowing transaction processing. We are also integrating io_uring to optimize asynchronous I/O and further reduce latency while improving throughput in the storage layer.
- Optimized KV Database for Reads:
Sei will optimize its key-value database for faster data retrieval using read-optimized structures. Techniques such as smart caching and preloading strategies will reduce latency and enhance performance.
Timelines
At the direction of network stakeholders, Sei Labs has already started on each of these work streams and expects to see major progress by the end of Q1 2025.
Developer Office Hours
Join Sei Foundation Dev Rel, and Sei Labs lead engineers for a special Office Hours in the Sei Discord.
Exploring the Giga Roadmap - Thursday, December 19th at 1pm EST
Join us
Sei Labs is working with network stakeholders on the most ambitious problems in crypto, and we’re aggressively expanding our research and engineering teams - if any of the challenges described above are interesting to you, please reach out to us at https://jobs.lever.co/SeiLabs