Besu

Boosting Besu Performance: 2025 Accomplishments and 2026 Roadmap

the Besu Maintainers |
Boosting Besu Performance: 2025 Accomplishments and 2026 Roadmap

Looking back to the beginning of 2025, improving performance in Besu was one of our major goals for the year. In this blog post we want to share what specific improvements we’ve made towards that goal in 2025, as well as some of the directions we’ll be focusing on for the year to come. 

Introduction

In 2025, our main focus for Besu performance has been on securing the 60 million mainnet gas limit and beyond, while keeping execution times safely within limits across all worst-case scenarios. A key part of this effort involved close collaboration with the Nethermind team, who led the research on identifying and reproducing these worst cases.

We made substantial improvements to ensure Besu can handle the most demanding workloads, enhanced our benchmarking framework to accurately reproduce and validate fixes for critical edge cases, and boosted sync performance to help nodes reach the head of the chain faster. Alongside these, we delivered multiple RPC and block processing optimizations, strengthening both worst-case resilience and general performance.

From 30M to 60M gas (Improve worst case scenarios)

At the beginning of 2025, the Ethereum mainnet block gas limit was 30 million gas per block. Raising this gas limit is part of the broader effort to scale mainnet Ethereum. All the Execution Layer (EL) client teams need to agree on raising this limit, and an important pre-requisite is to ensure that all clients can safely process blocks of this size within the time allotted for block processing, currently 3 seconds. For this, client teams need to consider worst case scenarios.

Thanks in large part to the Nethermind team, there are benchmarks that are shared among all the EL client teams to pinpoint areas that require development focus.

At the Berlin interop in June 2025, the client teams together made the decision to increase the mainnet block gas limit to 45 million gas per block. The ModExp precompile was a focus area for a number of EL teams, including Besu. We made several improvements to ModExp in Besu:

The default block gas limit for Ethereum mainnet was raised to 45 million in Besu 25.7.0. 

Fusaka is again raising the Ethereum mainnet block gas limit from 45 to 60 million gas per block. Testnets have been operating with a limit of 60 million for the last few months, with Hoodi the last to update in May 2025. 

To ensure any block with this gas limit can be processed within the available time, the Besu Performance team worked to improve the performance of a number of Opcodes, with a specific focus on worst cases: 

And precompiles:

Table showing Opcodes and Precompiles improvements, with max improvement being the improvement for known worst cases:

The default block gas limit for Ethereum mainnet was raised to 60 million in Besu 25.11.0. 

A 60 million block gas limit is a fine achievement, but it is not the final destination. Besu along with the other EL clients will continue working towards being able to further increase this number. 

Improve sync performance

We’ve made major improvements to Besu’s sync performance, especially on Ethereum mainnet where syncing takes the longest. These enhancements will benefit anyone running Besu, from stakers to enterprise users.

Key optimizations:

This graph shows Ethereum mainnet sync times (based on controlled virtual machine tests from Besu 2025 release burn-ins). Total sync time is shown, as well as the time taken for the different stages that make up sync. From this we can see how the cumulative sync optimizations have reduced the time taken for the different stages of syncing. The biggest improvements can be seen in two stages: Chain download and Snap DB heal. 

Sync times still depend on factors like hardware, network speed, and peer quality, but these results highlight the significant improvements made in 2025.

Improve RPC performance

RPC performance is important for a lot of use cases. We’ve made a few improvements to Besu RPC performance in 2025 including:

Improve block processing and general performance

Additionally, we have made a number of updates that enhance general block processing performance and a few improvements that are more general in nature.

Leading up to Fusaka, one of the tests performed was a long non-finality event. This enabled all the client teams to see how the clients performed in this scenario and led to these updates:

Improve benchmarking (EvmTool and JMH benchmarks)

Alongside performance improvements, we also made improvements to Besu benchmarking tools. Benchmarking of Opcodes (JMH) and precompiles (EvmTool) is critical for evaluating performance improvements, so this work enhances the effectiveness of the team. 

What's next for 2026

Onwards to 100 million: Besu, along with other EL clients, will continue to benchmark, observe and improve performance with the aim to raise the Ethereum mainnet block gas limit further.

Bonsai archive: we think there is an opportunity for L2s to utilize Bonsai archive, which will give us an environment where we can get feedback rapidly and make performance improvements. We want to get Bonsai archive performant enough to serve mainnet users but it isn’t there yet.

More sync improvements: while we have made some improvements to sync time and resource usage, we think there is more to do here. 

Building on the work done for MOD operations, we want to revamp the Besu EVM to have a UInt256 type on the stack that has its own more performant implementation for arithmetic operations, rather than using Java’s BigInteger which is not designed for performant UInt256 types. We are also productionizing the improvements we’re working on in Tuweni, creating a new version, which will potentially improve performance anywhere we use Bytes in Besu.

Conclusion

We hope you’ve enjoyed this look back at our performance journey in 2025 - a year filled with significant strides in Besu's performance! As we move forward into 2026, our priorities are clear: tackling scalability, boosting performance, and creating tools that empower the Ethereum ecosystem.

We’re committed to building on the progress we’ve made so far, and we’d love for you to join the conversation as we build a faster, smarter, and more efficient Ethereum client for the future. Connect with us on Discord. Whether it's ideas for new features, performance improvements, or fresh approaches to persistent challenges, your input helps shape the next phase of Besu.

Acknowledgements

We would like to thank the Nethermind team for their significant contributions in identifying and building the infrastructure to uncover worst-case scenarios, which has been instrumental in guiding our performance work.

We also extend our appreciation to the Ethereum Foundation team for their support in testing and validating our implementation against a known worst case, helping ensure the robustness and correctness of the improvements.
Back to all blog posts