In June, we unveiled Protocol, restructuring the Ethereum Foundation’s research & development teams to more effectively align with our current strategic objectives, Scale L1, Scale Blobs, and Enhance UX without sacrificing our dedication to Ethereum’s security and resilience.
In the upcoming weeks, we will share updates on each work stream, discussing their ongoing advancements, new initiatives, unresolved questions, and chances for collaboration. We commence today with Scale L1—anticipate follow-ups regarding Scale Blobs and Enhance UX shortly!
TL;DR
- Marius van der Wijden teamed up with Ansgar Dietrichs and Tim Beiko to co-lead Scale L1
- Mainnet’s gas cap raised to 45M following Berlinterop, marking a preliminary step on the journey to 100M gas and beyond
- All principal execution layer clients deployed Pre-Merge History Expiry, significantly diminishing node disk storage
- Block-Level Access Lists (BALs) are being evaluated as a feature for Glamsterdam
- Compute & state benchmarking efforts are in progress to enhance EVM resource pricing and performance limitations
- The route to zkEVM real-time proving is becoming clearer, with the prototyping of a ZK-based attester client active
- We are still seeking a Performance Engineering Lead: applications close on Aug 10
Geth-ing Serious About L1 Scaling
Scaling Ethereum necessitates balancing lofty concepts with engineering realism. To assist us in this endeavor, we’ve appointed Marius van der Wijden as co-lead for Scale L1 alongside Ansgar Dietrichs and Tim Beiko.
Marius’s vast engineering background on Geth, combined with his dedication to protocol security, makes him an ideal candidate to align our scaling strategy with Ethereum’s limitations.
Together, Ansgar, Marius, and Tim have established a series of pivotal initiatives that will enable us to Scale L1 as swiftly as possible.
Towards a 100M Mainnet Gas Limit
Our immediate objective is to safely expand Ethereum’s mainnet gas limit to 100M per block. Parithosh Jayanthi, with strong support from Nethermind’s PerfNet team, is spearheading our work advancing through each stepwise increase.
At the recent Berlinterop event, client teams dramatically enhanced their worst-case performance metrics, facilitating the recent increase to 45M gas—a first milestone on the trajectory toward 100M gas and beyond!
Moreover, client fortification has become a crucial component of the 100M Gas initiative. The Pectra upgrade deployment revealed several challenges attributed to network instability. It is essential to ensure clients remain resilient as throughput scales up, even if the network temporarily fails to achieve finality.
History Expiry
The History Expiry initiative, spearheaded by Matt Garnett, diminishes the historical data footprint of Ethereum nodes. The recent launch of Partial History Expiry eliminated pre-Merge historical data, allowing full nodes to reclaim roughly 300–500 GB of storage. This ensures they can operate efficiently with a 2TB disk.
Building on this, we are currently working on Rolling History Expiry, which will consistently prune historical data beyond a designated retention period. This will maintain nodes’ storage requirements manageable, even as Ethereum grows.
Block-Level Access Lists
Block-Level Access Lists (BALs), endorsed by Toni Wahrstaetter, are emerging as a prime candidate for integration in the Glamsterdam upgrade. BALs offer multiple essential advantages:
- Enable concurrent transaction execution within blocks.
- Facilitate simultaneous computation of state roots, greatly accelerating block processing.
- Permit preloading of necessary state at the commencement of block execution, optimizing disk access patterns.
- Enhance overall node synchronization efficiency, benefiting both new and archival nodes.
These enhancements collectively boost Ethereum’s ability to effectively manage higher gas limits and expedite block processing.
Benchmarking & Pricing
A persistent challenge in scaling Ethereum is aligning the gas expenses of EVM operations with their computational cost. The functionality of worst-case edge scenarios currently limits network throughput.
By enhancing benchmarking infrastructure and adjusting prices on operations that cannot be optimized by clients, we can achieve more consistent block execution times. If we bridge the gap between the worst and average case blocks, we can subsequently raise the gas limits accordingly.
Ansgar Dietrichs leads initiatives directed at targeted benchmarking and engineering measures, guided directly by PerfNet’s extensive benchmarking, to pinpoint and rectify compute-heavy bottlenecks. Significant advancements have already been achieved post-Berlinterop, particularly in managing worst-case compute situations.
Simultaneously, Carlos Pérez leads Bloatnet: an initiative aimed at benchmarking and optimizing state performance. This entails evaluating node performance under scenarios with state sizes twice the current mainnet and gas limits approaching 100–150M, to directly inform both repricing strategies and client improvements.
Both endeavors will contribute to Glamsterdam EIP proposals to standardize resource costs across operations, fostering further L1 scaling.
zkEVM Attester Client
Currently, Ethereum nodes carry out all transactions in a block upon receiving it. This is computationally intensive. To mitigate this computational expenditure, Ethereum clients could instead validate a zk proof of the block’s execution. For this to be feasible, proofs of the block need to be produced in real-time, which we are getting closer and closer to.
“““html
Kevaundray Wedderburn is spearheading efforts on a zkEVM attester client that presumes the availability of real-time proofs and employs them to carry out its validator responsibilities.
Once the prototype is prepared for mainnet deployment, it will be introduced as an optional verification system. We anticipate a select group of nodes will embrace this over the next year, enabling us to develop trust in its robustness and safety.
Subsequently, Ethereum nodes can progressively shift to zk-based validation, ultimately becoming the standard. At that juncture, L1’s gas capacity could rise significantly — potentially go beast mode!
RPC Performance & Hiring
As throughput expands, various node categories (execution, consensus, RPC) encounter unique hurdles. RPC nodes, in particular, face increased strain as they manage extensive historical and real-time state inquiries.
Within the organization, the EF’s Geth and PandaOps teams are diligently investigating the best configurations for various node types. We foresee the significance of this growing in the next few years and aspire to enhance our knowledge in this area.
In light of this, we’re actively recruiting for a Performance Engineering Lead. Applications conclude August 10. If you’re as eager as we are about enhancing the L1, we would be thrilled to connect with you!
Source link
“`
