Try   HackMD

Portal Network Implementers Call Notes

May 26th, 2025

1. Team update

  • General
    • Low attendance due to public holiday
    • Teams are encouraged to update/ catchup via notes later
  • Update on Trin & Glados by Milos
    • Progressed on ephemeral headers, ephemeral history bridge nearly finished (pending PR merge)
    • Client-side refactored how content is handled
    • Glados update:
      • Added Samba
      • Renamed Fluffy to Nimbus Portal
      • Refactored audit dashboard to allow display of random audits (pending PR merge)
      • Added page with recent transfer failures
    • Hive tests have been down for the last week — investigation ongoing
  • Update on Samba by Meld
    • Implemented validation for pre-merge block bodies and receipts
    • Added validation for merge-to-Capella headers, enabling full validation
    • Temporary implementation to support beaconStore compatibility (able to run on the latest version of Hive)
    • Updated the putContent endpoint to match latest spec changes
    • Fixed bugs related to data path issues and database initialization problems
  • Update on Shisui by Zhen Peng
    • Completed some code refactoring
    • Added support for ephemeral type for findContent
    • Addressed performance/stability issues, incl. concurrency bugs and memory leaks
    • Added eth api support

May 19th, 2025 - Call #56

1. Team update

  • Update by Piper

    • Preparing a co-authored document with Jason that outlines Portal’s role in the L1 scaling roadmap
    • Inviting teams to review and provide feedback prior to the public release

  • Update on Trin & Glados by Michael

    • Internal refactoring
    • Added new docker image for trin-execution
    • Moved consensus-related constants into ethportal-api
    • Added trace offer endpoint which support multiple content keys
    • Deployed e2hs bridges for data bridging
    • Glados: UTP transfer failures are now visible, aiding network health tracking

  • Update on Ultralight by ScottyPoi

    • Boot nodes redeployed with some issues to be fixed
    • Merged updated SSZ container for HistoricalSummariesStateProof, resolving Hive test failures

  • Update on Nimbus Portal Client (formerly Fluffy) by Kim

    • Added access from History network to historical summaries
    • Updated state bridge for more efficient gossiping
    • Fixed a bug in FindContent responses
    • Working on rate-limiting to better monitor gossip success rates
    • Renaming from "Fluffy" to "Nimbus Portal Client" underway across code and infrastructure
      • Dual support ("f" as Fluffy and "n" as Nimbus Portal) for transition phase in ENRs

  • Update on Shisui by Qi Zhou

    • Performance improvements via refactoring discv5 to pipeline the reloop and dispatch loop asynchronously
    • Fixed memory leak issues and block boundary check error
    • Refactored validation logic to reduce redundancy
    • Implemented support for retrieving ephemeral header from storage (code currently under review)

  • Update on Samba by Derek

    • Implemented the protocol versioning incl. accept codes and utp size prefix
    • Fixed an issue in history offer logic with large transfers
    • Integration with Besu
      • Improved logs and added parameter options for running Samba inside Besu
      • Made all Portal History JSON-RPC calls available as method calls for Java projects
      • Basic Besu node now operational with Samba integration
    • Plan to implement the beaconStore endpoint and Capella validation this week

2. Discussion topics

2.1 Beacon Spec Update

  • PR was submitted based on last week’s discussion
  • Open for team review before merging

2.2 Ephemeral Header Gossiping Mechanism Discussion

  • The problem: Current doc suggests neighborhood gossip for ephemeral headers, but this may
    • Restrict propagation to nodes near the content ID
    • Risk incomplete network coverage
  • Team consensus: Random gossip preferred over neighborhood gossip
    • Neighborhood gossip filters peers based on radius related to content ID, which is irrelevant for ephemeral headers
    • Random gossip can be adjusted to prioritize peers closer to the sender, but should spread widely to ensure all nodes receive the headers
  • Action item
    • Submit PR to update the spec to clarify random gossip as the mechanism for ephemeral headers

May 12th, 2025 - Call #55

1. Team update

  • Update on Trin & Glados by Milos

    • Implemented support for Pectra types on the Beacon network
    • Started adding initial support for Sepolia
    • Ongoing work on implementing the bridge that will follow the chain head and generate e2hs files
    • Glados: Updated areaGraph for cleaner visualization

  • Update on Fluffy by Kim

    • Fixed a bug in HistoricalSummaries storage in beacon_db
    • Explored experimental support for providing HistoricalSummaries as an Oracle to the History network (currently in testing, not merged yet)
    • Implemented several performance improvements on State network for synchronous data processing
    • Removed deprecated neighborhoodGossip logic
    • Proposed PR to clarify that HistoricalSummaries must be from finalized beacon state
      • Nodes should only respond to HistoricalSummaries requests if they have data for the requested epoch or newer
      • Otherwise, nodes should rely with an empty response (pending further clarification in the specs)

  • Update on Ultralight by ScottyPoi

    • Progress on HistoricalSummaries and updating SSZ types
    • Continued implementation of ephemeral headers
    • Started implementing typing and distance functions of Head-MPT State Network
    • Explored support for Sepolia recognized challenges in supporting multiple networks (mainnet & testnet) simultaneously

  • Update on Shisui by Qi Zhou

    • Fixed Hive test failures by addressing missing validation for historical network in findContent, especially for post-Shanghai data
    • Addressed a lightclient error due to an JSON-RPC typo
    • Split the ephemeral header PR into smaller parts for easier review
      • Implemented & merged the storage part
      • Remaining features will be implemented in the coming weeks
    • Created a new benchmark to compare performance between Trin vs Shisui using the same payload
      • Showed that discv5 from Geth implementation has significant overhead and slower compared raw uTP implementation
      • The benchmark tests are currently ad hoc, but can potentially be modularized and reused across clients for performance debugging/ validation

  • Update on Samba by Meld

    • Completed the accumulator functionality
    • Ongoing work on protocol versioning and accept code
    • Besu integration encountered some blockers (actively being addressed)

2. Discussion topics

2.1 LightClientUpdatesByRange Behavior Change Proposal

  • Current Spec: If a node is missing any update in a requested epoch range, it must not reply at all
  • Proposed Change: Nodes should reply with available updates starting from the earliest in the range, even if incomplete
    • Rationale: Syncing nodes benefit from partial progress (getting closer to chain head)
    • Impact: Able to avoids scenarios where nodes withhold all updates due to missing only the latest update
  • Team Consensus
    • No objections; PR to be submitted to formalize the changes

2.2 LightClientOptimisticUpdate Request/ Response Logic Optimization

  • Current Spec: Nodes request optimistic updates based on local clock’s current slot
  • Proposed Changes
    • Request Logic: Nodes should request the first missing slot (not just latest) if behind
    • Response Logic: Peers should always reply with their latest optimistic update, even if it's newer than the requested one
      • Example: If a peer stores only slot 100 but receives a request for slot 99, it should reply with slot 100
  • Key discussion points
    • Some clients (eg. Fluffy, Ultralight) store only the latest optimistic update, so replying with older ones isn't feasible
    • There's consensus to avoid introducing new content keys to existing data structure unless absolutely necessary to avoid complexity and redundancy
  • Team Consensus
    • General agreement reached; PR to be submitted to formalize the changes

2.3 Config Standardization for Testnets

  • Unified Config Approach
    • Proposed a unified YAML config format combining CL, EL, and Portal-specific fields to simplify setup
    • Can support presets or custom paths
  • Automation & Maintenance
    • Generate configs from upstream (CL/EL repo) to minimize manual updates
    • Tradeoff between simplified UX vs maintenance overhead
      • Possible to automate yaml generation to reduce maintenance burden

May 5th, 2025 - Call #54

1. Team update

  • Update on Trin & Glados by Kolby

    • Bug fixes & improvements
      • Fixed a bug related to gossiping content to nodes outside their radius
      • Replaced hardcoded fork activations with dynamic network config
      • Integrated Trin as a library, removing slow HTTP JSON-RPC and redundant tests to improve performance and dev experience
    • Enabled HistoricalSummaries for the Beacon Bridge
    • Added Deneb header proof type, generation code, and Deneb header validation
    • Added Electra types and created header proof using BeaconBlockElectra
    • In progress
      • PR to add HistoricalSummariesWithProof for Electra
      • PR to generate e2hs files
    • Glados：Added internal transfer failure chart

  • Update on Fluffy by Kim

    • Bug fixes & improvements
      • Improved eth_createAccessList to return access lists even when the EVM execution reverts
      • Ensured only a single instance of AsyncEvm created and shared across API handlers
      • Added architecture docs to the Fluffy guide
    • Tested & activiated block proof tests
    • Implemented & tested the adjustment of historical_summaries proof for Electra

  • Update on Samba by Meld

    • Continued implementing discv5 and awaiting upstream changes from Besu in the discv5 library
    • Continued implementing history endpoints, eg. historyTraceGetContent, historyRecursiveFindNodes
    • Nearly done with accumulator integration
    • Working toward release a Samba lib to be integrated into Besu client
    • Commit to complete the History network checklist soon

  • Update on Ultralight by acolytec3

    • Focused on implementing ephemeral headers, with progress on validation and purging old offers
    • Will work on historical summaries updates after current ephemeral headers implementation

  • Update on Shisui by Grapebaba

    • Converted ExecutionBlockProof from list to vectors
    • Added recursiveFindNodes in putContent logic
    • Fixed bug in ping extension upgrades
    • Working on ephemeral headers implementation

2. Discussion topics

2.1 History Network Update

2.2 Running Portal in Kurtosis

  • Goal
    • Enable Portal Network clients to run inside Kurtosis, for better devnet/ testnet support and integration testing
  • Initial target
    • Run local, small-scale devnets to experiment with setup, discover limitations, and determine necessary client-side changes
  • Topics discussed
    • Networking design consideration
      • Need for large super-nodes vs many small nodes in testing
      • Use of ephemeral devnets with isolated protocol IDs to avoid collisions
      • Flagged potential issues with ephemeral headers as client teams are still finalizing their implementation
    • Bridgework: Need mechanisms to insert historical data into the devnets for testing purposes
    • Protocol ID/ Chain ID: Discussed potentially using chain IDs for Kurtosis devnets, and how to map them to protocol IDs in Portal clients
    • Config consideration
      • Reuse existing EL/CL configs
      • Add Portal-specific settings via standardized spec file (shared across clients) and optional client-specific extensions
  • Collaboration
    • Kolby (Trin) - Will act as the point of contact from Portal devs
    • Kim (Fluffy) - Expressed interest to support Fluffy within Kurtosis
    • Barnabas (ethPandaOps) - Will provide Kurtosis integration support
    • Other teams/ clients are welcome and encouraged to contribute
  • Action items
    • Coordinate PRs and communication in ethpandaops repo
    • Create a tracking issue for Kurtosis integration blockers and client-side requirements to support Kurtosis devnet
    • Explore minimal devnet setups for initial testing

Apr 28th, 2025 - Call #53

1. Team update

  • Update on Trin & Glados by Mike

    • Added AcceptCodes to census and bridge metrics for better tracking
    • Beacon network is now required when History network is enabled
    • Simplified header proof generation code to avoid duplication
    • Changed ExecutionBlockProof from list to vector
    • PR to enable HistoricalSummaries bridge and prepare for future testnet support
    • Glados
      • Fixed a broken spec link
      • Added protocol version support by client graph

  • Update on Fluffy by Kim

    • Added protocol versioning support
    • Implemented ephemeral header encoding/ decoding (disabled for now until database update work is complete)
    • Implemented offer cache to hold content IDs of recent offers
    • Added metrics to count offer accept codes (Grafana dashboard will be updated accordingly)
    • Adjusted tests to support HeaderWithProof using SSZ vectors for Capella & Deneb
      • Adjust tests for supporting HeaderWithProof updated version with SSZ vector instead of list
      • Adjusted generation of test vectors and enabled validation accordingly
      • Updated portal-spec-tests test vectors
    • Revisited the PR of adding the HistoricalSummariesWithProof endpoint in Nimbus consensus client (expected to be merged soon)
    • PR to add HistoricalSummariesWithProof for Electra

  • Update on Shisui by Qi Zhou

    • Added ephemeral types support and test vectors
    • Improved performance with multi-threading validation, reducing rejection rates from ~66% to <40%
    • Continued pushing for integration with Geth but delayed due to recent Pectra upgrade priority

  • Update on Samba by Meld

    • Continued internal adjustements to integrate Samba with Besu
    • Working on completing all discv5 endpoints in the implementation
    • Working towards deploying a Besu-Samba integrated instance
    • Plan to differentiate client version string for tracking Besu-Samba integrated version and Samba standalone version

  • Update on Ultralight by acolytec3

    • Merged discv5 fixes into ChainSafe implementation (not yet deployed to boot nodes)
    • DevOps continued improving boot node stability through VM placement updates
    • Plan to update History network container for Deneb headers
    • Start to review ephemeral header gossip logic, and prepare to active the already-implemented basic types

2. Discussion topics

2.1 History Network Deployment & Monitoring

  • Goal
    • Finalize History network deployment to unblock State network development
    • Primary focus is to get the ephemeral head fully live and deployed
  • Action items
    • Create a meta-issue (led by Kolby) to track remaining tasks across client teams
    • Glados monitoring
      • Separate infra needed for tracking ephemeral head data
      • Minimal Beacon network monitoring to verify gossip availability
  • Next step
    • Client teams to contribute status updates via the meta-issue

2.2 Pectra Upgrade & Testnet Integration

  • Pectra Fork Readiness
    • Most Pectra-related updates are client-specific
    • No major cross-team sync needed now
    • Hive tests can be updated after the Pectra fork ships
  • Future Fork Testnet Integration
    • Integrate Portal into testnets before future forks to avoid post-fork delays
    • Prioritize Sepolia as the primary testnet for Portal-related work, as it contains pre-merge data

Apr 21st, 2025 - Call #52

1. Team update

  • Update on Trin & Glados by Milos

    • Improved bridge functionality
      • Supported e2hs files for gossiping history
      • Updated all bridges to use census / offer, enabling direct offer instead of gossip logic
    • Added ephemeral header content type definition (not active yet)
    • Ongoing PRs on Deneb proof generation & validation, and Electra consensus types
    • Glados: Implemented protocol version tracking for peers
      • Trin, Fluffy, and Shisui now all officially supports v1

  • Update on Ultralight by ScottyPoi

    • v1 is implemented but not publicly updated yet
    • Continued working with DevOps to revamp boot node setup
    • Awaiting ChainSafe to release an update that includes the bug fix discussed last week

  • Update on Shisui by Qi Zhou

    • Profiling ongoing to resolve performance issues (especially in comparison to Trin)
    • Implemented new RPC endpoints to support missing JSON-RPC APIs
    • Added caching for the highest compatible version to speed up lookups

  • Update on Samba by Meld

    • Focused on integration with Besu
      • Currently managed to run Samba inside Besu locally, but still need further polishing & adjustments
      • Aim to support Samba both within Besu and as a standalone solution
      • Working toward deployment / testing of Samba in Besu across versions to evaluate performance and behavior

2. Discussion: Head-MPT State Network Spec

  • Key problem
    • How to update slices of state for nodes whose section had no updates in a given block
  • Proposed Solution: Introduce block-level MPT diffs as a formalized content type
    • Each block contains a diff from the previous block's state
    • Allow nodes to update and re-anchor their data to the latest MPT root, even if their slice didn’t change
    • Enable a lightweight initial rollout - nodes can sync & operate without full account state storage initally
  • Sync Mechanism
    • New nodes can sync from the current trie and fast-forward using accumulated diffs
    • Trie diffs will be the main data gossiped on the network
    • While the protocol supports querying individual trie nodes, gossiping individual nodes isn't the focus
  • Implementation Roadmap for client teams
    • 1st priority: Finalize ephemeral header support in the History Network
    • 2nd priority: Push the State Network as close to the chain head as possible
    • Afterwards: Define block-level diff data type, and implement diff generation & propagation
  • Relevant PR/ issue: https://github.com/ethereum/portal-network-specs/pull/389

Apr 14th, 2025 - Call #51

1. Team update

  • Update on Trin & Glados by Kolby and Mike

    • Implemented Protocol versioning & v1
    • Fixed bug in JSON-RPC result for the offer method
    • Performance improvements of e2hs writer
      • 44× speedup for converting era1 to e2hs files
      • 6× speedup for converting era to e2hs files
      • Now pre-merge e2hs files can be generated in ~10s and post-merge e2hs files in ~5 minutes
    • Plan to run Bridges fully on e2hs files with public endpoints soon
    • Discussion on adopting e2hs format as a standard post-merge history format
      • Nearly ready for broader feedback & adoption once the full repo is generated and publicly available
    • Glados update
      • Finished schema change last week, allowing UTP health visibility
      • Deploy sync auditing this week
      • Plan to work on protocol versioning visibility next week

  • Update on Ultralight by ScottyPoi

    • v1 protocol version merged and unit tested
    • Fixed bugs in discv5 test suite
      • 3 PRs submitted to ChainSafe’s discv5 implementation to fix ID encoding, challenge re-sending, and byte/int conversion bugs
    • Continued revamping public node setup with devops team for better reliability

  • Update on Samba by Meld

    • First call with Besu team to align on plugin use and integration
      • Need to change certain discv5 endpoints from Besu side to integrate with Samba
    • Working in parallel on
      • Endpoint completion
      • Design shift from standalone to plugin-compatible mode

  • Update on Shisui by Qi Zhou

    • Implemented protocol versioning v1 and updated offer interface
    • Fixed a panic issue related to closing connections on errors
    • Addressing performance issue
      • Working on utp rate limiting to optimize gossip
      • Benchmarking ongoing to resolve performance gaps vs Trin

2. Discussion topics

2.1 Head-MPT State Network Spec

  • Design overview
    • The Head-MPT State Network will initially be a separate sub-network (not merged with the existing state network) for faster iteration, with potential future integration
  • Scope
    • Focused on the last 256 blocks, with nodes anchoring content via state roots or block hashes
  • Content Types
    • Gossiped data includes account trie nodes, bytecodes, and storage trie nodes
    • Each node anchors content to the state root
  • Gossip mechanism
    • Bridges split the trie diff into 256 subtries and gossip each as a key-value pair
    • Nodes store and update assigned subtrie based on their node ID
    • Nodes retain their subtrie for the last 256 blocks, discarding older ones
  • Querying & sync
    • Direct leaf queries are supported with proofs
    • Nodes can fetch missing individual trie nodes when bootstrapping or syncing
  • Implementation Phasing
    • Start with account trie only
    • Contract trie support will follow after validation
  • Next step
    • Client teams are encourated to review the PR and provide feedback
    • Deeper discussion scheduled for next week’s call (or the week after if there is light attendance next week due to Easter)
  • Relevant PR/ issue: https://github.com/ethereum/portal-network-specs/pull/389

Apr 7th, 2025 - Call #50

1. Team update

  • Update by Piper

    • Working on history expiry coordination & organization
    • In early onboarding phase for history state spec

  • Update on Trin & Glados by Kolby

    • e2hs update
      • Fixed an invalid BlockIndex type in e2hs
      • Pre-merge e2hs files generated & uploaded, except some minor bugs
      • Once fixed, e2hs file will replace era1, with plan to sync up to the chain head minus ~27h
    • Fixed a bug where the discv5 ENR cache was to small to handle 500+ or more active peers by scaling the cache size to match the max number of sessions
      • A cleaner upstream fix coming soon via the discv5 lib
    • Glados
      • Updated Glados to support ethportal-api 0.6.0 which incl. ping extension
      • Fixing the census issue, which now works even without online nodes

  • Update on Fluffy by Kim

    • Updated ping endpoint to support ping extension
    • Worked on async EVM improvement and investigating segfault/ bugs in CI tests
    • Protocol versioning
      • Progressing on protocol versioning, but blocked by an unexpected compiler bug
      • Need to update discv5 for protocol versioning to properly handle ENR access during request/ response
        • Plan to store ENR in session cache for reliable access
    • Implemented the changes of accept codes and add varient size prefix
      • Investigating a clean solution to manage versioning info along with the encoding/ decoding process

  • Update on Ultralight by ScottyPoi

    • Working on protocol versioning, added initial support but v1 features not yet active
      • Current priority: Finish the v1 changes implementation
    • Progressed on history expiry support within EthereumJS team
    • Plan to revamp boot node setup, shifting from 2 overloaded VMs to multiple smaller VMs, for better performance and reliability
    • Merged a Tauri-based web interface project built on Ultralight by Justin, currently allowing basic RPC lookups

  • Update on Samba by Meld

    • Implemented SSZ union removal
    • Fixed a bug in the putContent JSON-RPC endpoint, which is related to the utp lib usage
    • Passed local Hive tests against Trin, still debugging remaining test failures
    • Plan to begin regular biweekly sync with Besu team for planning and coordination

  • Update on Shisui by Grapebaba

    • Merged PRs, incl. implementation on accept codes and protocol versioning
    • Implemented add size prefix in FindContent
    • Fixed the FindNodes response bug, due to packet too long
    • Added rate limiting policy, where gossips are dropped when resource usage exceeds the limit
    • Fixed discv5 whoareyou response issue in Geth, through in Hive tests still see issues when communicating with Fluffy nodes
      • The issue with Fluffy nodes
        • When two portal requests are sent simultaneously before a session is established, only the 1st succeeds and the 2nd fails
        • eg. Fluffy’s background Ping (for node revalidation) races with test-triggered FindContent, would cause one to fail
      • Potential solution
        • Queuing on Fluffy side: Queue the concurrent requests until the session is setup
        • Retry logic: Clients could retry dropped requests automatically
        • Test adjustment: Modify Hive test to better reflect real-world behavior
      • Ultralight related notes
        • Identified 3+ encoding-related bugs when previously debugging Hive test failures related to discv5
        • Not sure if these fixes would resolve the concurrent-request problem

2. Discussion topics

2.1 Next step for Bridges

  • Curent status
    • Bridges are mainly run by the Trin team currently
  • Technical updates
    • Transition to generate e2hs file (much simplified without the accumulator) to replace era1 file for the entire range
    • Plan to maintain two bridge types: ephemeral bridge (handle reorgs & complex edge cases) and e2hs bridge
    • Dropping support for the older era1 bridge
  • Next step proposal
    • Community-runned bridges
      • Consensus: Shift from team-operated to community-operated bridges
      • Next steps: Document clear setup steps for external parties to get involved easily

2.2 Hive test

  • Achieved improved test coverage this week, with additional Portal Hive test now passing successfully
  • Strong cross-team collaboration as client teams worked together on troubleshooting and debugging

2.3 Team presence at Interop event

  • Some members from Portal team will attend the event in Berlin in June

Mar 31st, 2025 - Call #49

1. Team update

  • Update on Trin & Glados by Kolby

    • Formalized the new e2hs file format to handle all EL blocks, which simplifies bridging process
    • Implemented e2hs, incl. adding reader & writer in e2store crate
    • Published e2store crates available for external projects to use
    • Fixed a utp packet handling bug, where 1st part of transfer is ignored by recipient if the STATE response to SYN is dropped
    • Cleaned up redundant test vectors
    • Glados: Now display header by numbers on the latest block status graph

  • Update on Ultralight by ScottyPoi

    • Added protocol versioning to ENRs and preparing for the PR on accept/decline code change
    • Updating EthereumJS to support EIP-4444 and drop day by integrating Ultralight as a dependency
    • Progressed on ephemeral headers implementation

  • Update on Samba by Meld

    • Implemented new endpoints but facing delays with UTP library (putContent issue) and accumulator integration
    • Restructuring code for better modularity to integrate with Besu client
      • Need further exploration on the primary use case of Samba by the Besu client (eg. JSON-RPC serviing vs full block sync)

  • Update on Shisui by Qi Zhou

    • Added support for protocol versioning and accept code implementation
    • Fixed a race condition bug and merged ping extension PR
    • Investigating the Discv5 whoareyou-challenge response issue and performance bottlenecks in Go implementation

2. Discussion topics

2.1 Protocol Versioning Implementation

  • Most of the client teams have started implementing the protocol versioning
  • Teams are encouraged to implement miniimal versioning to enable incremental updates and compatibility testing

2.2 Ephemeral Headers Spec Update

  • The problem
    • The initial assumption that all EL headers could be sourced from LightClientUpdates was invalidated as the CL derives tx/ withdrawal data via SSZ instead of RLP/ MPT
  • The spec is being updated to
    • Clarify limitation: EL headers can't be fully sourced from the Beacon Network
    • Define handling
      • Headers should be organized chronologically in payloads
      • Clients should retain headers during reorgs until chain validity is re-verified
      • Avoid unnecessary network requests that could case DDOS issue
      • Allow headers to propagate naturally across the network
    • Note edge cases: Notes added for scenarios like reorgs and data propagation
  • Next step
    • Team working on the implementation is encouraged to read the PR link for details
  • Relevant github issue/ PR

2.3 Shisui & Ultralight Connectivity Issue

  • The problem
    • Milos obversed intermittent failures when sending Find_Nodes request to Shisui and Ultralight nodes
  • Pattern of failure
    • Find_Nodes requests fail at non-zero distance
    • Requests with zero distance or simple ping work fine
    • The issue isn't consistently reproducible but happens frequently
  • Investigation
    • Initial debugging attempts suggest the session or message keys might be incorrect
    • Equivalent Hive tests are passing, ruling out protocol-level bugs
    • Shisui identified a potential race condition in node assignment logic and a fix has been proposed & under review
  • Next step
    • Team agree to collaborate via a live debug session, with Shisui deploying a test node for targeted troubleshooting
  • Relevant link

2.4 EIP-4444 Coordination Updates

  • Sepolia testnet: The initial May 1st drop date is now Sepolia testnet drop date, with efforts underway to align an EL client version ready for Sepolia so CL clients can test against
  • Mainnet drop date: The mainnet drop date is postponed to sync with Pectra upgrade, due to CL's dependency on pre-merge logs

2.5 State Network Progress

  • State network rollout has stalled due to
    • Gossip performance issue: Bottlenecks in propagating new state data across the network
    • Dependencies: Protocol versioning, accept codes must be finalized first to unblock state sync features
  • Action plan
    • Short-term goal: Focus on shipping a client version that can sync state even if it's slighly behind the head of the chain
    • Priority:
      • History network will take precedence over the State network due to higher urgency
      • Ephemeral state work and performance optimizations may follow later, depending on team availability
  • Team bandwidth constraints
    • There will be reduced manpower on Trin side during summer due to vacation/ parental leave
    • Potentially exploring collaboration with Fluffy team on the State network development

2.6 Hive test

  • Two key failures remain unsolved
    • JSON-RPC extension: Some clients still need to expose required extensions via JSON-RPC
    • Post-merge proof validation: Minor bug persist in validating post-merge proof formats
  • Next step
    • Teams are urged to prioritize these fixes within the next two weeks to get Hive tests fully green, as prolonged failures risk being overlooked

Mar 24th, 2025 - Call #48

1. Team update

  • Update on Trin by Ognyan

    • Cleaned up history storage by removing old content value types
    • Refactored e2store header types for compatibility
    • Implemented ping extensions for JSON RPC endpoints
    • Added ephemeral store as a step toward ephemeral content for the History network

  • Update on Fluffy by Kim

    • Tuned the PoC syncing Nimbus EL from Portal Netowrk: incl. adjust recursive finContent concurrency (reducing requests from 3 to 2 slightly improved performance)
    • Observed that ~10% of recursive findContent requests timeout when sending 2-3 concurrent queries, potentially due to udp packet loss or async lib issue
    • Improved ContentDB pruning mechanism for speed
    • Integrated EVM and implemented eth_call, yielding a ~20% performance improvement
    • Started investigating the issue of discv5 protocol handling multiple concurrent requests

  • Update on Ultralight by ScottyPoi

    • Finalized ping extension for JSON RPC endpoints
    • Improved pruning mechanism (now more deterministic behavior with reduced overlaps)
    • Released major update with EthereumJS lib
    • Progressed on ephemeral header implementation
    • Preparing on Portal integration with EthereumJS client

  • Update on Shisui by Qi Zhou

    • Geth integration process update
      • PR re. repeat challenge in "who you are" handshake: Merged and currently being tested on Hive, have some issues with Fluffy nodes (need further investigation)
      • PR re. ping extension for JSON RPC endpoints: In final review stasge, expected to be merged shortly
    • Collaborated with Geth to make Shisui run as an independent sub-process for Portal Network access
      • Initial implementation planned by end of this month
    • Other work incl.
      • Participated in protocol versioning discussion
      • Start working on detailed offer decline code
      • Fixed race condition issues

  • Update on Samba by Meld

    • Close to passing all Hive tests, except a bug in putContent gossip handling
    • Slightly behind on SSZ union removal and ephemeral headers implementation
    • Working in parallel to integrate Samba as a plugin for Besu
    • Nearest goal is to fix the putContent bug and make initial progress on Besu plugin integration

2. Roll out network changes using Protocol Versioning

  • Deployment approach
    • Team agree to roll out offer/accept code changes using protocol versioning
    • Clients will start supporting both old & new versions and signal protocol version through ENR
  • Glados Monitoring
    • Need to add a tracking table on Glados to show the protocol versioning states
  • Issue discussed
    • Versioning semantic: Whether to rename portal wire protocol version to a more general portal version for future clarity
    • Lack of size prefix: FindContent/ Content request lacks a size prefix, making it difficult to distinguish between incomplete and invalid data, potentially leading to incorrect peer ban
  • Version 1 action items
    • Deploy offer/ accept codes
    • Add length prefix to uTP streams
  • Related PR/ issue

3. Banning/ Peer handling

  • Risk of peer banning
    • Could backfire in edge cases, eg. banning all peers during local network outage
    • May accidentally isolate nodes from the network
  • Peer scoring is preferred over hard bans
    • Prioritize good peers rather than punishing bad ones

4. Handling Ephemeral State Content

  • Main issue
    • Current trie-based model struggles with ephemeral state management during chain reorgs, as state content can exist on both canonical and non-canoncial chain
    • The issue arises when attempting to migrate state content to permanent storage or delete non-canonical chain related content after finality
  • Potential solutions
    • Solution 1: Modify content keys to include block hashes, enabling proof tracking across forks (but risks redundant storage)
    • Solution 2: Restrict trie-based model to finalized state only, using the flat model for ephemeral data near the chain head
  • Team consensus
    • Favored Solution 2 for simplicity
    • Open questions
      • Define the threshold of "finalized" (EL/ CL finality vs practical limits)
      • Document verification rules for headers near the head of the chain
  • Action items
    • Formalize flat model spec, led by Milos
    • Clarify finality checks in the state network specs
  • Related PR/ issue

5. Storage of Ephemeral Block Bodies/ Receipts

  • Main question
    • How clients choose to store the ephemeral block bodies/ receipts? Whether they should be store permanently?
  • Potential approach
    • Pre-finalized data: Keep in-memory cache (auto-purged on finalization)
    • Finalized data: Persist permanently (proofs are immutable post-finalization)
  • Client plans
    • Trin: Testing in-memory storage for pre-finalized bodies/ receipts
    • Fluffly: Exploring separate tables for headers/ bodies to enable efficient access & pruning, but still need further exploration

Mar 17th, 2025 - Call

1. Team update

  • Update on Trin by Mike

    • No major development update last week
    • Reth is able to sync the chain through Trin nodes
    • Two main areas for syncing performance improvement (from 34 hours to less than 10% of that)
      • Fix the issue of serving incorrect or corrupted data, potentially due to uTP
      • Address performance bottlenecks
    • Glados
      • Will focus on optimizing the Reth sync process, and maintain a scoreboard to track performance

  • Update on Ultralight by acolytec3

    • Ephemeral headers
      • Implemented the history network related features, incl. serving & storing content headers
      • Prepare to integrate the beacon network to track the head of chain and handle ephemeral headers
    • Question regarding the spec
      • The current spec requires storing the last 8,192 consecutive headers, which might be unnecessary and could potentially cause issues
      • Need further exploration/ clarification on how to handle headers once a sync period is finalized and historicalSummeries are pushed to the network

  • Update on Nethermind Portal implementation by FLCL

    • Plan to prepare the config and integrate Nethermind Portal Client into Portal Hive test

2. Discussion on the spec change

  • Relevant PR: https://github.com/ethereum/portal-network-specs/pull/370
  • Discussion over adding detailed offer decline codes to the spec
    • Concern: Adding the additional decline info might not be actionable and could overload the protocol
    • Beneift: Could be useful for optimizing content offering strategies and improving bridge operations
  • Team consensus
    • Would be better to clarify the purpose of adding the response codes
    • Will move forward with the change

Mar 10-11th, 2025 - Portal Summit Berlin

Mar 3rd, 2025 - Call #47

  • No call update this week as most of the team are travelling
  • There will be no call update next week (Mar 10th) due to in-person Berlin Summit

Feb 24th, 2025 - Call #46

1. General update

  • Update on Berlin summit by Piper
    • Working on multiple talks for the summit
  • Portal implementer call schedule for the next 2 weeks
    • March 3rd: Client team attendance uncertain; Confirmation of the call will be posted on the ethereum/pm github issue
    • March 10th: No call expected due to the Berlin in-person event

2. Team update

  • Update on Shisui by Grapepapa on EthR&D Discord

    • Ephemeral headers: Implemented the spec changes, incl. SSZ union removal, None removal in BlockHeaderWithProof, and add validation for post-Capella proof
    • trace_offer API: Work in progress
    • Geth integration: No updates this week
    • Berlin event: All team members have their visa ready

  • Update on Trin and Glados by Jason Carver, Milos and Nick

    • Trusted block root: Now embedded instead of requiring command-line input
    • Database migration
      • Added progress tracking to display stats during the migration
      • Decided not to migrate post-merge headers with proofs, but still keep the post-merge block & receipts
    • Beacon network: Added a pruning task to remove outdated light client bootstrap data
    • Glados
      • Now able to track database entries for every transfer failure, even if the audit succeeds
      • Working on the state audits for block 21 million, expected to have a PR soon
    • Add e2hs file format proposal
      • Proposed a new file format to be used in the History Network for storing pre & post-merge history data
      • Feedback from the Nimbus team is needed to understand why receipts were excl. and whether should be incl. directly or stored separately
      • Further discussions will take place in Berlin to gather feedback and finalize decisions on this format

  • Update on Ultralight by acolytec3

    • SSZ container discrepancy: Found and fixed a mismatch between the implementation and the spec for post-merge pre-Capella headers; Now passed all related portal spec tests
    • Hive tests
      • Aim to rerun the hive tests to verfiy the fix works
      • Noticed that some gossip tests for pre-merge blocks are failing, pending further investigation
    • Ephemeral headers
      • PR is ~25% complete, with fundamental work completed, incl. storing and advertising headers via the ping extension
      • Work remains on handling gossip and responses for content retrieval

  • Update on Samba by Meld

    • Hive tests: Passed 50+ tests related to the offer input, while some still failing
    • Recursive lookup: Work on implementing and finalizing the necessary endpoints
    • Additional endpoints: Implemented endpoints needed for testing the offer endpoint, including local content and history store
    • Besu integration: Planning to start research and integration in the coming weeks

3. Review ephemeral headers deployment plan

  • Deployment target for this week
    • Aim to remove SSZ union and enforce proofs for pre-merge headers
  • Client readiness
    • On the call, Trin and Ultralight are ready to deploy the change
    • Pending coordination with Fluffy and Shisui
  • Next steps
    • Confirm readiness from Fluffy and Shisui on discord
    • Team will move forward if the majority of the network is ready, even if some clients are temporarily offline

Feb 17th, 2025 - Call #45

1. General update

  • Update on Berlin summit by Piper
    • Organization & Deck preparation: Working on a new deck to present Portal from first principle, focusing on a deep dive into technical aspects

2. Team update

  • Update on Trin by Milos

    • History content migration: Progress on migrating history content to new types, incl. introducing ephemeral headers and updating the encodings for headers with proofs
    • Performance optimizations: Continued discussions on improving performance for syncing history network data
    • Glados: No major updates this week

  • Update on Ultralight by ScottyPoi

    • Union removal: Working on a PR to remove the bucket/ proof type and update data types
    • Ephemeral headers: Developing support for ephemeral headers with a PR in progress
    • Execution client rebuild: Discussed rebuilding the EthereumJS client based on Portal, potentially starting from the Ultralight client

  • Update on Fluffy by Kim

    • Ban list: PR merged, but choose to not activate it by default due to potential contact availability impacts
    • EVM integration: Basic eth_call support is functional, but no support for recursive call yet; testing shows similar performance to Trin
    • UTP benchmarks: Improved block processing, achieving 40 blocks/second at the 14M block height mark; noticed a UTP decoding error from Ultralight nodes
    • Historical Summaries: Updated a previous PR to add historical summaries to the REST API for Nimbus, pending to be merged

  • Update on Shisui by Qi Zhou

    • Geth integration: Waiting on Geth devs to review PRs for integration
    • History network: Worked on the new specs for the history network changes, incl. the nouns and SSZ union removal; code is ready, pending deployment plans
    • Beacon network: Progressed on beacon network integration, pending debug on some Hive tests

  • Update on Samba by Derek

    • Ping Extension: Completed the implementation of Ping extension
    • Next step: Setup infra to support ephemeral headers and recursive search
    • Explore collaboration with Besu: Considering align with Besu to maintain synchronization until beacon network integration is complete

3. Discussion on ephemeral bodies and receipts handling

  • General consensus
    • Ephemeral bodies and receipts should be gossiped when created and be proven using ephemeral headers
    • If the ephemeral bodies and receipts become invalid later, they can be deleted
    • No need for a separate content type for ephemeral bodies and receipts
  • Security concerns
    • DDoS attacks where the network could be flooded with invalid ephemeral bodies and receipts
      • The risk is generally low as reorgs are rare and shallow in the beacon chain (typically 1-2 blocks)
      • Invalid data wouldn't be propagated; it might linger in some clients but wouldn't infect the network
  • Next step
    • Update specs to clarify client behavior on handling ephemeral bodies and receipts

4. Coordination on ephemeral headers deployment plan

  • Merged the PR to remove the SSZ union from header with proofs
  • Post-merge bodies & receipts handling
    • General consensus
      • It's a client-specific decision on how to handle post-merge bodies and receipts that are not verified using proofs
      • Since invalid data would only affect the individual client, it shouldn't impact the overall network
    • Different handling approaches
      • Pruning approach: Use files with block header hashes to identify and prune invalid data, but that could require significant effort
      • Leave-as-it-is approach: Assume the data is valid and take no action, potentially cleaning up later if issues arise
  • Missing test vectors
    • Comprehensive test vectors for post-merge headers with proofs (eg. historical rules and summaries) are missing
    • Current tests only cover partial proofs, not the entire encoded content
    • Next step: Generate and verify complete test vectors before full implementation
  • Implementation status by client
    • Trin: Can implement new encoding and union removal by next week but won’t support ephemeral content until after Berlin summit
    • Fluffy: Supports pre-merge headers but hasn’t integrated new post-merge types; Limited availability next week for debugging issues
    • Ultralight: Ready for union removal and new encoding by next Monday
  • Deployment plan
    • Proposed order
      • Phase 1: Remove union types and enforce proofs for pre-merge headers
      • Phase 2: Add support for post-merge headers with new proof types
      • Phase 3: Implement ephemeral headers once each team is ready
    • Timeline
      • Teams agree to proceed with Phase 1 by next Monday
      • Each client will implement Phase 2 & 3 at their own pace

Feb 10th, 2025 - Call #44

1. General update

  • Update on Berlin summit by Piper
    • Preparing the opening talk for the Berlin event, aiming to provide a top-down overview of the Portal Network and get everyone onto the same page

2. Team update

  • Update on Trin by Nick & Ognyan

    • Performance improvements: Enabled WAL mode in SQLite databases, improving performance by 1.5x to 2x
    • UTP bug fix: Fixed a bug in the UTP library, which has been deployed
    • Hive testing: Merged a basic sync test for syncing light clients and added new JSON-RPC endpoints to the specs; portal clients will need to implement the new endpoints to ensure compatibility with the test
    • Minor refactor and new version of the Portal API
    • Header proof change: Preparing for the deployment of the new header with proof type, with the plan to convert/ update old data in the production database rather than trashing the database
    • Discussion on experiment with history network data download
      • Goal: To measure how fast data can be retrieved sequentially from the history network compared to geth sync
      • Results
        • Average download rate: c.5 blocks per second (c.20-23x slower than geth via devp2p).
        • UTP bottlenecks: Struggles with >60 connections and caps at 2-3 MB/s.
        • find_content latency: ~5 seconds (95th percentile).
      • Comparison with Fluffy
        • Can handle 50+ simultaneous transfer with larger batches
        • Performance is limited by chunked downloads and connection ramp-up time
      • UTP performance bottleneck: Currently struggled with inbound connection scaling, and encryption overhead
      • State network implication: The state network is much larger than the history network, requiring 10x more performance; current performance improvements (eg. range queries) may not suffice for state network requirements
      • Next step
        • Focus on UTP improvements, incl. improve UTP implementation and testing frameworks, and explore native discv5 connection methods
        • Further discussion on performance improvement strategy at Berlin summit

  • Update on Fluffy by Kim

    • UTP benchmarks: Improved UTP benchmarks to handle more nodes sending data simultaneously
    • Union removal & database migration: Optimized SQL statements for faster database migration, pending for further testing
    • EVM integration: Exploring EVM integration, potentially using EVMC API, allowing other EVMs to be plugged in
    • Others: Finished the bandwidth implementation, which is read to merge; working on a discv5 issue as part of ongoing improvement

  • Update on Shisui by Qi Zhou

    • Geth integration: Most PRs are submmitted, pending for review
    • Discv5 issue: Addressed discrepancies between Portal and Geth’s discv5 implementations, particularly around request-response handling
    • Glados issue: Fixed an issue of Glados, which now allows timestamp as sequence number
    • UTP benchmark & improvement: Significantly reduced memory footprint (from ~60 GB to ~100 MB) and improved performance (able to handle up to 2000 concurrent UTP streams, up from 500) with peak memory usage of c. 80 GB during high traffic

  • Update on Samba by Meld

    • UTP Implementation: Completed the first version of UTP, enabling read/write file exchange & the getContent endpoint, though some validation is still needed
    • Hive testing progress: Passing more Hive test, but currently still using an old version of the instance due to delay in ping extension implementation; expected to finish the update by the end of the week
    • Next step: Push a new instance of Samba, work on the implementation of offer endpoint, and continue discussion with Besu for integration

  • Update on Ultralight by acolytec3

    • Hive sync test: Implemented and merged new Hive sync tests for the beacon network
    • Header proof change: Ready to remove the union from block header proofs, passing all test factors; plan to start work on ephemeral headers this week
    • Browser client: Plan to reboot the browser client with help from the new contributor Justin

3. Coordination on ephemeral headers deployment plan

  • Timeline: Teams agreed to target 14 days from the call for deploying the ephemeral headers change
  • Database migration
    • Plan to delete headers without proof and convert pre-merge headers to the new format
    • Discussed whether to delete or keep post-merge block bodies and receipts; consensus leaned toward deleting the post-merge data due to lower importance for now
    • A temporary availability blip during migration is expected but considered acceptable
  • Beacon network dependency
    • Historical summaries from the beacon chain are required to verify headers with proofs, and team is ensuring the beacon network is ready before the migration
    • Clients will stop accepting headers without proofs and only accept those with valid proofs once the Beacon Network is ready

Feb 3rd, 2025 - Call #43

1. General update

  • Update on Berlin summit by Piper
    • A planning document is now available: https://notes.ethereum.org/_qEGASgDTLK4_EOAMSHctg?view
    • Team are encouraged to add topics for presentations & discussion
    • Expected attendance
      • Apart from Portal teams, members from Geth, Reth, Nimbus, Nethermind, and Besu are also expected to participate. Erigon won’t aattend due to a scheduling conflict.

2. Team update

  • Update on Trin & Glados by Jason Carver and Nick

    • Ping extension deployment
      • The deployment went smoothly, though ~30 nodes were lost, likely due to lack of upgrades
    • Debug improvements
      • Focused on fixing issues incl. setting gas_price when handling eth_call and resolving bugs in content storage checks
    • Node radius adjustment
      • Nodes starting at 100% radius were causing excessive work for state nodes
      • A fix to reduce the max-radiums param to 5% is in progress
    • Hive testing
      • Sync tests for the beacon network are being developed
      • Feedback and feature suggestions for Hive are welcome to drop in the #hive channel
    • Glados Metrics:
      • Improvements in global metrics were noted, with fewer failures observed recently
      • The exact cause is unclear but may relate to latest changes in header acceptance logic. Further investigation are welcome

  • Update on Ultralight by ScottyPoi and acolytec3

    • Glados metrics
      • Also observed a significant improvement in boot node census response rates on Glados
    • Ping extension
      • Discussed how to respond to unexpected type 2 pings without requiring a prior type 0 ping
      • Concluded that nodes will respond with the matching type without penalizing peers for unexpected type 2 pings
    • SSZ union removal
      • Worked on a PR to remove the union in BlockHeaderWithProof
    • Ephemeral headers
      • Plan to work on it later in the week, targeting a working implementation before the Berlin event
    • New contributor
      • A new contributor may join Ultralight soon to explore brower-based implementation

  • Update on Fluffy by Kim

    • Ban lists
      • A more generic, lower-level approach has been implemented, which can be applied on both discv5 and portal wire protocol
      • The part for discv5 has been merged, the part for portal wire is expected to be merged soon
    • Database pruning
      • A simple implementation is being tested, but it's inefficient due to the database layout
      • Challenges are the current approach is time-consuming, and pruning could potentially affect Glados metrics as nodes with pruned databases might behave differently
      • Input is needed from other clients (eg. Trin, Shisui) regarding whether to implement pruning or to delete & re-inject data to determine a consistant approach
    • Discv5 native stream: Start to explore and benchmark it

  • Update on Samba by Meld

    • UTP Integration
      • Partically completed the getContent & store endpoints, and successfully started receiving content via UTP
      • Plan to enable writing content through UTP and finalize changes for the ping extension
      • Aim to start passing more Hive tests by the end of the week

  • Update on Shisui by Qi Zhou

    • Integration with Geth: Several PRs have been submitted, incl.
      • Make APIs public for reuse in Portal Network
      • Introduce timeout parameters in discv5
      • Introduce node caching
      • Remove redundant ping challenges
      • Time consideration: Some changes may require updates to the discv5 spec, which could take additional time
    • Prepare for Berlin event
      • Team members are applying for visas and aim to resolve outstanding issues during in-person meetings in Berlin

Jan 27th, 2025 - Call #42

1. General update

  • Update on Berlin summit by Piper
    • Venue confirmed
    • More details, incl. event agenda, will be shared soon

2. Team update

  • Update on Trin by Ognyan and Milos

    • Added UTP duration metrics to measure performance
    • Fixed issues with the gas_price when handling eth_call
    • Experimented with block fetching from the network, generated metrics which could be further optimized
    • Made progress on developing testing for light client nodes on the beacon network
    • Initialized the big storage nodes, with basic tests (eg. eth_call, getbalance) showed some flakiness, indicating intermittent issues with content retrieval
      • Need mechanisms to propagate retrieved data back into the network for broader availability

  • Update on Samba by Meld

    • Worked on addressing issues with package exchanging using UTP

  • Update on Ultralight by acolytec3

    • Completed a fully working prototype for verifying post-Capella headers on Ultralight
    • Developed a bridge script to pull historical data, construct proofs, and push them to the nodes
    • Local multi-node gossip testing is planned to ensure proper header propagation

  • Update on Fluffy by Kim

    • Added support for ephemeral headers, incl. encoding/decoding
    • Created PR with test-vectors regarding the content key
    • Proposed removal of SSZ unions in BlockHeaderWithProof type by making the proof an SSZ encoded ByteList
    • Fixed bugs related to UTP socket handling and improved data validation process
    • Experimented running Glados within the Nimbus infra, but encountered a dependency on PandaOps
    • Identified an undefined failures field in the trace object spec, causing issues when running Fluffy

  • Update on Shisui by Qi Zhou

    • Most of the team members are on national holidays starting from last weekend, which may cause delays in responses
    • Merged the PR on ping extension and ready for deployment
    • 1st PR (adding stun protocol) submmit to Geth merged successfully
    • Prepared for the next PR to update the discv5 APIs, which will allow Shisui to run on top of Geth without the need to merge or modify Geth code directly

3. Coordination on Ping extension

4. Coordiantion on ephemeral headers implementation and SSZ union removal

  • Relevant link: https://github.com/ethereum/portal-network-specs/pull/362
  • Background
    • Multiple breaking changes, incl. getting ephemeral headers alive, remove SSZ union, and database migration, needs team coordination to avoid repeated operation
    • High priority task, as portal network approaches history network adoption
  • Goal
    • Setup a tracking issue to define timelines, migration plans, and readiness checks for these updates
    • Coordinate different teams to ensure client readiness and minimize disruption during the transition
    • Ideally within the next few weeks, before March
  • Kim will take the lead on coordination

Jan 20th, 2025

1. General update

  • Berlin Event update by Piper
    • Working on finalizing the event venue
    • Prepared event materials, incl. agenda, documentation etc.

2. Team update

  • Update on Trin & Glados by Ognyan, Kolby
    • Trin
      • General
        • Refactored the code to improve the repo structure
        • Improved test coverage for light clients and improved the content lookup service to check the local database at OverlayService level
        • Identified and working on a bug where queries persist during UTP transfer
      • Ping Extensions
        • Streamlined the ping extension by removing a serialization layer through combining the custom ping payload with the actual ping message (PR pending to be merged)
      • State Network
        • Trin node performance currently is the primary bottleneck
        • Working on achieving 10x performance improvement
    • Glados
      • Updated node visualization with weekly census tracking
  • Update on Fluffy by Kim
    • Ping Extensions
      • Proposed the streamline suggestion in the portal discord and implemented the ping extension
      • Pending team coordination for deployment
    • Peer Banning List: Working on implementing the temporary peer ban list (conditions for banning: validation failures and invalid offers)
    • UTP Reset: Identified a bug where connections timed out without sending a FIN or RESET packet
    • Node Deployment: Planning to deploy c.100 small nodes (2-4 GB) to test default user config
  • Update on Ultralight by ScottyPoi
    • Ping Extensions
      • Implemented and ready for deployment
      • Pending team coordination for deployment
    • UTP Reset: Implemented the reset packet to match team expectations and align with other clients' behavior
    • Peer Banning: Implemented peer banning at the UDP level using a mechanism discovered in discv5
  • Update on Samba by Meld
    • UTP Library: Completed a new version. Aim to integrate & pass through the hive tests, and refine the UTP library based on feedback
  • Coordination on Ping Extension
    • Deployment date: Aim to deploy next Monday, pending Shisui's readiness
    • Hive test: Coordinate to set up experimental Hive tests with merged client branches to validate changes before deployment

3. Other discussion

  • Chloe raised the point of making Portal's weekly meeting agenda and notes more publicly accessible through ethereum/pm github and Magician forum

4. Next week's call agenda

Jan 13th, 2025

1. General update

  • Upcoming Event update by Piper
    • Near confirmed dates for March 10th & 11th in Berlin. It will focus on collaboration with execution client teams for Portal integrations
    • Details will be confirmed and finalized soon

2. Team update

  • Update on Trin & Glados by Kolby, Nick
    • Trin
      • State Diff Bridges: Deployed to gather data, but high traffic and bandwidth bottlenecks seen on Trin nodes may hinder research on implementing range queries
      • State Nodes Deployment: Full state data from block 21 million will be deployed this Tuesday
      • Trin desktop: Added account-related features, allowing users to save accounts across sessions and look up data by account address
      • Ping Extensions: Planned for deployment alongside Fluffy, and will coordinate the update across all teams, incl. Ultralight, Samba, and Shisui
    • Glados
      • Fixed visualization crashes and added a weekly census count to track node fluctuations
  • Update on Fluffy by Kim
    • Portal sync: Worked on the portal sync proof of concept for execution client
    • JSON-RPC Methods
      • Added content validation + local storing to Portal history JSON-RPC methods
      • Raised the issue of validation requirements for getContent and putContent, where the spces are ambiguous and full validation significantly impacts bridge performance
      • Proposed adding a validation flag to toggle between different validation levels
    • Ping Extensions: Initial exploration completed, with plans to develop a basic version supporting standard extension types
    • Noticed unusual traffic patterns in metrics
      • Significant increase in UTP connections but decrease in connection duration
      • High session cache miss ratio (spiked to 40-50%), indicating many new nodes are connecting but not staying long
  • Update on Samba by Meld
    • UTP Library: Continued working to improve test coverage and synchronization. Aimed to expose getContent and offer
      s for Hive testing next week
  • Update on Shisui by Qi Zhou
    • UTP Performance: Testing a new UTP library in Glados, showing significant improvements (e.g., memory usage reduced from 10GB to <1GB).
    • Public APIs: Created reusable public APIs of the discv5 in node discovery, with plans to merge the PR into upstream
  • Update on Ultralight by ScottyPoi & acolytec3
    • LIbrary update: Work with ChainSafe (maintainer of discv5 lib in TS) to update the cryptography library used in EthereumJS
    • UTP transfer issue
      • Identified issue of receiving out-of-order packets from Trin nodes, causing duplicate responses and clogging the UTP stream
      • Current solution involved slowing down transfers and implementing packet queuing/ resorting
    • Ping extension: Will prioritize the implementation of ping extension
    • Hive test failure issue
      • All hive tests failed currently due to a low-level bug between Hive and the discv5 lib
      • Bug location has been identified and the fix is in progress

Jan 6th, 2025

1. General update

  • Upcoming team event: The event may either be a Portal Summit or an Execution client interop event, depending on attendance. The event is currently scheduled for March 10th & 11th in Berlin.

2. Team update

  • Update on Trin & Glados by Michael Ferris, Kolby, and Milos
    • Trin
      • Update portal_*Gossip to portal_*PutContent based spec
      • Work on implementing ping extensions
      • State network
        • Hardware availability & node deployment: Hardware delays are resolved, and 20 nodes are being deployed to gossip state data from block 20 million to the latest. Initialization of the nodes with a snapshot is expected to take a day, with gossiping completion estimated in 15-20 days.
        • Bridges & Testing: A few bridge nodes are being setup to help gossip heads and address missing state data during primary testing, ensuring smoother gossiping and synchronization.
        • Timeline: Near-head state is expected to be live by the end of Jan
    • Glados
      • No major updates on Glados over the holidays
      • A few PRs were merged, including a graph visualization for node census over time, helping track network dips and activity
  • Update on Samba by Meld
    • Work is ongoing on the UTP library to improve coverage and test
    • Work on content serialization and integration into the testing framework
  • Update on Shisui by Qi Zhou
    • Merge code updates to align with the latest PutContent spec
    • Pass all tests for the UPP code migrated from Trin
    • Fix issues with concurrency and large data transfers in the Go implementation
    • Test nodes on Glados to ensure stability

3. Other discussion

  • Hive test failure issue raised by Jason Carver
    • Hive has been pinging with failures since Christmas Eve, likely due to clients updating to the new PutContent spec
    • Teams are encouraged to investigate and resolve these failures to ensure test stability
Last changed by 
 
Chloe Zhu
https://github.com/Chloezhu010 https://x.com/Chloe_zhuX
0
1200

Read more

EPF cohort 6 华语区 Townhall

主要目的

Apr 22, 2025
Portal Summit Berlin 2025

Stewardship of Ethereum data

Mar 14, 2025
Setup Node Workshop

Download the geth's pre-built binary (in tar.gz) and the stable release (in .asc)

Feb 28, 2025
RollCall 8.1 Summary

Recording:

Oct 27, 2024
Read more from Chloe Zhu
Published on HackMD