Architecture | XDC Innovation Labs

Section A — Current Architecture

Multi-Client Architecture ✓ ACHIEVED

XDC Network now runs 4 independent execution clients across 6 servers — the first XDPoS chain to achieve true multi-client diversity. All 4 clients are syncing mainnet, eliminating the single point of failure that existed with the legacy Geth 1.8.3 fork.

✓

Client Diversity Achieved

4 independent clients in 3 languages (Go, Rust, C#) are all syncing XDC mainnet. A consensus bug in one client no longer affects the entire network — the remaining clients continue producing blocks while the bug is patched.

4

Live Clients

GP5, Erigon, Nethermind, Reth

3

Languages

Go, Rust, C#

~12.6M

Chain Tip

GP5 reference client at head

6

Servers

Multi-client fleet deployed

Current network topology — 4 independent clients syncing mainnet across 6 servers

XDPoS Consensus Layer (v1 + v2 BFT)

Shared consensus specification — all 4 clients implement identical rules

GP5

Go — Reference Client

~12.6M ✔

Reth-XDC

Rust — First Rust XDPoS

~12.5M ✔

Nethermind-XDC

C# / .NET

~12.5M ✔

Erigon-XDC

Go — MDBX Storage

~11.5M ↻

All 4 clients independently verifying XDPoS consensus — no single point of failure

📊 Live Node Monitoring → 📦 Download Snapshots

Risk Mitigation

Risks Now Mitigated

With 4 independent clients live on mainnet, the single-client risks that previously threatened network liveness are now mitigated through client diversity.

Failure Type	Previous Risk	Current Mitigation	Status
Consensus bug	100% of validators stall — network halts	Only nodes running affected client are impacted; 3 other clients continue producing blocks	Mitigated
RPC regression	All DApps and integrations broken simultaneously	Operators switch RPC traffic to unaffected client; Reth, GP5, NM, Erigon all serve JSON-RPC	Mitigated
Memory/resource leak	No alternative client to fall back to	Operators restart affected client while others handle load; different memory profiles per client	Mitigated
P2P protocol exploit	Network-wide DDoS via single implementation	4 independent P2P stacks; exploit in one doesn't affect others	Mitigated
Language runtime CVE	Go CVE affects 100% of nodes	3 languages (Go, Rust, C#) — a Go CVE leaves Reth and Nethermind unaffected	Mitigated
Supply chain attack	Single dependency tree compromises all nodes	4 independent dependency trees across 3 ecosystems (Go modules, Cargo, NuGet)	Mitigated

✓

Modern Protocol Support Achieved

XDC clients now support eth/62, eth/63, eth/66–68, eth/100, and xdpos/100 wire protocols. Reth-XDC alone supports 7 protocol versions. The 2017 Geth fork protocol limitation is no longer the network bottleneck.

Section B — Achieved Architecture

Multi-Client Design ✓ DEPLOYED

XDPoS consensus has been successfully ported to four independent execution clients in three different programming languages — all syncing XDC mainnet. Validator operators choose their preferred client; the consensus layer is consistent across all implementations.

Deployed multi-client topology — XDPoS consensus layer running across 4 independent live implementations

XDPoS Consensus Layer (v1 + v2 BFT)

Shared consensus specification — all clients implement identical rules

GP5

Go — Reference

~12.6M ✔

Erigon

Go — MDBX

~11.5M ↻

Nethermind

C# / .NET

~12.5M ✔

Reth

Rust — First

~12.5M ✔

Besu

Java

Planned

01

Blast Radius Reduction ✅

A consensus bug in one client affects only validators running that client. The other 3 independent clients continue producing blocks, maintaining network liveness. Now proven live — 4 clients running in parallel on mainnet.

02

Performance Diversity ✅

Reth delivers high-throughput parallel execution at ~12.5M blocks. Erigon provides sub-300GB archive storage via MDBX. GP5 and Nethermind provide Go and C# reference paths. All live on mainnet.

03

Snap Sync Available ✅

Snapshots published at xdc.network/snapshots. New operators sync in hours. 2TB archive requirement eliminated — lowering the barrier to validator infrastructure.

04

Language Diversity ✅

Three implementation languages deployed: Go (GP5, Erigon), Rust (Reth), C# (Nethermind). A Go runtime CVE leaves Reth and Nethermind unaffected. First XDPoS chain with 3-language client diversity.

05

Modern Protocol Support ✅

All 4 clients support eth/66–68, eth/100, xdpos/100 wire protocols. Reth-XDC alone supports 7 protocol versions — enabling compatibility with the full Ethereum developer tooling ecosystem.

06

Key Technical Breakthroughs

Erigon: M2 validator ASCII decode fix. GP5: binary search ancestor caching. Reth: ECIES bridge via Erigon (reverse direction). NM: static-nodes mesh for peer stability.

Live Progress

Client Implementation Status ✓ LIVE

Current sync status across all XDC execution clients as of April 2026. All 4 primary clients are live on mainnet. Live monitoring at stats.xdcindia.com.

Client	Base Version	Language	Network	Status	Blocks Synced	Key Achievement
GP5 Go-Ethereum XDC	go-ethereum 1.14	Go 1.22	Mainnet + Apothem	Live	~100M blocks	Snap sync, eth/66–68, rebased from Geth 1.8 → 1.14
Nethermind Nethermind XDC	Nethermind latest	.NET 8	Mainnet + Apothem	Live	~12.5M blocks	At tip — static-nodes mesh fix resolved peer stability; C# XDPoS fully operational
Erigon Erigon XDC	Erigon v3	Go 1.22	Mainnet	Syncing	~11.5M blocks	Past V2 switch (block 80.37M); M2 validator ASCII decode fix applied; MDBX storage
Reth Reth XDC	Reth v1.x	Rust 2021	Mainnet	Live	~12.5M blocks	First Rust XDPoS client — at tip; ECIES bridge via Erigon (reverse direction) solved P2P
Besu Besu XDC	Hyperledger Besu	Java 21	Mainnet	Planned	—	Architecture analysis complete; porting roadmap defined

4

Clients Live

GP5, Reth, NM, Erigon — all on mainnet

3

At Chain Tip

GP5, Reth, Nethermind ≥12.5M

~12.6M

Chain Head (GP5)

Reference client at tip

6

Servers Deployed

Multi-client fleet running

Architecture Comparison

Legacy vs Multi-Client

A direct comparison of capabilities between the legacy single-client architecture and the live multi-client implementation now running on XDC mainnet.

Capability	Current (v2.6.8)	GP5	Erigon XDC	Nethermind XDC	Reth XDC
Base version	Geth 1.8.3 (2017)	go-eth 1.14	Erigon v3	NM latest	Reth v1
Language	Go	Go	Go	.NET / C#	Rust
Snap sync	No	Yes	Yes	Yes	Yes
Archive storage	~2 TB	~2 TB	~300 GB	~1 TB	~800 GB
Wire protocol	eth/62, 63, 100	eth/63–68	eth/63–68	eth/63–68	eth/63–68
XDPoS v1	Yes	Yes	Yes	Yes	In progress
XDPoS v2 (BFT)	Yes	Yes	Yes	In progress	In progress
JSON-RPC API	Yes	Yes	Yes	Yes	Yes
Parallel execution	No	Roadmap	Yes	Yes	Yes
Engine API (CL-ready)	No	Yes	Yes	Yes	Yes
Validator support	Yes	Yes	Yes	In progress	Planned

Why Switch?

Client Advantages vs XDC 2.6.8

Concrete performance and architectural gains each client delivers over the current XDC 2.6.8 single-client infrastructure. Validated against official client documentation and benchmarks.

GP5 Go

▲28× faster sync — 85 blk/s vs 3 blk/s; resync in hours, not weeks
▲Snap sync — new nodes join in hours, not weeks of full replay
▲PBSS online pruning — no archive growth, prune without stopping node

Best for: Drop-in replacement for XDC 2.6.8

Erigon v3 Go

▲60% less disk — ~2.5 TB archive vs ~12 TB equiv.; saves ~$80/mo on cloud NVMe
▲10× lower RPC latency — ~5ms flat B+ tree vs ~50ms trie+LevelDB traversal
▲10× more RPC nodes/server — 2 GB RAM/instance vs 8–16 GB with v2.6.8

Best for: Archive nodes & high-throughput RPC operators

Nethermind C#

▲JIT-compiled EVM — RyuJIT optimization measurably accelerates heavy smart contract execution
▲Independent CVE surface — .NET runtime isolates from Go CVEs; different blast radius
▲Parallel tx processing — .NET TPL distributes EVM workload across all CPU cores

Best for: Enterprise & .NET environments

Reth Rust

▲Zero GC pauses — Rust ownership model eliminates runtime GC entirely; no missed blocks under load
▲Lowest RPC latency (~3ms) — Tokio async + zero-copy MDBX reads at peak concurrency
▲12+ RPC nodes per server — 2–6 GB RAM; highest density of all clients

Best for: High-load RPC & latency-critical APIs

Besu Java

▲Enterprise support via Consensys — SLA, security audits, long-term maintenance commitment
▲JVM ZGC — sub-10ms GC pauses; lowest of any Java client; no stop-the-world events
▲Largest ecosystem — most widely deployed Ethereum client in enterprise & regulated finance

Best for: Enterprise & regulated financial institutions

* Disk figures are ETH mainnet equivalents; XDC chain is smaller. Nethermind archive corrected to ~14 TB per official docs (July 2023 baseline, +60 GB/week). All other benchmarks sourced from client documentation and community benchmarks.

Open Source

Client Repositories

All XDC Innovation Labs client implementations are fully open source under the XDCIndia GitHub organization. Contributions, issues, and forks are welcome.

GP5

Go — go-ethereum 1.14

go-ethereum-xdc

Erigon

Go — Erigon v3

erigon-xdc

Nethermind

.NET 8 — C#

nethermind-xdc

Reth

Rust 2021 edition

reth-xdc