GATEWAY
ProductionGnosis Chain

Scaling and securing a public network with reliable RPC and validators

Operated a large share of Gnosis Chain’s live infrastructure for over three years , the primary public RPC, validators, and protocol engineering , at ~10.2B monthly requests, sub-22 ms latency, and 99.99% uptime.

~10.2BMonthly requestsServed at peak
<22 msMean latencyBasic EVM methods
99.99%Uptime in Europe99.98% in Southeast Asia
3.5k–4kRequests per secondSustained at peaks
40+Nodes~70% bare metal, hybrid public cloud
~9%Mainnet validatorsRun in private data centers

Overview

For over three years Gateway was Gnosis Chain’s strategic infrastructure partner, operating a large share of the network’s live infrastructure at scale — its primary public RPC, a significant validator set, and critical services such as the ENS resolver and interoperability (bridges). We also researched and tested major network upgrades, including the Merge to proof-of-stake, contributed client patches, and coordinated hard forks to keep the protocol current. Mixed client pools and established SRE practices kept the network fast, stable, and cost-efficient for builders.

The opportunity

Developers on Gnosis needed fast, reliable access to heavy EVM methods, wide log ranges, and stable endpoints that did not impose artificial limits. The goal was to make public RPC feel premium for experimentation, and to provide a seamless path to private RPC that scaled without friction.

Solution

Secure, scalable RPC and validator operations

01

Public and private RPC

  • Operated the core public endpoints for Mainnet and Chiado, including rpc.gnosischain.com, rpc.gnosis.gateway.fm, and rpc.chiado.gnosis.gateway.fm.
  • Provided a clean path to private RPC with unlimited API keys, requests, and traffic.
  • Sustained ~1,500 RPS across regions on private RPC for 100+ unique keys, covering eth_getLogs, traces, batch calls up to 500 items, and eth_call.
02

Traffic engineering for performance

  • Method-aware routing across mixed client pools — for example, eth_getLogs to Erigon and eth_getProof to Nethermind for optimal speed and correctness.
  • Supported batch sizes up to 500, full trace/debug, and log ranges from 0x to latest.
  • Proxy-level caching and automatic retries to reduce latency and error rates.
03

Industrial-grade deployment and analytics

  • Ran 40+ nodes across multiple data centers on a hybrid footprint — most workloads on bare metal (Dell R6615, AMD EPYC, NVMe SSDs) to cut cost, with critical components on public cloud to preserve reliability.
  • Maintained CI/CD pipelines covering integration, end-to-end, and load testing for node and RPC releases.
  • Provided per-project analytics with breakdowns by key, method, status code, and more.
04

Validators and protocol engineering

  • Operated ~9% of mainnet validators and one-third of Chiado in private data centers for resilience and decentralization.
  • Hosted the ENS resolver and interoperability (bridges) services, and participated in their governance.
  • Researched, tested, and shipped client upgrades — including the Merge to proof-of-stake — contributed patches, and coordinated hard forks to keep the network current.
  • Adopted CDK-Erigon on Gnosis and supported related devnets on Chiado; contributed infrastructure for Gnosis Pay L2 research and Optimism Bedrock-based experiments.
Response time vs. the public endpointLower is better · milliseconds
Gateway RPC Public gnosischain.com

Europe Mean response time, basic EVM methods

Frankfurt
10.30 ms
39.47 ms
London
11.17 ms
54.77 ms
Paris
16.83 ms
50.70 ms
Ireland
20.93 ms
62.97 ms
Milan
21.57 ms
50.43 ms
Zurich
22.20 ms
47.60 ms
Stockholm
27.73 ms
58.47 ms
Spain
38.47 ms
69.47 ms

Southeast Asia Mean response time, basic EVM methods

Singapore
4.53 ms
193.07 ms
Jakarta
21.53 ms
208.50 ms
Benchmarked with an independent third-party tool (comparenodes.com) on basic EVM methods (eth_blockNumber, eth_getBalance). Detailed reports: Europe · Southeast Asia.
Sustained throughput by keyPrivate endpoints, EU region · mean req/s
Key 01852 req/s
Key 02276 req/s
Key 03275 req/s
Key 04125 req/s
Key 05115 req/s

Peak per-key throughput from a representative EU sample; the private tier served 100+ unique keys in total. Key identifiers are anonymized.

Global uptimeRolling window, July 2024 – July 2025
Europe99.99%
Southeast Asia99.98%
Operational Alert window
Architecture

A dedicated RPC stack, replicated per region

Dedicated RPC stack deployed per region across three independent locations: user traffic enters through the Gnosis global load balancer, passes an aggregation and proxy layer, and is served by mixed consensus and execution node pools, with a parallel operations plane.

Ingress

User
Gnosis global load balancer

Aggregation & proxy

Aggregator (HTTP)
Aggregator (WS)
Proxy · consensus
Proxy · execution
Proxy · subscriptions

Operations

Admin UI
Rate policy · public / private
Analytics
Alerting

Consensus node pools

Lighthouse · archival
Lighthouse · non-archival

Execution node pools

Nethermind · archival
Erigon · archival
Nethermind · non-archival
Erigon · non-archival
  • Mixed node pools mapped per method profile for speed and reliability
  • Global routing with caching and retry logic at the proxy layer
  • Full trace/debug support and wide log ranges
  • Batch handling up to 500 items per request
  • Project-level telemetry and alerting to maintain SLOs
Outcome

Reliability and speed at network scale

  • Served ~10.2B monthly requests at peak, with sustained throughput of 3.5k–4k RPS.
  • Delivered mean latency under 22 ms for basic methods on public RPC in target regions.
  • Measured 99.98% uptime across Gnosis endpoints over the last year.
  • Delivered a developer experience without paywalls, compute-unit throttling, or narrow log windows — removing common blockers to building on the network.

Why it mattered for the ecosystem

High-quality public access lowered the barrier to entry for new teams, while private RPC with unlimited scaling supported production apps with demanding read and archive workloads. Together, this helped Gnosis attract more builders, increase network utility, and stay responsive under load.

In their words
…affordable and highly reliable RPC hosting services… during the Chiado testnet merge, processing the first post-merge block. We look forward to this continued…
Gnosis
Stefan GeorgeCTO, Gnosis

Related Gateway capability

Review the capability connected to this engagement, including its intended scope and operating boundary.

ExploreRPC infrastructure

Discuss the requirement behind this case.

Map your context, control boundary, and validation evidence with the Gateway team.