Solana RPC 2.0: Triton One and Solana Foundation Rebuild Network’s Read Layer from Scratch
Summary
Triton One and the Solana Foundation have announced RPC 2.0, a complete rebuild of Solana’s data read layer — the infrastructure that handles every balance check, transaction lookup, and token query on the network — marking the most significant infrastructure upgrade since the chain’s launch.
The core change is architectural: reads are being removed from the Agave validator client entirely and moved into two independent, open-source modules. The Accounts module replaces full database scans with adaptive indexes that build themselves from an application’s actual query traffic, dramatically reducing latency and eliminating the need for validator-class hardware — currently 768 GB of RAM — just to serve account data. The Historical module stores the entire Solana ledger in a columnar engine optimized for how apps query past transactions, collapsing what currently requires hundreds of sequential RPC calls for a single wallet’s history into one query returning results in milliseconds.
The existing read layer has remained largely unchanged since genesis, built as an afterthought alongside the validator rather than as purpose-built infrastructure. At Solana’s current scale, that architectural debt has become a hard ceiling: developers have been forced to either run expensive dedicated nodes, accept throttled public endpoints, or lock themselves into closed-source proprietary APIs — all workarounds for a problem that RPC 2.0 addresses at the foundation level.
Both modules ship open-source under the AGPL license, governed through a neutral RPC organization managed by the Solana Foundation. The project is being built in stages throughout 2026, with technical benchmarks and deployment guides to be published as each milestone ships.
Triton One said developers can access the infrastructure ahead of the open-source release via a Triton endpoint.