Skip to content

SSMT State Engine

SSMT (Scalable State Merkle Trie) is Altius Labs' parallel state-root engine. Computing the state root — hashing a Merkle Patricia Trie over every account and its storage — is traditionally a largely serial bottleneck on an execution client's hot path. SSMT moves that work into a dedicated, multi-threaded engine that computes the root in parallel and returns it to the client.

SSMT's advantage grows with the state it operates on: as accounts and storage accumulate and the trie deepens, root computation comes to be dominated by the latency of reading trie nodes from disk. This is the work SSMT parallelizes.

On traditional blockchains, the standing solution to the problem of state bloat is to increase the hardware; this approach will eventually become economically impractical. Holding a 766 GiB state resident, for instance, means roughly a terabyte of RAM per node. And for public chains with permissionless node operators, greater hardware requirements adversely affect the chain's decentralization.

What it is

  • A single-process engine plus a server binary (ssmt-server) that connects to one execution client over an IPC transport (a Unix domain socket today).
  • The client streams per-block state changes to SSMT; SSMT computes the new state root on a pool of worker threads and returns it for the client to validate against the block header.
  • Deliberately decoupled from client internals (built on the stable Alloy libraries) so the engine can be reused across EVM implementations.

What it is not

To set expectations clearly: SSMT is a state-root computation engine, not a distributed database. Its parallelism comes from worker threads within one process, not from sharding state across multiple machines. There is no multi-node replication or cross-node fault-tolerance layer.

How it fits the platform

SSMT accelerates state-root computation for the Reth-based execution layer of Altius L1, helping the chain keep state handling off the critical path as state grows. See How It Works for the internals, and Benchmarks for how it performs against Reth's state root.