Transaction Lifecycle
Ingestion: Transactions are pulled from the Data Availability layer.
Interpretation: Transactions are sent to the Interpreter and executed optimistically in parallel, generating a read/write set for each.
Resolution: The read/write sets are passed to the Resolver, which identifies instruction-level dependencies and conflicts.
Coordination: The Coordinator (OCCDA) schedules the transactions.
Re-Execution: Any instructions identified as conflicted are re-executed in parallel.
Finalization: The Coordinator finalizes the block's state transition.
Settlement: The final state root is published to the Settlement layer.
State Management: State management is handled entirely by the SSMT architecture. The state is not viewed as a single MPT on one machine but as a sharded, distributed database. Access is memory-first for speed, with the sharded cluster of SSDs providing persistence.
Fault Tolerance: Fault tolerance is built into the Storage (SSMT) layer. Each state shard is stored across different nodes, providing redundancy. The failure of a single node will not halt the network or result in data loss, as the state is replicated and can be recovered from other nodes in the cluster.
Last updated