%20Workflow.png){kind=link}
A runtime execution model that allows blockchain transactions to be executed in parallel speculatively, while ensuring a deterministic final order. It reduces lock contention, adapts throughput dynamically to network conditions, and improves stability under high load without sacrificing consensus safety.
............................................
Deterministic Speculative Scheduling (DSS) is a new runtime model for Solana-style blockchains. It improves throughput and stability by executing transactions speculatively in parallel, while committing results in a deterministic final order.
Solana’s Sealevel runtime already supports parallel execution, but in practice it runs into bottlenecks because:
-
Lock contention • Transactions that access the same account can’t run in parallel. • When contention spikes (NFT mints, liquidations, meme coin rushes), throughput drops sharply.
-
Static batching / slot sizing • The runtime schedules transactions in fixed-size “windows” (slots). • If the window is too small → wasted capacity during high load. • If it’s too large → validators choke on validation and execution.
-
Non-deterministic ordering risks • Parallelism introduces race conditions: two validators might produce different final orderings if execution isn’t tightly controlled. • Solana solves this today by being conservative → which reduces parallel efficiency.
-
Speculative parallel execution • Transactions are allowed to run in parallel even if they might conflict. • If a conflict is detected, DSS rolls back only the colliding transactions (not the whole batch). • This reduces idle cores and wasted cycles from waiting.
-
Adaptive window sizing • Instead of a one-size-fits-all slot, DSS dynamically adjusts batch size based on: • Network conditions (latency, congestion) • Validator hardware capacity (fast nodes can do bigger windows) • Transaction type mix (DeFi vs. NFTs vs. payments) • This keeps the pipeline “right-sized” → smooths out throughput, avoids sudden congestion cliffs.
-
Deterministic ordering preserved • Even with speculation, DSS enforces a canonical order in the final ledger. • All validators converge to the same state → no consensus breaks.
- +15–20% throughput under stress (NFT mints, liquidation cascades).
- Lower latency variance → smoother p95/p99.
- Adaptive scaling based on hardware + network conditions.
- Speculative parallel execution.
- Deterministic final ordering.
- Adaptive execution window sizing.
Author: Kaleb Barnhart X (@xkal3b)
💬 Telegram: @kb441