parallel execution
Running multiple transactions at the same time on different processor cores, which requires knowing in advance what state each transaction touches to avoid conflicts.
As defined in Solana: Faster, Cheaper, More Scalable
Where it comes up · 1
Most important claims · 6 of 6Search all claims for “parallel execution” →
- The only known solution to safe parallel transaction execution is for each transaction to declare in advance which parts of global state it will touch, allowing non-overlapping transactions to run concurrently and overlapping ones to run serially.Solana: Faster, Cheaper, More Scalable
- Solana runs transactions natively on modern NVIDIA graphics cards with thousands of cores, so a card with about 4,000 cores can execute thousands of jobs in parallel, versus the handful of cores a laptop uses for serial code.Solana: Faster, Cheaper, More Scalable
- Solana enables parallel execution because each transaction header specifies which part of global state it will touch, so the system can run non-overlapping transactions simultaneously on GPUs and serialize only those with dependencies.Solana: Faster, Cheaper, More Scalable
- Within a window of seconds to a minute, most transactions in a block do not depend on one another, so they can be executed in parallel, with chronology mattering only for the minority that touch shared state.Solana: Faster, Cheaper, More Scalable
- Writing parallel transaction processing is hard because low-level resource management is difficult, and few engineers know it, so Ethereum's team, lacking distributed-systems backgrounds, defaulted to serial execution as Windows and Mac took three to four years to support multi-core.Solana: Faster, Cheaper, More Scalable
- Solana's ability to run parallel execution comes from an execution-layer design choice rooted in its team's high-performance and distributed systems background, not from its consensus algorithm or proof of history.Solana: Faster, Cheaper, More Scalable