Forerunner: Constraint-based Speculative Transaction Execution for Ethereum

Ethereum is an emerging distributed computing platform that supports a decentralized replicated virtual machine at a large scale. Transactions in Ethereum are specified in smart contracts, disseminated through broadcast, accepted into the chain of blocks, and then executed on each node. In this new Dissemination-Consensus-Execution (DiCE) paradigm, the time interval between when a transaction is known (during the dissemination phase) to when the transaction is executed (after the consensus phase) offers a window of opportunity to accelerate transaction processing through speculative execution. However, the traditional speculative execution, which hinges on the ability to predict the future accurately, is inadequate because of DiCE’s many-future nature.

Forerunner proposes a novel constraint-based approach for speculative execution on Ethereum. In contrast to the traditional approach of predicting a single future and demanding it to be perfectly accurate, Forerunner speculates on multiple futures and can leverage speculative results based on imperfect predictions whenever certain constraints are satisfied. Under these constraints, a transaction execution is substantially accelerated through a novel multi-trace program specialization enhanced by a new form of memoization. The fully implemented Forerunner is evaluated as a node connected to the worldwide Ethereum network. When processing 13 million transactions live in real time, Forerunner achieves an effective average speedup of 8.39x on the transactions that it hears during the dissemination phase, which accounts for 95.71% of all the transactions. The end-to-end speedup over all the transactions is 6.06x. The code and data sets are publicly available.