Performance Analysis of the Solana Turbine Protocol : A Simulation-Based Approach

Abstract

Solana's blockchain network accommodates thousands of nodes that require fast and scalable propagation. The Turbine protocol is a critical mechanism of the Solana's blockchain that deals with data availability within the network. Turbine splits data into smaller packets called shreds and delivers them efficiently, reducing network congestion and latency while maintaining low block confirmation times. This thesis investigates the efficiency and trade-offs of different Turbine data distribution strategies via simulation of three different models: a Random model, a Greedy Oracle model, and a Weighted Shuffle model. These models, written in Python, implement different propagation behaviors under varying levels of network awareness and optimization strategies. Outcomes demonstrate that while the Greedy Oracle and Weighted Shuffle models offer performance benefits, the Random model has significant performance limitations. Though the Greedy Oracle strategy is impossible to implement in a real network, the results show that the Weighted Shuffle strategy performs comparably in certain tests.