An analytical model for churn process in Bitcoin network with ordinary and relay nodes

Dynamic participation of nodes, also known as churn, is quite common in P2P networks, and Bitcoin network is no exception. The impact of churn becomes more pronounced when nodes with different roles and functionality exist. In this paper we analyze node churn in Bitcoin network where ordinary nodes generate transaction traffic and distribute and verify blocks, while relay nodes that just distribute blocks, but have a larger number of connections. We use the tools of probabilistic analysis and Markov chains, and show that churn of relay nodes has a higher impact on traffic performance than that of ordinary nodes. We introduce two different Continuous Time Markov chains (CTMC) for ordinary and relay nodes to model each type of node’s behavior separately. Our results indicate that node sleep impacts the performance more in networks where the proportion of relay nodes is lower, and also that block and transaction delivery times are higher in those cases. Also, we find that the time needed for a node to synchronize upon rejoining the network after sleep is dominated by the time required for block and transaction verification.