blockchain-simulator

An experimental blockchain simulator written in python. Current algorithms available for testing are the Longest Chain Protocol and GHOST. Current algorithms in development include Longest Chain With Pool (i.e. Prism v1) and Prism.

Requirements

• Python 3
• graph-tool

Description

Parameters:

• Number of nodes (Note: current network model has all nodes connected to one another)
• Number of adversaries
• Maximum number of transactions per block
• Transaction error probability
• Fork choice rule
  • Longest chain protocol - add block to longest current chain
  • GHOST protocol - add block to heaviest subtree
• Block proposal rate
• Transaction dataset
• Poisson - a transaction occurs based on poisson distribution with rate 1 transaction/second
• Deterministic - a transaction happens at fixed timestamps
• Transaction scheduling rule
• FIFO
• Network model
  • Decker-Wattenhorf - based on “Information Propagation in Bitcoin Network” (https://ieeexplore.ieee.org/document/6688704/), we determine that the delay should be approximately 19/600 sec/tx * txs + 1 sec. Then, we use this delay as the parameter for an exponential distribution (https://docs.scipy.org/doc/numpy/reference/generated/numpy.random.exponential.html)
  • Constant Decker-Wattenhorf - for testing purposes. Same as above without randomness of exponential distribution
• Duration of simulation in seconds
• Logging - enable or disable logging

Overall Flow of Simulation

• 1) Create nodes
• 2) Generate transaction dataset
• 3) Generate proposal events
• 4) Run simulation until all proposals and transactions exhausted.
  • Check for next event in simulation (proposal or transaction)
  • If transaction:
  • Source node broadcasts transaction to all neighboring nodes with factored delay. Neighboring nodes add to their ‘buffer’
  • If proposal:
  • Choose random node
  • Random node creates a proposal
  • Node processes its buffer by adding received transactions into its local queue and adding received proposals into its local blocktree
  • Create a new block limited by number of transactions per block or timestamp
  • Broadcast created proposal to rest of network
• 6) Compute metrics and log results
• 7) Metrics are computed: main chain arrival timestamp, finalization timestamp
  • Main chain arrival timestamp: The main chain consists of blocks common to ALL nodes in the network. When a block on the main chain is proposed, all the tx’s in that block ‘arrive’
  • Finalization timestamp: When a block is proposed and ends up on the main chain, the block k blocks above is finalized. The parameter k is a function of the number of adversaries and the number of nodes

Logged Results

Blocks.csv

• Block id
• Block parent id
• Proposal timestamp - when the block was initially proposed
• Finalization timestamp - when the block was finalized on global blocktree
• Depth - the block’s depth along the main chain
• Finalized - whether or not the block was finalized
• Transactions - ids of all transactions in the block

Stats.csv

• All the parameters
• Time elapsed (real time, NOT simulation time)
• Network latencies for blocks and transactions
• Finalization depth
• Number of blocks
• Length of main chain
• Fraction of main blocks
• Expected fraction of main blocks computed according to 1.0/(1+proposal_rate*block_delay)
• Fraction of orphan blocks
• Expected fraction of orphan blocks computed according to 1-1.0/(1+proposal_rate*block_delay)
• Expected arrival rate of blocks onto main chain
• Expected finalization latency of a block

Transactions.csv

• Id
• Source node
• Generated timestamp
• Main chain arrival timestamp (on global blocktree)
• Finalization timestamp (on global blocktree)

Blocktree.png

Visualization of longest chain of blocktree’s main chain created via graphtool library