Stochastic load spike modeling for bitcoin transactions

go run run/main.go [--load <float>] [--bs <float>] [--nb <int>] [--ns <int>]

--load the percentage of the maximum TPS of the bitcoin network. Setting this parameter will run the simulation with this single load for the entirety of the simulation. Abscence of this parameter will cause the simulation to run the designation SpikeProfile described in run/main.go

--bs the maximum block size for the simulation

--nb number of blocks to create in a single iteration, roughly upperbounds the maximum time before transaction confirmation times are unrecorded by the simulation

--ns number of iterations to repeat using the above parameters, higher = more accurate

Custom spike profiles can be defined in the run/main.go file. Spikes are (time, load) pairs where time is the completion percentage [0,1) of the simulation (in terms of block numbers) and load is the percentage [0, infinity) of the maximum TPS (Transactions Per Second) of the network, currently set to 3.5 as per the Bitcoin Traffic Bulletin.

These spikes must occur in increasing order of time, the simulation will validate provided spike profiles and panic if they do not meet the above requirements requirements.

Data generated from each simulation is written to a file named /data/load-spike-%f:%f-%d-%d.cl-dat, where the format specifiers are replaced with the time, load, number of blocks, and number of iterations, respectively.

Each files contains rows corresponding to <bucket-number> | <log-of-txn-confirmation-time> | <probability> | <cumulative-probability>.

python plotter.py will accrue all files in the /data folder with the format load-spike-*.cl-dat and attempt to plot them all in a single chart. The resulting chart is then written to /plots/load-spike-cumulatives.png.

The time_series_logger.go is still in the works, hence the lack of comments and support in the plotter. Our goal is to first mimic the results seen in the Bitcoin Traffic Bulletin before introducing other improvements.