This is the first and OG implementation of RainTree used to prove out the concept implemented by @andrewnguyen22.
This simulator executes raintree WITHOUT the cleanup layer to better represent the coverage before the extra redundancy.
config.json
{
"NumberOfNodes": the 'starting' number of nodes in the simulated network - as you can run multiple simulations at once,
"EndingNumberOfNodes": the 'ending' number of nodes in the simulated network - as you can run multiple simulations at once,
"DeadNodePercentage": % of nodes not responding/propagating,
"FixedDeadNodes": if true, use the FixedDeadNodesIndexArray inorder to kill specific node as certain indices,
"FixedDeadNodesIndexArray": if FixedDeadNodes is on, which nodes would you like to kill?,
"RandomizePartialAddressBooks": would you like to randomize the partial address books? If not, though they'll be partial, they'll be in a fixed order,
"ViewershipPercentageFixed": would you like to have a fixed viewership curve? if true, use ViewershipCurveArray below to set specific values,
"ViewershipCurveArray": if ViewershipPercentageFixed=true, set a specific curve of viewership for each node. Ex. [90,80,70] means 90%, 80%, 70& viewership respectively ,
"TargetPartialViewershipPercentage": the global median viewership percentage of partial addressbooks,
"PartialViewershipStdDev": how far do you want the global std deviation to be from the TargetPartialViewershipPercentage median,
"InvertCurve": invert the viewership curve for fun doomsday scenarios,
"RedundancyLayerRightOn": turn on the right side redundancy layer (not the cleanup layer),
"RedundancyLayerLeftOn": turn on the left side redundancy layer (not the cleanup layer),
"MaxHotlist": how many nodes on the 'hotlist' for first level - initial propagation,
"ShowIndividualNodeResults": show individual node results in the results.json/csv,
"ShowIndividualNodePartialAddressBooks": show individual node addrbooks in the results.json/csv,,
"ResultFileOutputName": the prefix of the .json/.csv output files,
"OriginatorIndex": the initial sender index, use -1 for random
}
results.json
{
"NumberOfNodes": how many nodes in the network?,
"Levels": what was the number of levels in the network?,
"AverageRedundancy": what was the avg number of messages received?,
"NonDeadCoveragePercentage": what percentage of the network was hit before the cleanup layer,
"DeadCount": how many dead nodes?,
"ConsecutiveLevelZeroMatrix": {
how many consecutive zeroes before the cleanup layer: how many occurences
},
results.csv
Nodes,Levels,Comms,Redundancy,Coverage,Missed,LongestMiss
Number of nodes, number of levels, number of (total) communications, avg redundancy, NonDeadCoveragePercentage, how many missed?, longest consecutive miss?
This is an alternative implementation of RainTree that was implemented by @Olshansk to research, understand, and evaluate RainTree.
This was originally proposed in pokt-network/pocket#316, but we decided that it does not belong in the official V1 implementation.
- Alternative approach to implementation: this uses Breadth First Search to queue up and simulate the network propagation.
- Research Collaboration: Python is a "more friendly" language when collaborating with the research community.
- Evaluation: Visualization & evaluation libraries are more readily available in Python
See python/README.md for more details on how to use this library