Distributed Replication Simulator

Simulator to build and test causal consistency algorithms. Still very much a WIP.

TODO - Simulator:

• Client simulation
  • Configuration - Reading values from config file
  • Operations being sent in a closed loop
  • Distinction between read/write operations
  • Individual Clients
  • Access specific keys to prepare for partial replication
• Configuration - Create an easy-to-read sample config file
• Collecting Metrics
  • Latency
  • Operations/s
  • Visibility Times
• Heap / Scheduler change - A node can only receive one operation on a specific timestamp.
• Transport - Latency Matrix / Throughput into consideration
  • Read and process latency matrices to a format acceptable for Peersim
  • Generate an "optimized" tree graph based on it - Python Script (altered Prim algorithm basically)
  • Import the matrices into Peersim, the Latency ones using E2ENetwork/E2ETransport, and the Tree Overlay using WireFromFile.
  • Node Bandwidth Support
• Protocol - Base Meta-protocol
  • Internal event queue for non processed events
  • Changes to account for difference between reads/writes
  • Base protocolMessage container class getSize() for Transport etc...
  • New layer to add message processing time support instead of being done through the Heap
  • Partial Replication support
    • Each node includes the list of Partitions of every node.
    • Operation propagation according to Partitions
    • Reads / Writes accessing correct partition
    • Migrations
• Operation propagation through the system
  • AllToAll Broadcast and Overlay Init - 1 Click
  • Saturn Tree Propagation

TODO - Protocols

• C3
• Saturn
• Cure or Gentlerain

Running and Compiling the Simulator:

Compile: mvn compile assembly:single

Run: java -cp target/peersim.jar peersim.Simulator config/config.txt

Latency Matrix

• Africa Cape Town (0)
• US West (N. Cal) (19)
• Asia Pacific Sydney (7)
• Canada (8)
• EU Paris (14)
• EU London (13)
• São Paulo (16)