libp2p-test-plans/dht

Plan: The go-libp2p DHT behaves

IPFS can safely rely on the latest DHT upgrades by running go-libp2p DHT tests directly

What is being optimized (min/max, reach)

• (Minimize) Numbers of peers dialed to as part of the call
• (Minimize) Number of failed dials to peers
• (Minimize) Time that it took for the test
• (Minimize) Routing efficiency (# of hops (XOR metric should decrease at every step))

Plan Parameters

• Network Parameters
  • instances - Number of nodes that are spawned for the test (from 10 to 1000000)
• Image Parameters
  • Single Image - The go-libp2p commit that is being tested
  • Image Resources CPU & Ram

Tests

Test: Find Peers

• Test Parameters
  • auto-refresh - Enable autoRefresh (equivalent to running random-walk multiple times automatically) (true/false, default: false)
  • random-walk - Run random-walk manually 5 times (true/false, default: false)
  • bucket-size - Kademlia DHT bucket size (default: 20)
  • n-find-peers - Number of times a Find Peers call is executed from each node (picking another node PeerId at random that is not yet in our Routing table) (default: 1)
• Narrative
  • Warm up
    • All nodes boot up
    • Each node as it boots up, connects to the node that previously joined
    • Nodes ran 5 random-walk queries to populate their Routing Tables
  • Act I
    • Each node calls Find Peers n-find-peers times

Test: Find Providers

• Test Parameters
  • auto-refresh - Enable autoRefresh (equivalent to running random-walk multiple times automatically) (true/false, default: false)
  • random-walk - Run random-walk manually 5 times (true/false, default: false)
  • bucket-size - Kademlia DHT bucket size (default: 20)
  • p-providing - Percentage of nodes providing a record
  • p-resolving - Percentage of nodes trying to resolve the network a record
  • p-failing - Percentage of nodes trying to resolve a record that hasn't been provided
• Narrative
  • Warm up
    • All nodes boot up
    • Each node as it boots up, connects to the node that previously joined
    • Nodes ran 5 random-walk queries to populate their Routing Tables
  • Act I
    • p-providing of the nodes provide a record and store its key on redis
  • Act II
    • p-resolving of the nodes attempt to resolve the records provided before
    • p-failing of the nodes attempt to resolve records that do not exist

Test: Provide Stress

• Test Parameters
  • auto-refresh - Enable autoRefresh (equivalent to running random-walk multiple times automatically) (true/false, default: false)
  • random-walk - Run random-walk manually 5 times (true/false, default: false)
  • bucket-size - Kademlia DHT bucket size (default: 20)
  • n-provides - The number of provide calls that are done by each node
  • i-provides - The interval between each provide call (in seconds)
• Narrative
  • Warm up
    • All nodes boot up
    • Each node as it boots up, connects to the node that previously joined
    • Nodes ran 5 random-walk queries to populate their Routing Tables
  • Act I
    • Each node calls Provide for i-provides until it reaches a total of n-provides