Skip to content

karknu/rndds

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

rndds

Overview

rndds is a distributed application where a fixed set of peers periodically sends a random number between 0 and 1 to all other peers. All messages are ordered using Lamport Timestamps. Sending of new messages stops after a set time (-k argument), and the application waits an additional time (-l argument) for outstanding messages. When exiting each peers prints the timestamp and the random value they received in transmission order.

At the start all peers can talk to each other, but connectivity between some (or possibly all after a while) peers is expected to fail. The cluster will attempt to overcome this by utilizing a Probabilistic Reliable Multicast scheme, as presented by Jun Luo, Patrick Th and Eugster Jean-pierre Hubaux in Probabilistic Reliable Multicast published in Ad Hoc Networks (Ad Hoc Networks 2 2004 volume 2.)

rndds uses TCP as transport protocol. For TCP sockets the 'distributed-process' package provides unreliable in order delivery messages. This is not ideal, instead a transport layer on top of UDP would be preferable. With an UDP transport rndds could detect missing packets directly and start asking other peers for it.

The broadcasted stream of random values is deterministic and based on a seed (-s argument). The peer's 'gossip' functionality is also seeded with values derived from the provided seed, but not completely deterministic. The 'gossip' functionality should differ between the nodes which has the effect that changing the number of nodes will result in different gossip being sent. The 'gossip' functionality also depends on timing of network events.

Building

stack build

Run

rndds uses distributed-process-simplelocalnet to find the initial set of peers. Each peer is started in slave mode, and a master node is used to distribute configuration and start the broadcasting.

  1. Start the slaves For each slave do:
    • stack exec rndds-exe -- -h "ip address to use" -p "port to use"
    • stack exec rndds-exe -- -h "another ip address to use" -p "port to use"
  2. Start the master, and run for 180s. stack exec rndds-exe -- -m -k180

The test directory contains the script simple.sh which can be used to start 10 peers on the local computer. After completion the script verifies that all peers agree on ordering and value of the random numbers sent between them.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published