Melvin's digital garden

Distributed computing for new blood

[2014-10-18 Sat 10:14] speaker: Raymond Tan, HP Labs event: Geekcamp 2014 ** definition of distributed computing overcome time for information to travel dealing with failure of networks/servers ** why do we need it?

  • scalability
  • availability ** related work
  • distributed objects, eg Java RMI
  • remote procedure call
  • distribteud shared mutable state ** FLP impossibility result In an async model, where one processor might fail, there is no distributed algorithm to solve the consensus problem. ** CAP theorem C - linearizability A - every request to a non-failing node must complete successfully P - network partition can happen

Practical systems are CP or AP. Other forms of consistency possible. ** to scale, partition consistent hashing ** to be resilient, replicate need to solve consensus

Links to this note