Re:A generalised solution to distributed consensus
by pathseeker on Saturday March 09, 2019 @12:47AM
(#19344706)
Can you explain how? In many cases when a "generalized result" comes out there was already a generalized result that covered all but obscure corner cases.
[ Parent
| on HN ]
Re:A generalised solution to distributed consensus
by DannyBee on Saturday March 09, 2019 @01:51AM
(#19344875)
This is the first paper i've seen to generalize consensus while only using immutable state.
That is, you can describe PaxOS, Raft, etc using the generalization given here.
The result is fairly understandable.
(Of course, as mentioned, Raft is trivial in her framework because there is only one thing deciding values, Raft is spending it's time in the leader election part to find that one thing)
Re:A generalised solution to distributed consensus
by DennisP on Saturday March 09, 2019 @09:13AM
(#19346169)
Also, from the abstract, "the surprising result of this analysis is a substantial weakening to the quorum requirements of these widely studied algorithms."
Re:A generalised solution to distributed consensus
by thethirdwheel on Sunday March 10, 2019 @03:39PM
(#19353912)
The big thing here is that its analytical framework allows the authors to clearly analyze Paxos and some important variants and identify performance improvements (e.g. steps in the algorithm you can elide) that don't undermine the guarantees provided.