there is a lot of literature and practice in this area. (it is close enough to a distributed hash table.)
there are a LOT of failure modes that we've seen in practice and i would think you'd want to cover those. the things 0mq help with here are not the problem; the main issues are how to start (discovery) and how to repair the consistency of teh whole table after things like network splits and node failures. On Feb 24, 2011, at 7:01 AM, Michiel ten Hagen wrote: > All, > > My current project is a distributed ' latest value' system where 'items' > have several key/value pairs on which updates are received. Several > architectural principles are the same as described in your whale shark > system, I will follow closely and if possible might contribute. > > Is the goal to provide a base layer on which distributed stream processing > can be done or will be a complete processing system such as boreals > http://www.cs.brown.edu/research/borealis/public/? > > A base framework like 'whaleshark' could be beneficial if it does not impose > too many restrictions on the applications on top. > > Michiel ten Hagen > > > -- > Regards, > > Michiel ten Hagen > > _______________________________________________ > zeromq-dev mailing list > [email protected] > http://lists.zeromq.org/mailman/listinfo/zeromq-dev ------------------ Andrew Hume (best -> Telework) +1 623-551-2845 [email protected] (Work) +1 none currently AT&T Labs - Research; member of USENIX and LOPSA
_______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
