ATOM implements a novel variant of the Active Object Pattern that uses microthreads instead of native threads, and continuations to model Futures. Microthreads provide massive scalability enabling the creation of thousands of atoms and they considerably reduce communication costs thanks to lightweight context switching. Continu- ations hidden in RPC stubs permit to elegantly mask synchronous calls as asynchronous messages. The resulting code is easier to follow since it respects data-flow semantics.
ATOM also offers advanced abstractions like multi-call invocations, su- pervisors, remote reference passing, and exception handling to ease the implementation of distributed algorithms. ATOM has been implemented in Python over the Stackless Python project. We have validated ATOM against Erlang, Akka and an implementation of ATOM using native threads.
atompy
An Active Object Middleware with Microthreads and Continuations
Status: Abandoned
Brought to you by:
rmondejar
Downloads:
0 This Week