You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@storm.apache.org by jeff saremi <je...@hotmail.com> on 2014/07/03 21:05:35 UTC

in-memory versus serialized

What can be said about the same memory being used for a topology? under jvm and under multi -lang?

let me elaborate please:
- if i manage to contain my topology to only one machine while maximizing the utilization of cores in that machine will i be able to speed up the processing by eliminating the object/tuple copying?
If that is even possible in one JVM in Java what happens if i use other languages which are "process"-based? I guess the memory mapped files would still work?
Has anyone had similar experiences or thoughts? thanksJeff