> 1. The performance problem (along the CPU/memory tradeoff) happens on 
> process level (process consuming too much memory etc.) not on library 
> level. All the libraries loaded to the process contribute to the 
> problem. Thus, the solution should be preferably at process level.

Yes and no.  Some libraries contribute more to the problem, because they 
allocate more memory.  (Or, allocate memory more frequently.)  In my case, 
where I'm processing millions-to-billions of records, it's the per-record 
allocation overhead that I care about.  And so that points to two libraries in 
particular: the data serialization library (Avro), and the transport library 
(zmq).  I only have to get those two libraries to use the different allocator 
to see a performance win.

–doug

Attachment: PGP.sig
Description: This is a digitally signed message part

_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Reply via email to