> Yes, everyone just seems to transfer in chunks.

Just to add info, that's because the current thrift implementation needs 
to buffer the entire content in memory.  So a "large file" would consume a 
"large" amount of memory in your process.  Even worse, a single block of 
contiguous memory, most likely.

        /r$

--
STSM, WebSphere Appliance Architect
https://www.ibm.com/developerworks/mydeveloperworks/blogs/soma/

Reply via email to