Yeah - you slice from socket to pipe, then from pipe to the file descriptor. It keeps all the data kernel-side and, in page-aligned cases, may be able to remap data directly from the NIC if it supports DMA.

Some esnet folks were able to double single-stream performance using this technique:

http://www.es.net/assets/pubs_presos/eScience-networks.pdf

However, they went from 22Gbps to 39Gbps -- may indicate there's something else hidden here.

—
Reply to this email directly or view it on GitHub.

Use REPLY-ALL to reply to list

To unsubscribe from the XROOTD-DEV list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-DEV&A=1