Genes MailLists wrote: > On 06/01/2011 02:20 PM, Bill Davidsen wrote: > as squeezed out if I remember right to make room > >>> for the cookie) ... >>> >>> and therefore some performance degradation when the machine gets busy >>> ... so its never been totally problem free in that sense ... >>> >>> >>> >> Depending on what you do, more than "some." As physical distance goes up and >> speed goes up, the penalty for small window size goes up as well. Pulling a >> TB/day or so from NY to CA I used large window sizes to make it possible. >> >> > As the internet has gotten faster, that was exactly my concern ... > > Note that other tuning is appropriate for making lots of connections with small amounts of data on each, vs. a single socket with very large transfers. We were using multi-GB aggregations, so tuning the initial window size was less of an impact, while it is very important for delivering smaller bytes/socket. -- Bill Davidsen<davidsen@xxxxxxx> We are not out of the woods yet, but we know the direction and have taken the first step. The steps are many, but finite in number, and if we persevere we will reach our destination. -me, 2010 -- users mailing list users@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe or change subscription options: https://admin.fedoraproject.org/mailman/listinfo/users Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
