Hi, we have recentily upgraded the swift cluster hosting images for wikimedia from folsom to icehouse and it has been very smooth! thank you! After the upgrade we have noticed a substantial increase in the bandwidth used by backends, in particular it seems to line up with the upgrade of frontends to icehouse (which happened last). You can see the effect I'm talking about here together with the upgrade timeline: https://wikitech.wikimedia.org/wiki/Swift/Icehouse#post-upgrade It seems to coincide with an increase in CPU on the backends too: http://ganglia.wikimedia.org/latest/?r=custom&cs=07%2F07%2F2014+00%3A00+&ce=07%2F21%2F2014+00%3A00+&m=cpu_report&s=by+name&c=Swift+eqiad&h=&host_regex=&max_graphs=0&tab=m&vn=&hide-hf=false&sh=1&z=small&hc=4 I was wondering what has changed that could cause that increase and if it is expected, also what would be a good way to graph replication bandwidth vs "serving" bandwidth? that'd be useful too I think. thanks in advance! filippo -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20140721/a7a6cd02/attachment.html>