>mdbox format is a cross between mbox and sdbox. > >The idea is that it keeps up to mdbox_rotate_size sized mbox files >which contain mails. Using zlib here will not help, because zlib is >not applied to the full mdbox file, but individual mails within. > >Also not sure why you think that adding compression would make things >faster? =) atop shows a saturated lvm/disk, some were telling me that enabling compression would alleviate on the disk io's in exchange of higher cpu utilization.
>I am not fully sure how cephfs works but you might get better results >with sdbox format. > I think the one message / one file, is bad for cephfs storage, as I understand ceph creates quite a bit of overhead there. https://www.mail-archive.com/ceph-users@lists.ceph.com/msg48054.html >> On 14 January 2019 at 17:30 Marc Roos <m.r...@f1-outsourcing.eu> wrote: >> >> >> >> I have test environment to determine what would be best settings. I have >> been told that enabling zlib compression would be good to save iops on >> storage. But doing the test now, I get worse results. >> >> [@test2 ~]# pr -m -t mail04-mdbox-vdb-append-64kb-6.log >> mail04-mdbox-vdb-append-64kb-8.log |less >> Logi Sele Appe Logi Sele Appe >> 100% 100% 100% 100% 100% 100% >> 1 1 6 1/ 1 1 1 6 1/ 1 >> 0 0 9 1/ 1 0 0 8 1/ 1 >> 0 0 7 1/ 1 0 0 8 1/ 1 >> 0 0 10 1/ 1 0 0 9 1/ 1 >> 0 0 7 1/ 1 0 0 6 1/ 1 >> 0 0 8 1/ 1 0 0 8 1/ 1 >> 0 0 9 1/ 1 0 0 8 1/ 1 >> 0 0 10 1/ 1 0 0 8 1/ 1 >> 0 0 9 1/ 1 0 0 7 1/ 1 >> 0 0 11 1/ 1 0 0 7 1/ 1 >> 29 3 1093 ms/cmd avg 43 2 1232 ms/cmd avg >> Logi Sele Appe Logi Sele Appe >> 100% 100% 100% 100% 100% 100% >> 0 0 10 1/ 1 0 0 9 1/ 1 >> 0 0 9 1/ 1 0 0 7 1/ 1 >> 0 0 10 1/ 1 0 0 7 1/ 1 >> 0 0 9 1/ 1 0 0 6 1/ 1 >> 0 0 11 1/ 1 0 0 8 1/ 1 >> 0 0 9 1/ 1 0 0 7 1/ 1 >> 0 0 8 1/ 1 0 0 6 1/ 1 >> 0 0 11 1/ 1 0 0 7 1/ 1 >> 0 0 9 1/ 1 0 0 5 1/ 1 >> 0 0 9 1/ 1 0 0 9 1/ 1 >> 0 0 1063 ms/cmd avg 0 0 1413 ms/cmd avg >> Logi Sele Appe Logi Sele Appe >> 100% 100% 100% 100% 100% 100% >> 0 0 8 1/ 1 0 0 8 1/ 1 >> 0 0 10 1/ 1 0 0 8 1/ 1 >> 0 0 7 1/ 1 0 0 8 1/ 1 >> 0 0 10 1/ 1 0 0 6 1/ 1 >> 0 0 10 1/ 1 0 0 7 1/ 1 >> 0 0 11 1/ 1 0 0 8 1/ 1 >> 0 0 7 1/ 1 0 0 8 1/ 1 >> 0 0 11 1/ 1 0 0 6 1/ 1 >> 0 0 8 1/ 1 0 0 7 1/ 1 >> 0 0 10 1/ 1 0 0 8 1/ 1 >> 0 0 1077 ms/cmd avg 0 0 1382 ms/cmd avg >> Logi Sele Appe Logi Sele Appe >> 100% 100% 100% 100% 100% 100% >> 0 0 7 1/ 1 0 0 6 1/ 1 >> 0 0 9 1/ 1 0 0 4 1/ 1 >> 0 0 6 1/ 1 0 0 7 1/ 1 >> 0 0 9 1/ 1 0 0 7 1/ 1 >> 0 0 8 1/ 1 0 0 7 1/ 1 >> 0 0 10 1/ 1 0 0 5 1/ 1 >> 0 0 9 1/ 1 0 0 8 1/ 1 >> 0 0 7 1/ 1 0 0 6 1/ 1 >> 0 0 9 1/ 1 0 0 6 1/ 1 >> 0 0 7 1/ 1 0 0 8 1/ 1 >> 0 0 1246 ms/cmd avg 0 0 1524 ms/cmd avg >> ... >> ... >> ... >> Logi Sele Appe Logi Sele Appe >> 100% 100% 100% 100% 100% 100% >> 0 0 8 1/ 1 0 0 5 1/ 1 >> 0 0 6 1/ 1 0 0 4 1/ 1 >> 0 0 6 1/ 1 0 0 4 1/ 1 >> 0 0 7 1/ 1 0 0 4 1/ 1 >> 0 0 5 1/ 1 0 0 5 1/ 1 >> 0 0 5 1/ 1 0 0 4 1/ 1 >> 0 0 6 1/ 1 0 0 4 1/ 1 >> 0 0 7 1/ 1 0 0 5 1/ 1 >> 0 0 6 1/ 1 0 0 4 1/ 1 >> 0 0 7 1/ 1 0 0 5 1/ 1 >> 0 0 1596 ms/cmd avg 0 0 2308 ms/cmd avg >> Logi Sele Appe Logi Sele Appe >> 100% 100% 100% 100% 100% 100% >> 0 0 7 1/ 1 0 0 4 1/ 1 >> 0 0 1583 ms/cmd avg 0 0 4 1/ 1 >> 0 0 2331 ms/cmd avg >> Totals: >> Logi Sele Appe Totals: >> 100% 100% 100% Logi Sele Appe >> 1 1 1978 100% 100% 100% >> 1 1 1145 >> >> >> >> >> >> >> > >
