*** fuentess has quit IRC | 02:12 | |
*** sgarzare has joined #kata-general | 08:17 | |
*** auk has quit IRC | 09:43 | |
*** sameo has quit IRC | 12:47 | |
*** pvdp665564 has quit IRC | 12:52 | |
*** sameo has joined #kata-general | 13:03 | |
*** devimc has joined #kata-general | 13:21 | |
kata-irc-bot | <s257117> Hello @jose.carlos.venegas.m, I'm still not able to have an accurate benchmark for the I/O. What I find out is that if using kata-qemu (9p) there is no way to bypass the host cache (actually only the guest page cache is bypassed), instead with kata-qemu-virtiofs the host page cache is not used. Why we have such a behaviour? ```host:``` | 13:32 |
---|---|---|
kata-irc-bot | <s257117> Hello @jose.carlos.venegas.m, I find out that with kata-qemu(9p) is not possible to bypass the host cache whereas with kata-qemu-virtiofs we don't use it. Why we have such a behaviour? Here an example: ```host_cache = 1614 MB container(9p) runs : dd if=/dev/zero of=testfile bs=4k count=262144 oflag=direct host_cache = 2665 MB =========================================================== host_cache = 1614 MB container(virtio-fs) runs: | 13:38 |
kata-irc-bot | dd if=/dev/zero of=testfile bs=4k count=262144 oflag=direct host_cache = 1783 MB``` | 13:38 |
*** fuentess has joined #kata-general | 14:04 | |
*** devimc has quit IRC | 15:19 | |
*** devimc has joined #kata-general | 15:19 | |
*** sgarzare has quit IRC | 17:34 | |
*** fuentess has quit IRC | 19:47 | |
*** fuentess has joined #kata-general | 20:05 | |
*** pvdp665564 has joined #kata-general | 20:29 | |
*** devimc has quit IRC | 22:08 | |
*** CeeMac has quit IRC | 22:26 | |
*** CeeMac has joined #kata-general | 22:30 | |
*** fuentess has quit IRC | 23:24 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!