At the moment I have two VPS servers, I'll name them Xen and Hyper for this discussion. Both of them have Centos 4.5 as the operating systems Xen uses Apache and Bind. Has 256mb Ram and 544mb swap partition (no burst ram) Hyper uses Lightpd and Tiny Dns. Has 256mb Ram and 128mb Burst Ram (no swap) Now, I've recently moved my sites over to Hyper, And I estimate due too promotion that my traffic has doubled since the move. I've been having a stability problem with Lightpd and thanks to some help from some guys Here we have established that I need more ram on this machine. Now getting to the point: Xen was my first VPS, From the first time I ever logged in, with no dns/domains/traffic the ram usage was in the orange. about 150mb from memory. I didn't know how much ram the op system is supposed to use etc. But now that I have Hyper when I first logged in there was only about 70mb used. I realize that there are vast differences between these two machines though, And this is the reason for my post. There is still one site hooked up to Xen the moment, but it's a very small forum with practically no traffic. Logging in now, Lxadmin reports ram usage of 224mb the output from 'ps aux' for Xen is: USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND root 1 0.0 0.2 2560 540 ? S Sep06 0:00 init [3] root 2 0.0 0.0 0 0 ? S Sep06 0:00 [migration/0] root 3 0.0 0.0 0 0 ? SN Sep06 0:00 [ksoftirqd/0] root 4 0.0 0.0 0 0 ? S< Sep06 0:00 [events/0] root 5 0.0 0.0 0 0 ? S< Sep06 0:00 [khelper] root 6 0.0 0.0 0 0 ? S< Sep06 0:00 [kthread] root 7 0.0 0.0 0 0 ? S< Sep06 0:00 [xenwatch] root 8 0.0 0.0 0 0 ? S< Sep06 0:00 [xenbus] root 12 0.0 0.0 0 0 ? S< Sep06 0:00 [kblockd/0] root 29 0.0 0.0 0 0 ? S Sep06 0:00 [pdflush] root 30 0.0 0.0 0 0 ? S Sep06 0:00 [pdflush] root 31 0.0 0.0 0 0 ? S Sep06 0:00 [kswapd0] root 32 0.0 0.0 0 0 ? S< Sep06 0:00 [aio/0] root 174 0.0 0.0 0 0 ? S Sep06 0:00 [kseriod] root 281 0.0 0.0 0 0 ? S Sep06 0:00 [kjournald] root 454 0.0 0.0 0 0 ? S< Sep06 0:00 [kauditd] root 793 0.0 0.1 2416 428 ? S<s Sep06 0:00 udevd root 921 0.0 0.0 0 0 ? S Sep06 0:00 [kjournald] root 1247 0.0 0.2 3364 544 ? Ss Sep06 0:00 syslogd -m 0 root 1251 0.0 0.1 2096 384 ? Ss Sep06 0:00 klogd -x named 1276 0.0 1.1 37660 2944 ? Ssl Sep06 0:00 /usr/sbin/named - rpc 1322 0.0 0.2 1944 532 ? Ss Sep06 0:00 portmap rpcuser 1341 0.0 0.2 1860 712 ? Ss Sep06 0:00 rpc.statd root 1413 0.0 0.1 5240 376 ? Ss Sep06 0:00 rpc.idmapd root 1525 0.0 0.4 5916 1164 ? Ss Sep06 0:01 /usr/sbin/sshd root 1538 0.0 0.3 2752 796 ? Ss Sep06 0:00 xinetd -stayalive root 1551 0.0 0.1 2364 448 ? S Sep06 0:00 /usr/lib/courier- root 1553 0.0 0.0 2928 212 ? S Sep06 0:00 /usr/lib/courier- root 1560 0.0 0.1 3372 444 ? S Sep06 0:00 /usr/lib/courier- root 1563 0.0 0.0 2432 212 ? S Sep06 0:00 /usr/lib/courier- root 1567 0.0 0.1 2872 468 ? S Sep06 0:00 /usr/lib/courier- root 1569 0.0 0.1 3164 396 ? S Sep06 0:00 /usr/lib/courier- root 1575 0.0 0.1 2660 448 ? S Sep06 0:00 /usr/lib/courier- root 1577 0.0 0.0 1704 208 ? S Sep06 0:00 /usr/lib/courier- root 1616 0.0 0.1 1936 336 ? Ss Sep06 0:00 gpm -m /dev/input root 1897 0.0 0.4 5768 1236 ? S Sep06 0:00 /bin/sh /usr/bin/ mysql 1943 0.0 2.0 25112 5372 ? Sl Sep06 0:14 /usr/libexec/mysq qmails 1969 0.0 0.1 2864 392 ? S Sep06 0:00 qmail-send qmaill 1976 0.0 0.1 2160 420 ? S Sep06 0:00 splogger qmail root 1977 0.0 0.1 2952 328 ? S Sep06 0:00 qmail-lspawn |pre qmailr 1978 0.0 0.1 2568 348 ? S Sep06 0:00 qmail-rspawn qmailq 1981 0.0 0.1 2500 324 ? S Sep06 0:00 qmail-clean root 1982 0.0 0.3 6280 928 ? Ss Sep06 0:02 crond root 2000 0.0 0.1 2380 416 ? Ss Sep06 0:00 /usr/sbin/atd dbus 2013 0.0 0.3 3132 796 ? Ss Sep06 0:00 dbus-daemon-1 --s root 2035 0.0 0.8 5080 2112 ? Ss Sep06 0:00 hald root 2042 0.0 0.1 2256 460 xvc0 Ss+ Sep06 0:00 /sbin/agetty xvc0 root 3739 0.0 5.6 27752 14840 ? Ss Sep06 0:00 /usr/sbin/httpd root 4731 0.0 0.8 7564 2320 ? Ss 09:45 0:00 sshd: root@pts/0 root 4733 0.0 0.5 5740 1384 pts/0 Ss+ 09:45 0:00 -bash lxlabs 4838 0.0 0.6 4532 1820 ? S Sep06 0:00 /usr/local/lxlabs apache 5589 0.0 4.7 27884 12376 ? S 10:03 0:00 /usr/sbin/httpd root 11829 0.0 0.8 8212 2300 ? Ss 12:18 0:00 sshd: root@pts/1 root 11831 0.0 0.5 6096 1376 pts/1 Ss 12:18 0:00 -bash root 11956 0.0 0.2 3004 768 pts/1 R+ 12:20 0:00 ps aux root 23130 0.0 0.7 8996 1992 ? Ss 04:02 0:00 cupsd apache 23178 0.0 4.9 28128 13028 ? S 04:02 0:00 /usr/sbin/httpd apache 23179 0.0 4.9 28124 13004 ? S 04:02 0:00 /usr/sbin/httpd apache 23180 0.0 4.9 28128 13028 ? S 04:02 0:00 /usr/sbin/httpd apache 23181 0.0 4.9 28128 13040 ? S 04:02 0:00 /usr/sbin/httpd apache 23182 0.0 4.9 28120 13020 ? S 04:02 0:00 /usr/sbin/httpd apache 23183 0.0 4.9 28112 13000 ? S 04:02 0:00 /usr/sbin/httpd apache 23184 0.0 4.9 28164 13056 ? S 04:02 0:00 /usr/sbin/httpd apache 23185 0.0 4.9 28124 13044 ? S 04:02 0:00 /usr/sbin/httpd apache 23657 0.0 4.9 28164 13088 ? S 04:47 0:00 /usr/sbin/httpd apache 23658 0.0 4.9 28120 13012 ? S 04:47 0:00 /usr/sbin/httpd apache 23659 0.0 4.9 28132 13040 ? S 04:47 0:00 /usr/sbin/httpd Code (markup): With XEN ram is supposed to be pre allocated so... Where has my ram gone? lol And should there be so many instances of apache? Or... is this normal for a XEN setup? Rebooting Xen Now to see Ram usage once it restarts... Its back with only 132mb used (this will slowly creep up again though) showing 8 instances of Apache. Thanks in advance for any help in making sense of this
I've seen Apache use almost all of my memory under my dedicated server. Maybe Apache and Lighttpd has something to do with your issue here.
Thanks yyahmee, Maybe this is normal for Apache? Maybe that's why I have such a huge ram usage difference between the two. In anycase I think it's obvious that I will need to get more ram. But I'd really like to know if this is normal for a XEN and Apache VPS with lxadim.
HP's OpenVZ and Xen comparison showed that Xen used more RAM then OpenVZ but I don't know by how much. Try putting Lighttpd and see if the RAM usage changes
hmm... I am not sure if I can do this with lxadmin. I'll have a look though the OpenVZ one is the one that I'm keeping, But I still want to get a handle on this, so that I know what's normal
that could be a clue. On the Apache Machine (no traffic): total used free shared buffers cached Mem: 256 144 111 0 20 64 -/+ buffers/cache: 60 195 Swap: 543 0 543 Code (markup): And the Lightpd Machine (with traffic): total used free shared buffers cached Mem: 325 143 181 0 0 0 -/+ buffers/cache: 143 181 Swap: 0 0 0 Code (markup): So maybe this has something to do with buffers/cache.
Ok... I found out how to swap my new vps to Apache And all my ram is gone on this machine as well... lol Which makes me just a little embarrassed about this thread Well, I've got to learn somehow... and before I had nothing to compare it with. So, for any other Noobs out there thinking about a VPS for proxy hosting... 256mb ram, even with another 128mb burst is just not enough!
The purpose of RAM is to avoid use of swap and reduce HDD access as much as possible. A good OS / server environment makes use of all RAM whenever possible. Hence the more RAM you have - the more RAM will be "gone" ( used ) - at the benefit of reduced access times for data and reduced wear and tear for hard drives. example: on my "workhorse" -laptop, an acer ferrari 5000, i have 4 GB ram, I work a lot and have plenty of difference applications open at almost any given time. > free -m total used free shared buffers cached Mem: 3965 3909 56 0 262 1953 -/+ buffers/cache: 1694 2271 Swap: 5145 0 5145 on my server with quiet some traffic ( around 9GB / day with currently daily average of 559303 hits / day I have > free -m total used free shared buffers cached Mem: 2011 1721 289 0 89 1277 -/+ buffers/cache: 355 1656 Swap: 2047 0 2047 Hence in both cases most of the RAM used ... = success! both machines are fast and I will definitely have a much larger RAM on my next generation server to speed things up even more and to reduce HDD access further. hence for you the RAM used is just 1 factor - the most important however is the access speed / page load speed from your server - including mySQL, CPU and other data to serve pages. the only reliable comparison data thus MUST include a test of a sample page loaded under precise condition - that of course also needs to consider your local LAN / WLAN / ISP environment!
Your running a lot more than just web server + php. Theres a mail server, mysql, bind dns (named), and so on, which all consume a lot of memory. If you strictly ran Lighttpd + PHP + xcache, that would be plenty of RAM for a proxy site. A bare bones CentOS machine uses maybe 50MB of RAM.