Hi Sven, > after reading all that statements, that SQL databases cannot handle the > amount of flows of a "normal" line, I am interested about the facts. So > I would like to start a little survey, how many rows of flows per time > which SQL db can handle on which kind of hardware and in which > configuration (or not).
In the interests of comparability, I invented a little benchmark: [EMAIL PROTECTED] chris]$ vi pmacct-0.11.0/sql/pmacct-create-db_v6.mysql (remove the lines specifying the pmacct database to avoid dropping it) [EMAIL PROTECTED] chris]$ cat pmacct-0.11.0/sql/pmacct-create-db_v6.mysql | mysql test -u root -p<password> [EMAIL PROTECTED] chris]$ time perl -e 'print "insert into acct_v6 (ip_src,ip_dst) values (rand(),rand());" x 10000' | mysql test -u root -p<password> real 0m17.149s user 0m1.373s sys 0m1.036s This works out at about 583 rows/second, or 50 million rows per day. Setup: Celeron 366 128 MB RAM Single IDE hard disk (ancient 20 GB) Linux 2.6.10-1.771_FC2 MySQL 5.0.26-standard We are using this machine to monitor, firewall and NAT an ADSL line, 8 mbit down, 384 kbit up (approx). The machine has crashed once due to load from pmacctd threads, but that was when it was running a 2.4 kernel on FC1, and it has since been upgraded to FC2. Cheers, Chris. -- (aidworld) chris wilson | chief engineer (http://www.aidworld.org) _______________________________________________ pmacct-discussion mailing list http://www.pmacct.net/#mailinglists
