Date: Wed, 19 Sep 2012 19:18:06 +0400 From: Gleb Smirnoff <glebius@FreeBSD.org> To: Ivan Alexandrovich <ivsan@ngs.ru> Cc: freebsd-net@FreeBSD.org Subject: Re: getting counters for a plenty of vlan ifaces Message-ID: <20120919151806.GK85604@FreeBSD.org> In-Reply-To: <web-25219141@mx16.intranet.ru> References: <web-25219141@mx16.intranet.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Sep 16, 2012 at 09:41:19PM +0700, Ivan Alexandrovich wrote: I> Hi I> I> We are running freebsd9.0 on a router with I> more than 1000 of subscriber's vlan interfaces. I> Outgoing packet rate is approximately 40 kpps. I> I> There's a need to collect bytes and packets I> counters for all those vlan interfaces every I> minute (or even twice a minute) and store them I> in a plain text file: I> <em1.757> <bytes_in> <bytes_out> <pkts_in> <pkts_out> I> <em1.761> <bytes_in> <bytes_out> <pkts_in> <pkts_out> I> ... I> Also I'd like to copy the whole arp table I> into a file (not so frequently). I> I> Our observations show that using common tools I> like a snmp daemon can create a significant I> CPU load. If I'm not mistaken this is due to I> high rate of context switches that are need I> to access kernel data from the userspace. That's strange. Per-minute read shouldn't induce large CPU load. What snmp daemon do you use? I remember that several years ago net-snmp daemon from ports used a single linked list for all ARP entries, and thus it consumed a lot of CPU when receiving a single ARP change from routing socket. Don't know whether this is still valid. And I hope in base bsnmpd doesn't have such problem. -- Totus tuus, Glebius.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20120919151806.GK85604>