Date: Fri, 1 Dec 1995 11:20:28 -0500 From: "Garrett A. Wollman" <wollman@lcs.mit.edu> To: Brian Tao <taob@io.org> Cc: questions@FreeBSD.org Subject: Re: What is in_rtqtimo ? Message-ID: <9512011620.AA13050@halloran-eldar.lcs.mit.edu> In-Reply-To: <Pine.BSF.3.91.951201012600.13078D-100000@flinch> References: <9511301605.AA08576@halloran-eldar.lcs.mit.edu> <Pine.BSF.3.91.951201012600.13078D-100000@flinch>
next in thread | previous in thread | raw e-mail | index | archive | help
<<On Fri, 1 Dec 1995 01:28:25 -0500 (EST), Brian Tao <taob@io.org> said:
>[I wrote:]
>> The parameters are all tunable, in the net.inet.ip branch of the MIB:
>>
>> net.inet.ip.rtexpire = 3600 # current max lifetime
>> net.inet.ip.rtminexpire = 10 # minimum max lifetime
>> net.inet.ip.rtmaxcache = 128 # how many is ``too many''
> On servers that see a lot of connections from different hosts
> coming and going (like on our IRC server), would turning up the
> rtexpire and or rtmaxcache values be a good thing to do?
It's difficult to say; it depends a lot on whether any useful
statistics are getting saved in the cache. If you see a significant
amount of Transaction TCP traffic, increasing rtmaxcache or
rtminexpire might help. Be aware that every entry costs on the order
of 256 bytes of memory, which is why we make an effort to dynamically
limit them in the first place.
You can tell whether you are getting useful statistics by looking at
the output of `netstat -p tcp' and searching for the following:
181 connections closed (including 19 drops)
27 connections updated cached RTT on close
27 connections updated cached RTT variance on close
5 connections updated cached ssthresh on close
You can examine the statistics by using the `route get' command; for
example:
------------------------------------
root@khavrinen(17)# route get freefall.freebsd.org
route to: freefall.FreeBSD.ORG
destination: freefall.FreeBSD.ORG
gateway: radole
interface: ed0
flags: <UP,GATEWAY,HOST,DONE>
recvpipe sendpipe ssthresh rtt,msec rttvar hopcount mtu expire
16384 16384 1656 438 500 0 1500 3538
------------------------------------
The ssthresh value is the switchover point between linear and
exponential TCP window expansion; rtt is the estimated round-trip
time; and rttvar is the ``variance'' of the rtt measurement. (I put
`variance' in quotes because the value is not computed as a true
statistical variance, but rather as an average deviation.) A correct
ssthresh value can substantially increase the performance of short
connections; rtt and rttvar come into play when determining how long
to wait for an acknowledgement before timing out and retransmitting
the packet.
-GAWollman
--
Garrett A. Wollman | Shashish is simple, it's discreet, it's brief. ...
wollman@lcs.mit.edu | Shashish is the bonding of hearts in spite of distance.
Opinions not those of| It is a bond more powerful than absence. We like people
MIT, LCS, ANA, or NSA| who like Shashish. - Claude McKenzie + Florent Vollant
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9512011620.AA13050>