At Friday 09:31 AM 2/4/00 , Mike Horwath wrote:
>On Fri, Feb 04, 2000 at 09:19:56AM -0500, Brandon Applegate wrote:
> > We are getting avg. 25 msec on ping tests from the following link:
> >
> > pa-t3 <---- 600 mile DS3 (clear channel, full) ----> pa-t3
> >
> > Packet size does not affect our times.
> >
> > Can anyone give me their $ .02 on this ?? I am mainly concerned with the
> > delay induced by the circuit itself, as we are able to tweak the router
> > config on each end.
>
>25ms isn't all that bad.
>
>You are traveling 600 miles and the speed of light is still an
>unbroken rule.
>
>If I did my math right, the minimum latency is 3.17ms based on 189Kmph
>for light through a vacuum.
>
>Add in hardware issues and switching issues and 25ms doesn't seem all
>that bad at all.
Hardware and serialization should come in at a cool 2ms
at most. The speed of light in fiber is not exactly light speed.
http://www.slac.stanford.edu/comp/net/wan-mon/tutorial.html has
a reference using 0.6*c :
600mi*2 / (0.6 * 189,000 miles/s) = 0.0106 s = 10.6ms . And that's RTT!
Knowing how strange carriers are routing circuits, its a safe guess
to assume that they are routing your DS3 through a very long detour.
Given that DS-3 TCP session performance is just atrocious when you
go from say: 15 to 25 ms, you might have wanted to write the maximum
acceptable propagation delay into your contract with the carrier.
I seem to recall a NANOG posting a few years ago where a T1 route from
Denver to SF encompassed a detour through Kansas City. You can take
a good guess which carrier was involved. Moan and bitch at your
provider for routing your DS3 through neverneverland and exposing
you to an unduely increased risk of an outage (more fiber miles =
more chance of backhoe-attack).
Performance hit by going from 15 to 25ms RTT for single TCP session:
bandwidth delay product goes from roughly 84 to 140 KBytes - this
is the size your window must have (at the least) to get 45Mbps
throughput on a per-session basis, which you would have to set on
both servers and clients that you are using across the link.
If this is a link transporting public IP traffic, you can safely assume
that all Windows machines using it have a measly 8K window, throttling
the performance to roughly 8K per RTT (in reality its slightly more),
yielding you 533KB/s @15ms and 320KB/s @25ms.
bye,Kai
-- kai@conti.nu "Just say No" to Spam Kai Schlichting Palo Alto, New York, You name it Sophisticated Technical Peon Kai's SpamShield <tm> is FREE! http://SpamShield.Conti.nu | | LeasedLines-FrameRelay-IPLs-ISDN-PPP-Cisco-Consulting-VoiceFax-Data-Muxes WorldWideWebAnything-Intranets-NetAdmin-UnixAdmin-Security-ReallyHardMath
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:09 EDT