DTE rate latency measured
John Navas
From time to time there are posts noting that one of the advantages of
locking the serial port at 115,200 bps, rather than 57,600 bps, is reduced
latency. For those that might be interested, here is a measurement of
latency effect of the DTE rate.
My test was to measure the 'ping' response time from my Windows 95 machine
to the router in a local Internet POP. I get solid, nearly error-free
28,800 connections to this POP, but to be safe I locked the speed down to
26,400. The modem at my end was a USR Courier V.Everything. The POP uses
USR MP-series racks. From prior tests I know that the POP is locked at
115,200 bps.
I ran 90 pings at both 57,600 and 115,200 with Microsoft's ping applet,
which has millisecond timing resolution. I then discarded the 15 longest
times in order to minimize disturbances caused my the normal load on the
POP. Finally, I ran a statistical analysis of the results:
57,600 115,200
Mean 118 Mean 110.68
Standard Error 0.173983534 Standard Error 0.231064906
Median 118 Median 110
Mode 118 Mode 110
Standard Deviation 1.506741607 Standard Deviation 2.001080789
Sample Variance 2.27027027 Sample Variance 4.004324324
Kurtosis -0.703023561 Kurtosis 0.27110332
Skewness -0.12176195 Skewness 0.757041193
Range 6 Range 9
Minimum 115 Minimum 107
Maximum 121 Maximum 116
Sum 8850 Sum 8301
Count 75 Count 75
Confidence Level(95.0%) 0.346669801 Confidence Level(95.0%)0.460406931
The bottom line is that 115,200 bps reduced the round-trip ping time by
roughly 8 ms as compared to 57,600 bps. The difference is small (too small
to be noticeable in an interactive mode), but it is there.
--
Best regards,
John mailto:JNa...@NavasGrp.com http://web.aimnet.com/~jnavas/
28800 Modem FAQ: http://web.aimnet.com/~jnavas/modem/faq.html
John Navas
[Posted to comp.dcom.modems]
lu...@srv.net (Rob Luke) wrote:
>In article <318e7847...@news.aimnet.com>, JNa...@NavasGrp.com says...
>
>-From time to time there are posts noting that one of the advantages of
>-locking the serial port at 115,200 bps, rather than 57,600 bps, is reduced
>-latency. For those that might be interested, here is a measurement of
>-latency effect of the DTE rate.
>
>..................[SNIP test results].................
>
>-The bottom line is that 115,200 bps reduced the round-trip ping time by
>-roughly 8 ms as compared to 57,600 bps. The difference is small (too small
>-to be noticeable in an interactive mode), but it is there.
>
>Hey, would someone with 230kbps DTE hardware (Bob, AMQUEST ?) do a similar
>comparison locking at 115k and 230k? Just call me curious.
Assuming you could find an indial locked at 230.4 Kbps, and that the modems
at both ends were fully up to the job, you would expect to see a further
decrease in latency of 4 ms in moving up to 230.4 Kbps from 115.2 Kbps.
Geoffrey Welsh
In article <4mrn8v$4...@mars.hyperk.com>, lu...@srv.net (Rob Luke) wrote:
>In article <318e7847...@news.aimnet.com>, JNa...@NavasGrp.com says...
>-roughly 8 ms as compared to 57,600 bps. The difference is small (too small
>-to be noticeable in an interactive mode), but it is there.
>
>Hey, would someone with 230kbps DTE hardware (Bob, AMQUEST ?) do a similar
>comparison locking at 115k and 230k? Just call me curious.
It'll be approximately half, or 4 ms.
--
Geoffrey Welsh, Developer, InSystems Technologies Inc.
Temporary: crs...@inforamp.net; At work: insy...@pathcom.com
At home: ge...@zswamp.uucp or [xenitec.on.ca|m2xenix.psg.com]!zswamp!geoff
TYPING IN ALL CAPS IS GROUNDS FOR IMMEDIATE DISMISSAL.