Copyright 1996-2017 The
Navas GroupSM, All Rights
Reserved.
Permission is granted to copy for private non-commercial
use only.
Posted as <http://modemfaq.navasgroup.com/latency.htm>.
One of the advantages of locking a 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 the latency effect of the DTE rate.
The difference in latency was measured by comparing the 'ping' response times from a Windows 95 machine to a router in a local Internet POP (Point Of Presence). Solid, nearly error-free 28,800 connections were possible, but to be safe the modem speed was locked down to 26,400 bps. The modem at the Windows 95 end was a USRobotics Courier V.Everything; the modems at the POP end were USRobotics MP-series racks. Prior tests established that the POP was locked at 115,200 bps.
90 pings were performed at both 57,600 and 115,200 bps using Microsoft's ping applet, which has millisecond timing resolution. The 15 longest times were discarded in order to minimize disturbances caused by momentary loads on the POP. Finally, the results were subjected to statistical analysis:
| 57,600 | 115,200 | |
|---|---|---|
| Mean | 118 | 110.68 |
| Standard Error | 0.173984 | 0.231065 |
| Median | 118 | 110 |
| Mode | 118 | 110 |
| Standard Deviation | 1.506742 | 2.001081 |
| Sample Variance | 2.27027 | 4.004324 |
| Range | 6 | 9 |
| Minimum | 115 | 107 |
| Maximum | 121 | 116 |
| Count | 75 | 75 |
| Confidence Level (95.0%) | 0.34667 | 0.460407 |
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.
Email comments and suggestions to John Navas
|
|
|