📄 rfc956.txt
字号:
Offset Count Offset Count ------------- ------------- 0 sec 35 (continued) 30 3 -30 50 60 8 -60 42 90 3 -90 8 120 1 -120 4 150 1 -150 2 180 0 -180 1 210 0 -210 0 240 0 -240 1 270 0 -270 0 > 270 2 < -270 2 Table A3. Offset Distribution < 270 sec A total of 138 of the 163 host clocks are within a minute in accuracy, while a total of four host clocks are off more than 4.5 minutes. It is considered likely that most host clocks, with the exception of the 16 identified above as probably synchronized to a radio clock, are set manually by an operator. Inspection of the raw data shows some hosts to be very far off; for instance, SRI-UNICORN.ARPA is off more than ten hours. Note the interesting skew in the data, which show that most host clocks are set slow relative to standard time. A2. ICMP Timestamp Messages Experiment The the second experiment four ICMP Timestamp messages were sent at about three-second intervals to each of the 1775 hosts and 110 gateways listed in the NIC Internet host table. A total of 1910 samples were received from 504 hosts and gateways and compared with a local reference based on a WWVB radio clock, which is known to be accurate to within a few milliseconds. Support for the ICMP Timestamp messages is optional in the DoD Internet protocol suite, so it is not surprising that most hosts and gateways do not support it. Moreover, bugs are known to exist in several widely distributed implementations of this feature. The situation proved an interesting and useful robustness test for the clustering algorithm described in the main body of this note. While the complete table of ICMP offsets by host is too large to reproduce here, the following Tables A4 through A7 show the interesting characteristics of the distribution. The raw statistics computed from the weighted data indicate a mean of -2.8E+6 msec, together with a maximum of 8.6E+7 msec and a minimum of -8.6E+7 msec. Setting a local clock on the basis of theseMills [Page 21] RFC 956 September 1985Algorithms for Synchronizing Network Clocks statistics alone would be ridiculous; however, as described in the main body of this note, use of the clustering algorithm improves the estimate to within 8 msec of the correct value. The apparent improvement of about six orders in magnitude is so remarkable as to require a closer look at the distributions. The reasons for the remarkable success of the clustering algorithm are apparent from closer examination of the sequence of histograms shown in Tables A4 through A7. Table A4 shows the distribution in the scale of hours, from which it is evident that 80 percent of the samples lie in a one-hour band either side of zero offset; but, strangely enough, there is a significant dispersion in samples outside of this band, especially in the negative region. It is almost certain that most or all of the latter samples represent defective ICMP Timestamp implementations. Note that invalid timestamps and those with the high-order bit set (indicating unknown or nonstandard time) have already been excluded from these data. Offset Count Offset Count ------------- ------------- 0 hr 204 (continued) 1 10 -1 194 2 0 -2 0 3 0 -3 2 4 0 -4 17 5 0 -5 10 6 0 -6 1 7 0 -7 22 8 0 -8 20 9 0 -9 0 > 9 0 < -9 13 Table A4. ICMP Offset Distribution < 9 hours Table A5 shows the distribution compressed to the range of 4.5 minutes. About half of the 370 samples remaining after the outliers beyond 4.5 minutes are excluded lie in the band 30 seconds either side of zero offset, with a gradual tapering off to the limits of the table. This type of distribution would be expected in the case of host clocks set manually by an operator.Mills [Page 22] RFC 956 September 1985Algorithms for Synchronizing Network Clocks Offset Count Offset Count ------------- ------------- 0 sec 111 (continued) 30 25 -30 80 60 26 -60 28 90 13 -90 18 120 7 -120 19 150 5 -150 9 180 3 -180 10 210 3 -210 6 240 1 -240 2 270 2 -270 2 > 270 29 < -270 105 Table A5. ICMP Offset Distribution < 270 sec Table A6 shows the distribution compressed to the range of 27 seconds. About 29 percent of the 188 samples remaining after the outliers beyond 27 seconds are excluded lie in the band 3 seconds either side of zero offset, with a gradual but less pronounced tapering off to the limits of the table. This type of distribution is consistent with a transition region in which some clocks are set manually and some by some kind of protocol interaction with a reference clock. A fair number of the clocks showing offsets in the 3-27 second range have probably been set using the UDP Time protocol at some time in the past, but have wandered away as the result of local-oscillator drifts. Offset Count Offset Count ------------- ------------- 0 sec 32 (continued) 3 15 -3 22 6 9 -6 12 9 6 -9 8 12 13 -12 8 15 5 -15 5 18 8 -18 9 21 8 -21 7 24 9 -24 3 27 6 -27 3 > 27 114 < -27 202 Table A6. ICMP Offset Distribution < 27 sec Finally, Table A7 shows the distribution compressed to the range of 0.9 second. Only 30 of the original 504 samples have survived and only 12 of these are within a band 0.1 seconds either side ofMills [Page 23] RFC 956 September 1985Algorithms for Synchronizing Network Clocks zero offset. The latter include those clocks continuously synchronized to a radio clock, such as the DCNet clocks, some FORDnet and UMDnet clocks and certain others. Offset Count Offset Count ------------- ------------- 0 sec 6 (continued) .1 3 -.1 6 .2 1 -.2 3 .3 1 -.3 0 .4 0 -.4 0 .5 1 -.5 2 .6 0 -.6 0 .7 1 -.7 0 .8 4 -.8 2 .9 0 -.9 0 > .9 208 < -.9 266 Table A7. ICMP Offset Distribution < .9 sec The most important observation that can be made about the above histograms is the pronounced central tendency in all of them, in spite of the scale varying over six orders of magnitude. Thus, a clustering algorithm which operates to discard outliers from the mean will reliably converge on a maximum-likelihood estimate close to the actual value. A3. Comparison of UDP and ICMP Time The third experiment was designed to assess the accuracies produced by the various host implementations of the UDP Time protocol and ICMP Timestamp messages. For each of the hosts responding to the UDP Time protocol in the first experiment a separate test was conducted using both UDP and ICMP in the same test, so as to minimize the effect of clock drift. Of the 162 hosts responding to UDP requests, 45 also responded to ICMP requests with apparently correct time, but the remainder either responded with unknown or nonstandard ICMP time (29) or failed to respond to ICMP requests at all (88). Table A8 shows both the UDP time (seconds) and ICMP time (milliseconds) returned by each of the 45 hosts responding to both UDP and ICMP requests. The data are ordered first by indicated UDP offset and then by indicated ICMP offset. The seven hosts at the top of the table are continuously synchronized, directly or indirectly to a radio clock, as described earlier under the firstMills [Page 24] RFC 956 September 1985Algorithms for Synchronizing Network Clocks experiment. It is probable, but not confirmed, that those hosts below showing discrepancies of a second or less are synchronized on occasion to one of these hosts. Host UDP time ICMP time ------------------------------------------------- DCN6.ARPA 0 sec 0 msec DCN7.ARPA 0 0 DCN1.ARPA 0 -6 DCN5.ARPA 0 -7 UMD1.ARPA 0 8 UMICH1.ARPA 0 -21 FORD1.ARPA 0 31 TESLA.EE.CORNELL.EDU 0 132 SEISMO.CSS.GOV 0 174 UT-SALLY.ARPA -1 -240 CU-ARPA.CS.CORNELL.EDU -1 -514 UCI-ICSE.ARPA -1 -1896 UCI-ICSC.ARPA 1 2000 DCN9.ARPA -7 -6610 TRANTOR.ARPA 10 10232 COLUMBIA.ARPA 11 12402 GVAX.CS.CORNELL.EDU -12 -11988 UCI-CIP5.ARPA -15 -17450 RADC-MULTICS.ARPA -16 -16600 SU-WHITNEY.ARPA 17 17480 UCI-ICSD.ARPA -20 -20045 SU-COYOTE.ARPA 21 21642 MIT-MULTICS.ARPA 27 28265 BBNA.ARPA -34 -34199 UCI-ICSA.ARPA -37 -36804 ROCHESTER.ARPA -42 -41542 SU-AIMVAX.ARPA -50 -49575 UCI-CIP4.ARPA -57 -57060 SU-SAFE.ARPA -59 -59212 SU-PSYCH.ARPA -59 -58421 UDEL-MICRO.ARPA 62 63214 UIUCDCSB.ARPA 63 63865 BELLCORE-CS-GW.ARPA 71 71402 USGS2-MULTICS.ARPA 76 77018 BBNG.ARPA 81 81439 UDEL-DEWEY.ARPA 89 89283 UCI-CIP3.ARPA -102 -102148 UIUC.ARPA 105 105843 UCI-CIP2.ARPA -185 -185250 UCI-CIP.ARPA -576 -576386 OSLO-VAX.ARPA 3738 3739395Mills [Page 25] RFC 956 September 1985Algorithms for Synchro⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -