1、 TECHNICAL REPORT ATIS-0100021 ANALYSIS OF FCC-REPORTABLE SERVICE OUTAGE DATA ATIS is the leading technical planning and standards development organization committed to the rapid development of global, market-driven standards for the information, entertainment and communications industry. More than
2、250 companies actively formulate standards in ATIS 18 Committees, covering issues including: IPTV, Service Oriented Networks, Energy Efficiency, IP-Based and Wireless Technologies, Quality of Service, and Billing and Operational Support. In addition, numerous Incubators, Focus and Exploratory Groups
3、 address emerging industry priorities including “Green”, IP Downloadable Security, Next Generation Carrier Interconnect, IPv6 and Convergence. ATIS is the North American Organizational Partner for the 3rd Generation Partnership Project (3GPP), a member and major U.S. contributor to the International
4、 Telecommunication Union (ITU) Radio and Telecommunications Sectors, and a member of the Inter-American Telecommunication Commission (CITEL). For more information, please visit . Notice of Disclaimer this reflects the difference between the reporting criteria which has an absolute 30-minute duration
5、 threshold, while no such reporting threshold exists for customers affected alone. However, the magnitude weight curve has a more sustained rise and does not approach its asymptote until well beyond its inflection point. This reflects the outage index sensitivity to the impact of affecting larger gr
6、oups of customers relative to duration impacts. The number of customers affected is determined in the following way: Step 1: Calculate the number of users affected by taking the maximum of the Number of Potentially Affected fields under Effects of the Outage - i.e., Wireline Users, Wireless (non-pag
7、ing) Users, Paging Users, Cable Telephony Users, and Satellite Users. Step 2: Calculate the number of equivalent customers affected based on the reported number of Blocked Calls. If the Historic field is checked for Blocked Calls, then this number is equal to the reported number of Blocked Calls. If
8、 the Real-Time field is checked for Blocked Calls, then this number is equal to the reported number of Blocked Calls divided by 3. Step 3: Calculate the number of equivalent customers affected based on the reported number of DS3s affected. This number is equal to the reported number of DS3s affected
9、 multiplied by 666.667. This conversion is based on the equivalency of 672 DS0 circuits to a DS3 as discussed in FCC 04-188. Step 4: Calculate the number of equivalent customers affected based on the reported number of Lost SS7 MTP Messages. If the Historic field is checked for Lost SS7 MTP Messages
10、, then this number is equal to the reported number of Lost SS7 MTP Messages multiplied by 30,000 and ATIS-0100021 8 divided by 167,000. This conversion is based on the Alcatel equivalency of 500,000 lost MTP messages with 90,000 blocked calls as discussed in FCC 04-188. If the Real-Time field is che
11、cked for Lost SS7 MTP Messages, then this number is equal to the reported number of Lost SS7 MTP Messages multiplied by 10,000 and divided by 167,000. Step 5: The number of customers affected for determining the magnitude weight is the maximum of the four values calculated in Steps 1 through 4. 6.3
12、Outage Index Examples 6.3.1 Outage Index Values Given Outage Duration and Outage Magnitude Metric Each column of Table 1 provides outage indexes for a given duration and a variety of magnitudes. Each row of the table provides outage indexes for a given magnitude and a variety of durations. The magni
13、tude can be expressed in different ways (e.g., blocked calls, customers, DS3s) as well as historic or real-time measures. Each row has an equivalent magnitude for the purposes of the outage index calculation; for example, 10 million customers affected is equivalent to 15,000 DS3s affected or 30 mill
14、ion real-time blocked calls. ATIS-0100021 9 Table 1: Outage Index Given Magnitude and Duration A. E911 Not Affected DS3s Customers MinutesHistoricReal-TimeHistoricReal-TimeAfectdAfecd 3013612173010 30 57 1,670 10 0.00.000.000.000.000.000.000.030 90 1,670 5,01 30 0.00.000.000.000.000.000.000.050 1,50
15、 2,783 8,350 50 0.00.000.000.000.000.000.000.01,000 3,000 5,567 16,700 1,000 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.002,000 6,000 11,133 33,400 3 2,000 0.00 0.00 0.01 0.01 0.01 0.01 0.01 0.014,000 12,000 22,267 66,800 6 4,000 0.01 0.02 0.02 0.03 0.03 0.03 0.03 0.0310,000 30,000 55,667 167,000 15 10,00
16、0 0.08 0.11 0.15 0.17 0.19 0.19 0.20 0.2030,000 90,000 167,000 501,000 45 30,000 0.72 0.95 1.36 1.56 1.69 1.75 1.80 1.8050,000 150,000 278,333 835,000 75 50,000 2.00 2.63 3.79 4.35 4.69 4.86 4.99 5.00100,000 300,000 556,667 1,670,000 150 100,000 6.57 8.65 12.43 14.27 15.40 15.97 16.39 16.42300,000 9
17、00,000 1,670,000 5,010,000 450 300,000 18.74 24.67 35.48 40.72 43.93 45.57 46.75 46.84500,000 1,500,000 2,783,333 8,350,000 750 500,000 25.75 33.89 48.73 55.94 60.34 62.59 64.22 64.341,000,000 3,000,000 5,566,667 16,700,000 1,500 1,000,000 34.66 45.63 65.61 75.32 81.24 84.27 86.46 86.633,000,000 9,0
18、00,000 16,700,000 50,100,000 4,500 3,000,000 44.08 58.02 83.43 95.77 103.30 107.16 109.94 110.165,000,000 15,000,000 27,833,333 83,500,000 7,500 5,000,000 46.43 61.11 87.88 100.87 108.81 112.87 115.80 116.0310,000,000 30,000,000 55,666,667 167,000,000 15,000 10,000,000 48.26 63.53 91.36 104.86 113.1
19、1 117.33 120.39 120.62Magnitude DurationBlocked Calls Lost SS7 MTP Messages Hours DaysB. E911 Affected DS3s Customers MinutesHistoric Real-Time Historic Real-Time Affected Affected 30 1 3 6 12 1 7 30100 300 557 1,670 100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00300 900 1,670 5,010 300 0.00 0.00 0.00 0
20、.00 0.00 0.00 0.00 0.00500 1,500 2,783 8,350 500 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.001,000 3,000 5,567 16,700 1,000 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.002,000 6,000 11,133 33,400 3 2,000 0.01 0.01 0.01 0.01 0.01 0.02 0.02 0.024,000 12,000 22,267 66,800 6 4,000 0.03 0.03 0.05 0.06 0.06 0.06 0.06
21、 0.0610,000 30,000 55,667 167,000 15 10,000 0.16 0.21 0.30 0.35 0.37 0.39 0.40 0.4030,000 90,000 167,000 501,000 45 30,000 1.44 1.90 2.73 3.13 3.37 3.50 3.59 3.6050,000 150,000 278,333 835,000 75 50,000 4.00 5.27 7.57 8.69 9.37 9.72 9.98 10.00100,000 300,000 556,667 1,670,000 150 100,000 13.14 17.29
22、 24.87 28.55 30.79 31.94 32.77 32.83300,000 900,000 1,670,000 5,010,000 450 300,000 37.49 49.35 70.96 81.45 87.86 91.13 93.51 93.69500,000 1,500,000 2,783,333 8,350,000 750 500,000 51.49 67.78 97.47 111.88 120.68 125.18 128.44 128.691,000,000 3,000,000 5,566,667 16,700,000 1,500 1,000,000 69.33 91.2
23、6 131.23 150.63 162.48 168.54 172.93 173.263,000,000 9,000,000 16,700,000 50,100,000 4,500 3,000,000 88.16 116.04 166.87 191.54 206.60 214.31 219.89 220.325,000,000 15,000,000 27,833,333 83,500,000 7,500 5,000,000 92.86 122.23 175.76 201.75 217.62 225.74 231.61 232.0610,000,000 30,000,000 55,666,667
24、 167,000,000 15,000 10,000,000 96.53 127.06 182.72 209.73 226.23 234.67 240.77 241.24Blocked Calls Lost SS7 MTP Messages Hours DaysMagnitude DurationATIS-0100021 10 6.3.2 Outage Durations and Outage Magnitude Metrics that Result in a Given Outage Index Value Figures 4 and 5 present contour plots of
25、outage index versus magnitude (customers affected) and duration (hours). Each line of the plot represents an outage index value; for example, the line marked 10 in the right margin of the plot gives all magnitude and duration combinations that result in an outage index of 10. Figure 4 presents resul
26、ts when E911 is not affected, while Figure 5 presents results when E911 is affected. 0 0.5 1 1.5 2 2.5 305010015020025015101520250.50.10.01Duration (Hours)ThousandsofCustomersAffectedThousandsofCustomersAffectedFigure 4: Outage Index Contours Given Magnitude and Duration (E911 Not Affected) ATIS-010
27、0021 11 15101520250.50.10.01Duration (Hours)ThousandsofCustomersAffected0 0.5 1 1.5 2 2.5 305010015020025050ThousandsofCustomersAffectedFigure 5: Outage Index Contours Given Magnitude and Duration (E911 Affected) 6.4 Definition of Aggregated Outage Index The aggregated outage index represents the im
28、pact of a set of FCC-reportable service outages. The simplest function defining such an aggregated outage index is the sum of the outage indexes for the individual service outages in the set. 7 CONTROL CHARTS This section addresses techniques for the creation of control charts for outage frequency a
29、nd aggregated outage index. It emphasizes the application of appropriate statistical tests rather than the derivation of these tests. In contrast to trend analysis, control charts of outage frequency and aggregated outage index are used to gauge the state of network reliability at a particular momen
30、t in time rather than its change over a period of time. Control charts are constructed based on past data which represents the norm of the process. The levels in the control chart identify regions in which new data depart from the norm, positively (number of outages is much less than the norm) or ne
31、gatively (number of outages is much greater than the norm). It is important to establish a useful level of ATIS-0100021 12 significance for the control charts that balances errors of commission (false identification of departures from the norm) against errors of omission (not identifying departures
32、from the norm). 7.1 Control Charts for Outage Frequency Control charts for outage frequency are based on the assumption that the number of outages in a time period follows a Poisson distribution. The mean parameter for the distribution is based on the average number of outages experienced in past ti
33、me periods of similar length. The time periods used for defining should be representative of the process. While data may be available for long periods of time, much of it may be out-of-date and thereby unrepresentative of the current process; data from those periods too far in the past to be conside
34、red representative should not be included. The model can be extended to consider changes in the network size; however, this is beyond the scope of this document at this time. Example: The table below gives the number of outage reports in the last 20 quarters for a network. In the first four quarters
35、, the network implementation was at the tail end of a major technological change and growth followed by relative stability in the subsequent 16 quarters. For this reason, the control chart mean was based on the average number of outage reports in years 2 through 5: = (60 + 51 + 51 + + 55 + 49) / 16
36、= 51.0. Year 1 Year 2 Year 3 Year 4 Year 5 Q1 41 60 42 57 43 Q2 36 51 49 54 57 Q3 35 51 65 46 55 Q4 51 49 41 47 49 Control limits can be constructed using the Poisson distribution and the control chart mean . The Upper Control Limit (UCL) is a value such that: Pr(X UCL) and Pr(X UCL-1) where X is a
37、random variable from a Poisson distribution with mean and is a pre-defined one-sided level of significance. The UCL is set so that the probability of exceeding it is small if the process has not changed. Since the Poisson distribution has integer values only, the UCL can be any one of an infinite nu
38、mber of values between two consecutive integers; typically, the UCL is set as the halfway point between two consecutive integers. Similarly, the Lower Control Limit (LCL) is a value such that: Pr(X where X is a random variable from a Poisson distribution with mean and is a pre-defined one-sided leve
39、l of significance. (In general, a different level of significance is permissible but generally symmetry is preserved between the UCL and LCL by using the same level of significance.) For control charts on outages and their impacts, the focus is generally on the UCL in order to identify if current re
40、liability is worse than expected from past experience. However, the LCL can be useful as well in identifying improvements in reliability or changes in the process. ATIS-0100021 13 Example: The table below shows probabilities from a Poisson distribution with mean 51.0, the control chart mean from the
41、 earlier example. x Pr(X x) 38 0.024997 64 0.033070 39 0.035471 65 0.024623 Assuming that = 0.025 has been decided as the one-sided level of significance for both the UCL and the LCL: LCL = 37.5 since Pr(X UCL = 65.5 since Pr(X UCL) = Pr(X 65) = 0.024623 and Pr(X UCL-1) = Pr(X 64) = 0.033070 . Note
42、that 37.3 and 65.6 would also work as the LCL and the UCL in this example (as well as an infinite quantity of values between 37 and 38 and between 65 and 66). The midpoints of these intervals have been chosen primarily for purposes of plotting on charts. The control chart for this example is shown b
43、elow. Note that the use of the midpoint of the interval facilitates the visual identification of a data point being above or below a control limit. The data point for Year 6 Quarter 1 is left blank for completion once the data is known. This point can then be compared with the control limits. A poin
44、t above the UCL would be cause for concern that reliability in Year 6 Quarter 1 was significantly worse than in the past. A point below the LCL would be an indication that reliability in Year 6 Quarter 1 was significantly better than in the past. A data point between the two control limits indicates
45、 that reliability in Year 6 Quarter 1 was not significantly different from the past. ATIS-0100021 14 303540455055606570Y2Q2Y2Q3Y2Q4Y3Q1Y3Q2Y3Q3Y3Q4Y4Q1Y4Q2Y4Q3Y4Q4Y5Q1Y5Q2Y5Q3Y5Q4Y6Q1QuarterNumberofOutageReportsDataUCLMeanLCLFor some data sets with low numbers of outage reports, an LCL may not be ca
46、lculable. For example, if mean = 3.0, then Pr(X = 0) = 0.0498. An LCL does not exist for = 0.025 since it is impossible to find a region of low values with probability of occurrence less than 0.025. However, for = 0.05, LCL = 0.5 since the Pr(X UCL) = where X is a random variable from a Gamma distri
47、bution with shape and scale parameters as given above and is a pre-defined one-sided level of significance. The UCL is set so that the probability of exceeding it is small if the process has not changed. Similarly, the Lower Control Limit (LCL) is a value such that: Pr(X UCL) = Pr(X 366.3) = 0.025.
48、ATIS-0100021 17 8 TREND ANALYSIS This section addresses techniques for the identification and testing of trends in outage frequency and aggregated outage index over time. 8.1 Trend Analysis of Outage Frequency This section provides a methodology for trend analysis of the number of outages in a time
49、period from FCC-reportable service outage data. It emphasizes the application of appropriate statistical tests rather than the derivation of these tests. The primary purpose of collecting outage data is to better understand the state of network reliability. Since outages are not deterministic, this requires a statistical model which uses the data to describe the underlying process. The data can be tracked across time, thus providing insight to trends in the performance of the network. A first step in analyzing the data is to
