Digital subscriber line profile selection method and apparatus

Abstract

An Initial Profile Application Apparatus (IPAA) is operable to apply an initial profile to a modem pair connection system, the modem pair connection system comprising a first modem, a corresponding second modem and a metallic wire connection, wherein the first and second modems are operable to establish a data connection between themselves over the metallic wire connection. The IPAA comprises: a receiver; an evaluator; a line database; a comparator; a determiner; and an applicator.

Claims

1. A method of applying an initial profile to a modem pair connection system, the modem pair connection system comprising a first modem, a corresponding second modem and a metallic wire connection, wherein the first and second modems are operable to establish a data connection between themselves over the metallic wire connection and wherein the metallic wire connection is a new connection to be used for the first time in its current form for supporting the data connection between the first and second modems, the method comprising: receiving new-order information associated with a new order placed by a customer for a new broadband data connection to be supplied over the modem pair connection system; evaluating a set of risk factors for the new order based on the received new-order information; comparing the set of evaluated risk factors for the new order with corresponding risk factors for established modem pair connection systems, wherein none of the modem pairs in the established modem pair connection systems comprise both the first modem and the second modem, using a database storing information about a plurality of established modem pair connection systems which information includes values for the corresponding risk factors of each of the plurality of established modem pair connection systems, or information permitting the evaluation of such values, and information about a profile applied to each of the plurality of established modem pair connection systems; determining an initial profile to apply to the modem pair connection system based on the comparison of risk factors; and applying the determined initial profile to the modem pair connection system.

2. Processor implementable instructions stored on a non-transitory processor-readable storage medium for causing a processing device to carry out the method of claim 1 during execution of the instructions by the processing device.

3. A non-transitory computer-readable carrier medium carrying processor implementable instructions to, when loaded into and executed by a processor, cause the processor to perform the method as claimed in claim 1.

4. An apparatus for applying an initial profile to a modem pair connection system, the modem pair connection system comprising a first modem, a corresponding second modem and a metallic wire connection, wherein the first and second modems are operable to establish a data connection between themselves over the metallic wire connection and wherein the metallic wire connection is a new connection to be used for the first time in its current form for supporting the data connection between the first and second modems, the apparatus comprising: a receiver for receiving new-order information associated with a new order placed by a customer for a new broadband data connection to be supplied over the modem pair connection system; an evaluator for evaluating a set of risk factors for the new order based on the received new-order information; a line database storing information about a plurality of established modem pair connection systems which information includes values for the corresponding risk factors of each of the plurality of established modem pair connection systems, or information permitting the evaluation of such values, and information about a profile applied to each of the plurality of established modem pair connection systems; a comparator for comparing the set of evaluated risk factors for the new order with corresponding risk factors for established modem pair connection systems, wherein none of the modem pairs in the established modem pair connection systems comprise both the first modem and the second modem, using information from the line database; a determiner for determining an initial profile to apply to the modem pair connection system based on the comparison of risk factors; and an applicator for applying the determined initial profile to the modem pair connection system.

5. The apparatus of claim 4 further comprising a Dynamic Line Management system for monitoring the operation of the modem pair connection system after having applied an initial profile to the system and from time to time applying a different profile to the system in dependence upon the results of monitoring the system, in order to drive the operation of the system towards a target level of operation based on observed operational parameters of the system.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) In order that the present disclosure may be better understood, embodiments thereof will now be described, by way of example only, with reference to the accompanying drawings in which:

(2) FIG. 1 is a schematic illustration of a part of an access network illustrating typical cases in which different DSL technologies may be used, and within which aspects of the present disclosure may be employed.

(3) FIG. 2 is a schematic block diagram of an initial profile application apparatus together with a modem pair connection system.

(4) FIG. 3 is a schematic block diagram illustrating the initial profile application apparatus of FIG. 2 in greater detail.

(5) FIG. 4 is a flow chart illustrating the tasks performed by the apparatus of FIG. 3.

(6) FIG. 5 is a flow chart illustrating the tasks performed by the Compare Risk Factors subroutine of FIG. 4 in accordance with a risk factor by risk factor comparison.

(7) FIG. 6 is a flow chart illustrating the tasks performed by the Compare Risk Factors subroutine of FIG. 4 in accordance with a whole set of risk factors comparison.

DETAILED DESCRIPTION OF EMBODIMENTS

(8) FIG. 1 illustrates a portion of an access network. As shown, the illustrated portion includes several customer premises 10a-10e, a Distribution Point 20 with an associated DSLAM 22, a cabinet 30 with an associated cabinet DSLAM 32, some twisted metallic pairs 15a-15e, 25a-25e, 35a-35e and some optic fibers 35abc, 35de. The Figure illustrates that as an access network evolves it places optic fiber closer and closer towards the customers' premises. Thus before any optic fiber was installed beyond a local exchange building, each customer 10a-10e would have a single metallic twisted pair extending all the way from the customer premises via the distribution point 20 and the cabinet 30 back to a local exchange building (not shown). When a fiber optic is installed at the cabinet 30 all of the lines are capable of being diverted to a DSLAM 32 located at the cabinet and the signals between the cabinet and the local exchange can instead be sent over a fiber optic cable 35abc. The sections of the metallic twisted pairs 35a-35e from the cabinet to the exchange are usually left in place and a customer only has their line switched to the cabinet DSLAM 32 when they wish to upgrade to a new service (e.g. a VDSL2 service which can typically provide much higher speeds than the ADSL service originally provided from the local exchange over the entire twisted metallic pair). The old twisted metallic pair may still be used to carry POTS signals between the customer's telephone and the local exchange with only data signals being terminated at the DSLAM 32.

(9) A similar process occurs when a customer wishes to upgrade to an even newer faster service (e.g. a G.FAST service) via a DSLAM 22 located at a distribution point 20 with a corresponding fiber optic backhaul connection 35de from the distribution point 22 to the local exchange. It should be noted that in such a case, the new connection (e.g. a VDSL2 connection from cabinet 30 to customer A 10a, or a G.FAST connection from distribution point 20 to Customer E 10e) passes over a twisted metallic pair connection (25a and 15a or 15e alone respectively) which is fundamentally different to that which the previous service passed over (15a, 25a and 35a for Customer a and either 15e, 25e and 35e for an ADSL to G.FAST upgrade or 15e and 25e only for a VDSL2 to G.FAST upgrade for Customer E) even though parts of the twisted metallic connection are common in all cases (i.e. 15a-15e).

(10) FIG. 2 illustrates just one of the connections shown in FIG. 1 together with an additional Initial Profile Application apparatus 100 according to embodiments of the present disclosure. The initial profile application apparatus 100 which is shown in greater detail in FIG. 3, is operable to apply an initial profile to a modem pair connection system 5 (also hereinafter referred to as line 5) which is made up of a modem 33 within the DSLAM 32, a modem 11 within the CPE 10a and the twisted metallic pair connection 25a, 15a extending between the modems 33, 11. The profile (or an identification thereof) applied by the initial profile application apparatus is sent to the DSLAM which then ensures that when the line first synchronizes after having been suitably instructed by the initial profile application apparatus, it does so in accordance with the applied initial profile in a conventional manner.

(11) FIG. 3 illustrates the Initial Profile Application Apparatus (IPAA) 100 of FIG. 2 in greater detail. As shown it comprises an Interface/Receiver/Applicator 110 through which new order information is received as well as through which an initial profile is applied to the line 5 associated with the new order. The IPAA 100 also includes a processor 120 a memory 130 and a Line Database 140. The memory 130 stores various code means each of which, when executed by the processor 120 causes a corresponding module to be implemented. The various code means stored within the memory 130 include Evaluator code means 132, comparator code means 134 and determiner code means 136 which give rise to an evaluator module, a comparator module and a determiner module respectively. The evaluator module evaluates risk factor values for the new order when new order information is received. In the present embodiments, the new order information includes: an indication of the product which has been ordered (from which it is possible to ascertain whether the end customer will self-install his/her CPE modem or whether an engineer for the CP company will perform the installation of the CPE modem) the identity of the Communications Provider (CP) which has made the new order (from which it may be possible to estimate which CPE modem will be used as discussed below), the telephone number or other identifier of the actual line on which the service is to be provided from which the IPAA can ascertain (via external databases storing equipment inventories and line plant records) the loop length of the section of twisted metallic pair 15a,25a between the Cabinet 30 and the CPE 10a, the vendor of the DSLAM 32 and the postcode or other geographical region identifier of the cabinet 30. Note that in alternative embodiments, rather than the IPAA obtaining the ascertained information itself this could be done by an external device and the new order information might then contain only the information actually required by the IPAA to perform the following described operations.

(12) Having ascertained the following information about the new order: installation option of the new order, CP identity for the new order, DSLAM vendor of DSLAM 30, geographical region identifier (e.g. postcode) and loop length, the evaluator module assigns risk factor values as follows: set Product risk factor to SI for self install or to MI for engineer install; set CP to a CP ID code A, B, C, D, . . . according to CP identity, set DSLAM vendor to corresponding ID code for DSLAM vendors X, Y, Z, . . . according to the DSLAM vendor, set Loop Category to one of Short, Medium, Long or Very Long according to predefined ranges in accordance with identified loop length, and set Region to first half of postcode value.

(13) The comparator module is operable to compare this set of risk factors with corresponding risk factors of established lines contained within the line database 140 in a manner described in greater detail below with reference to FIGS. 4 and 5 (for a first embodiment) and with reference to FIGS. 4 and 6 (for a second embodiment). The output provided by the comparator module is a set of profile value attributes and these are then input to the determiner module which determines from the set of profile attribute values an initial profile to apply to the line 5. In the present embodiments, the profile attributes are rate cap (Max Speed), a specified minimum permitted speed (Min Speed), a target margin (Margin) and a level of impulse noise protection (INP). Each of a plurality of different profiles used in the present embodiments specify different values for these profile attributes. If the profile attributes input to the determiner module exactly match a single profile, then that profile is chosen. Otherwise, the determiner module is operable to select a profile which represents a best fit to the input profile attributes, in the present embodiment, by selecting the profile for which each of the attributes of the profile is as close as possible to the input attribute without being less conservative than the corresponding input parameter. In the present embodiments, Max Speed goes from less to more conservative as the value decreases, Min Speed goes from less to more conservative as the value decreases, Margin goes from less to more conservative as the value increases and INP goes form less to more conservative as the value increases. In alternative embodiments, the determiner module may alternatively generate a new profile based on the input profile attributes output from the comparator module.

(14) The Line database 140 stores records of established lines including a line index identifier, risk factor values, profile ID of the currently applied profile, information about the CPE modem device of the line and information about the time since a fault was last identified on the line and the time since the line was last reset, for each line in the database. Some example records are set out in the examples which follow below after a discussion of FIG. 4-6.

(15) FIG. 4 illustrates the tasks performed by the IPAA 100 in order to apply an initial profile to a line 5 associated with a new order. The method starts at s10 in which the new order information is received by the IPAA 100 via the interface/receiver/applicator 110. The method then proceeds to s20 in which the evaluator module evaluates the risk factors for the new order based on the information received at s10. The method then proceeds to the subroutine s30 in which the risk factors evaluated in s20 are compared with the risk factors of established lines using the information contained within the line database 140. As a result of the comparison, the comparator module outputs a set of profile attribute values. The method then proceeds to s40 in which the profile attribute values output by the comparator module at s30 are provided to the determiner module which determines an initial profile to apply to the line 5 associated with the new order. Finally, at s50 the determined profile is applied (via the interface/receiver/applicator 110) to the new line 5.

(16) Referring now to FIG. 5, a first comparison method employed by a comparator module according to a first embodiment is now described. The subroutine s40 of this first embodiment (option 1) begins at s510 in which a first risk factor (e.g. product installation methodtaking a value of SI or MI) is obtained and then passed to s520. At s520, an average set of profile attributes for that risk factor is then calculated. In the present embodiment, this is done by, for each profile attribute Max Speed, Min Speed, Margin and INP, ordering the attribute values of each line in the database having the same risk factor value and then selecting the mode attribute value within the identified lines sharing the same risk factor value. As an alternative, here instead of using the mode, the median attribute value for the selected lines could be chosen or a percentile value (e.g. the 60.sup.th percentile value where the lines are ranked for each profile attribute from least conservative to most conservative so that the profile attribute value of the 60.sup.th percentile profile is as conservative or more conservative than the profile attribute value of the median (50.sup.th percentile) profile. The method then proceeds to s530 in which the average attribute values determined in this way (for the risk factor associated with the current iteration) is stored. The method then proceeds to s540 in which it is determined whether there are more risk factors to be evaluated. If so the method proceeds to s550 in which the next risk factor to be evaluated is obtained and then the method reverts to s520.

(17) If at s540 it is determined that all of the risk factors have been evaluated, the method proceeds to s560 in which a super-averaged attribute set is calculated from the stored attribute sets associated with each risk factor. In the present embodiment, this is done by generating a linear combination of all of the attribute factors associated with the different attribute sets, weighted by a suitable weighting associated with the risk factor. An example of this is given in the below example. Alternative embodiments may employ a median or percentile combination approach and an example of this variant is also given in the below example.

EXAMPLE 1

(18) Input Example Data.

(19) TABLE-US-00001 Speed Time compared since to last Line DSLAM CPE expected fault Pro- index Product CP Vendor ID (fraction) Region (days) file 1 MI A A CPE1 1.1 A1 500 1 2 SI B A CPE2 0.7 A1 3 3 MI A A CPE1 0.9 A1 1 4 MI A B CPE2 1.0 A1 1 5 SI B B CPE2 0.4 A2 5 6 SI C B CPE1 0.8 A2 100 3 7 SI C A CPE2 0.65 A2 4 8 MI B B CPE1 1.3 A2 1 9 MI C A CPE1 0.7 A2 4
Example Calculations for Option 1
Product Risk
Ordered product is Self-install (SI) from CP A.
The most common profile for SI is:

(20) TABLE-US-00002 Speed Time compared since to last Line DSLAM CPE expected fault Pro- index Product CP Vendor ID (fraction) Region (days) file 1 MI A A CPE1 1.1 A1 500 1 2 SI B A CPE2 0.7 A1 3 3 MI A A CPE1 0.9 A1 1 4 MI A B CPE2 1.0 A1 1 5 SI B B CPE2 0.4 A2 5 6 SI C B CPE1 0.8 A2 100 3 7 SI C A CPE2 0.65 A2 4 8 MI B B CPE1 1.3 A2 1 9 MI C A CPE1 0.7 A2 4
Determine most common profile:
Profile 1=0
Profile 2=0
Profile 3=2
Profile 4=1
Profile 5=1
Therefore, the recommended profile is Profile 3.
Equipment Risk
Equipment is determined by the combination of the most common CPE for an ordering CP and the expected DSLAM vendor. Customer is due to be provided on vendor B.

(21) TABLE-US-00003 Speed Time compared since to last Line DSLAM CPE expected fault Pro- index Product CP Vendor ID (fraction) Region (days) file 1 MI A A CPE1 1.1 A1 500 1 2 SI B A CPE2 0.7 A1 3 3 MI A A CPE1 0.9 A1 1 4 MI A B CPE2 1.0 A1 1 5 SI B B CPE2 0.4 A2 5 6 SI C B CPE1 0.8 A2 100 3 7 SI C A CPE2 0.65 A2 4 8 MI B B CPE1 1.3 A2 1 9 MI C A CPE1 0.7 A2 4
Ordering CP Is CP A.

(22) TABLE-US-00004 Speed Time compared since to last Line DSLAM CPE expected fault Pro- index Product CP Vendor ID (fraction) Region (days) file 1 MI A A CPE1 1.1 A1 500 1 2 SI B A CPE2 0.7 A1 3 3 MI A A CPE1 0.9 A1 1 4 MI A B CPE2 1.0 A1 1 5 SI B B CPE2 0.4 A2 5 6 SI C B CPE1 0.8 A2 100 3 7 SI C A CPE2 0.65 A2 4 8 MI B B CPE1 1.3 A2 1 9 MI C A CPE1 0.7 A2 4
Most common CPE for CP A is CPE1.
This gives a final combination of DSLAM vendor B, CPE 1:

(23) TABLE-US-00005 Speed Time compared since to last Line DSLAM CPE expected fault Pro- index Product CP Vendor ID (fraction) Region (days) file 1 MI A A CPE1 1.1 A1 500 1 2 SI B A CPE2 0.7 A1 3 3 MI A A CPE1 0.9 A1 1 4 MI A B CPE2 1.0 A1 1 5 SI B B CPE2 0.4 A2 5 6 SI C B CPE1 0.8 A2 100 3 7 SI C A CPE2 0.65 A2 4 8 MI B B CPE1 1.3 A2 1 9 MI C A CPE1 0.7 A2 4
Determine most common profile:
Profile 1=1
Profile 2=0
Profile 3=1
Profile 4=0
Profile 5=0
Therefore, the recommended profile is 1 or 3. In the case of equal numbers, choose the most conservative, i.e. Profile 3.
CP Risk
Installation practice, modem and helpdesk instruction all contribute to CP performance. Ordering CP is CP A:

(24) TABLE-US-00006 Speed Time compared since to last Line DSLAM CPE expected fault Pro- index Product CP Vendor ID (fraction) Region (days) file 1 MI A A CPE1 1.1 A1 500 1 2 SI B A CPE2 0.7 A1 3 3 MI A A CPE1 0.9 A1 1 4 MI A B CPE2 1.0 A1 1 5 SI B B CPE2 0.4 A2 5 6 SI C B CPE1 0.8 A2 100 3 7 SI C A CPE2 0.65 A2 4 8 MI B B CPE1 1.3 A2 1 9 MI C A CPE1 0.7 A2 4
Determine most common profile:
Profile 1=3
Profile 2=0
Profile 3=0
Profile 4=0
Profile 5=0
Therefore, the recommended profile is Profile 1.
Geography Risk
Customer is located in A2.

(25) TABLE-US-00007 Speed Time compared since to last Line DSLAM CPE expected fault Pro- index Product CP Vendor ID (fraction) Region (days) file 1 MI A A CPE1 1.1 A1 500 1 2 SI B A CPE2 0.7 A1 3 3 MI A A CPE1 0.9 A1 1 4 MI A B CPE2 1.0 A1 1 5 SI B B CPE2 0.4 A2 5 6 SI C B CPE1 0.8 A2 100 3 7 SI C A CPE2 0.65 A2 4 8 MI B B CPE1 1.3 A2 1 9 MI C A CPE1 0.7 A2 4
Determine most common profile:
Profile 1=1
Profile 2=0
Profile 3=1
Profile 4=2
Profile 5=1
Therefore, the recommended profile is profile 4.
Calculation of Profile

(26) Different weighting factors are used for each parameter type. This is because the influence on performance can vary, i.e. equipment type drives the maximum speed parameter more than others due to memory restrictions, or higher quality components used in the build process. Installation method dominates on error performance due to use of unbalanced home wiring.

(27) This gives a table of weighting factors:

(28) TABLE-US-00008 Risk factor Max speed Min speed Margin INP Product risk 0.3 0.3 0.25 0.5 Equipment risk 0.4 0.5 0.25 0 CP risk 0.2 0 0.25 0.1 Geography risk 0.1 0.2 0.25 0.4
Profiles selected are:
Product risk=profile 3
Equipment risk=profile 3
CP risk=profile 1
Geography risk=profile 4
Profile attributes are:

(29) TABLE-US-00009 Profile Max speed Min speed Margin INP 1 100 2 6 0 2 100 2 9 1 3 75 2 6 1 4 50 2 6 2 5 25 2 6 8
Simplified calculation for three presented risk factors is:
(CP_risk_weightprofile attribute)+(Product_risk_weightprofile attribute)+(Equipment_risk_weightprofile attribute)+(geography_risk_weightprofile attribute)=determined profile attribute.
For max speed:
(0.2100)+(0.375)+(0.475)+(0.150)=77.5 Mbps.
For min speed:
(02)+(0.32)+(0.52)+(0.22)=2.
For Margin
(0.256)+(0.256)+(0.256)+(0.256)=6.
For INP
(0.10)+(0.51)+(01)+(0.42)=1.3
Finally, a look-up against the profile table is used. The closest fit is profile 3.

Variant of Example 1

(30) For a given line:

(31) TABLE-US-00010 Risk Factor Max Rate Weighting CP 40 0.1 DSLAM 60 0.45 CPE 30 0.55
Sort by attribute:

(32) TABLE-US-00011 Risk Factor Max Rate Weighting DSLAM 60 0.45 CP 40 0.1 CPE 30 0.55
Normalize weightings so they add to 1:

(33) TABLE-US-00012 Risk Factor Max Rate Weighting DSLAM 60 0.41 CP 40 0.09 CPE 30 0.50
The median would be 40. The 60.sup.th percentile (from less to more conservative) would be 30.

(34) With some discrete attributes it might be more appropriate (low processing complexity) to use mode, e.g. an option to retransmit errored data segments (ReTX) On or Off. Mode, however, is the simplest calculation so it is good for solutions requiring minimum processing resources to be expended.

(35) To recap, therefore, the method of operation according to the first embodiment operates thus:

(36) For each risk, the existing DSL lines are grouped by that factor and a ranked list of percentage of lines on each DSL profile is calculated.

(37) A consolidated database may be constructed using these ranked lists and the lines can be grouped via database queries where the attribution under consideration is matched to the line due to be provisioned. For example, for a given CP, modem combination: Select providing_CP, CPE_modem, count_profile, group by providing_CP, CPE_modem, where providing_CP=CPx, CPE_modem,=Modemxxxx sort by count_profile

(38) By running these queries for each attribute, you will get the most common profile on the existing line base for that grouping. These different risks then need to be combined to give a single profile selection.

(39) The choice of profile is linked to your profile design. If you use banded profiles (i.e. with an upper and lower rate bound), then you need to use the lowest recommended profile to prevent lines.

(40) If you have target margin control, or capped (i.e. a maximum rate only), you are best using a weighted average to combine the profiles. This will require the profiles to be resolved to their individual attributions such as margin, noise protection and line rates.

(41) TABLE-US-00013 Profile Max speed Min speed Margin INP 1 100 2 6 0 2 100 2 9 1 3 75 2 6 1 4 50 2 6 2 5 25 2 6 8
The risk factors would then pick their most common profile: CP risk=profile 3 Plant risk=profile 2 Product risk=profile 4 Geography risk=profile 1 Equipment risk=profile 1

(42) Each risk factor would have an assigned weighting factor. However, this cannot operate on the profile number, as the variation in each one is non-linear. Instead the scalar value for each column would need to be assessed and matched to the nearest profile, i.e.: CP_risk_weightprofile attribute+plant_risk_weightprofile attribute+Product_risk_weightprofile attribute+Geography_risk_weightprofile attribute+Equipment_risk_weightprofile attribute=determined profile attribute.

(43) This would be completed for each profile attributei.e. maximum speed, minimum speed, Margin and INP and for each directionupstream and downstream to product a set of determine profile attributes. This would then be matched to the profile table for the best fit.

(44) This profile would then be applied to the line as part of the provisioning process.

(45) Referring now to FIG. 6, the tasks performed by the comparator module according to a second embodiment commence at s610 with the getting of all of the risk factors evaluated for the new order. The method then proceeds to s620 where the target risk factor values or ranges of values are evaluated for a first iteration of the method. In the present example, this involves simply setting the target values to equal the values of the risk factors obtained at s610. As an example these might be product install type=MI (indicating an engineer performed installation) CP=A (indicating that the CP making the order is CP A) DSLAM vendor=X (indicating that the DSLAM 32 is made vendor X), CPE ID=CPE1 (indicating that the most likely CPE device identity is CPE1in the present example this is estimated based on identifying the mode CPE ID for all lines associated with CP A), Loop category=Short, Region=A1 (corresponding to a postcode region A1). In the present embodiment, a target risk factor is also specified for a time since last fault. This is something which does not necessarily depend upon the new order information but could simply be a value chosen by the network operator in accordance with a desired policy of the network operator. In the present example the value is always set to neverindicating that only established lines which have not experienced any faults should be included in the target set of established lines.

(46) Having thus specified the target values for the target set to be selected, a search is performed at s630 to identify established lines within the line database 140 which satisfy the specified target criteria values for the associated risk factors of the established lines. The number of established lines meeting these specified criteria is then compared with a threshold (which in the present example is set to 200) to determine if a sufficiently large target set has been found. If not, then the method proceeds to s650 in which the target risk factor values are expanded to form ranges (or increased ranges) rather than individual target values using a number of pre-specified expansion rules. These rules can take the form of various IF THEN statements based on heuristics known to network operators. For example experience may show that the type of installation has little impact on very long or long lines, thus an expansion rule could take the form of IF target loop length risk factor=Long or Very Long THEN set product installation target value/Range to ALL values (i.e. either MI or SI). The expansion rules are given in an order and as soon as a single expansion rule is found which causes the target risk factor value ranges to be expanded, the method returns to S630. If the situation arises where no further expansion rules are applicable at s650 and yet the number of lines in the target set is below the threshold amount, a default initial profile can be applied to the new order line.

(47) Once at s640 it is determined that the target set has enough members, the method proceeds to s660 in which an averaged attribute set is obtained by ranking the lines by order of their attribute values and then selecting a predetermined percentile value (or a mode value if no ranking of the attribute value is possible).

(48) This method is illustrated in the following second Example:

EXAMPLE 2

(49) Example of a Few Lines Out of 1000 of Raw Data

(50) TABLE-US-00014 Time Time since last since DLM Line DSLAM CPE Loop last fault reset index Product CP Vendor ID Category Region (days) (days) Profile 1 MI A X CPE1 Long A1 500 10 1 2 SI B X CPE2 Medium A1 400 3 3 MI A X CPE1 Short A1 60 1 4 MI A Y CPE2 Very A1 5 1 Long 5 SI B Y CPE2 Short A2 1001 5 6 SI C Y CPE1 Very A2 100 100 3 Long 7 SI C X CPE2 Medium A2 67 4 8 MI B Y CPE1 Short A2 97 1 9 MI C X CPE1 Very A2 20 4 Short . . .
Setup Process Repeated Every Week in Offline Process
Discard data from lines with time since reset <=20 days because they have not had time to stabilize, or time since user connect <=20 (not shown here).
Group the time since last fault into 0-10 days, 10-30 days, 30-100, >100 or never.

(51) TABLE-US-00015 Time Time since last since DLM Line DSLAM CPE Loop last fault reset index Product CP Vendor ID Category Region (days) (days) Profile 2 SI B X CPE2 Medium A1 Never 400 3 3 MI A X CPE1 Short A1 Never 60 1 5 SI B Y CPE2 Short A2 Never 1001 5 6 SI C Y CPE1 Very A2 30-100 100 3 Long 7 SI C X CPE2 Medium A2 Never 67 4 8 MI B Y CPE1 Short A2 Never 97 1 9 MI C X CPE1 Very A2 Never 20 4 Short . . .
For each possible combination of Product, CP, DSLAM Vendor, CPE ID, Loop Category, Region, Time since last fault (days) select the lines.
First Example Set of Risk Factors:

(52) TABLE-US-00016 Time since last DSLAM CPE Loop fault Product CP Vendor ID Category Region (days) MI A X CPE1 Short A1 Never
Imagine there are 200 lines in this set and we now have the profile index for each line.
200 lines are deemed to be enough lines to give a valid result.
The following shows the attributes for some profiles:

(53) TABLE-US-00017 Profile Max speed Margin INP 1 100 6 0 2 100 9 1 3 75 6 1 4 50 6 2 5 25 6 8 . . . 34 75 3 1 . . .
Given the profiles for each attribute compute the 60.sup.th (where 60.sup.th is close to fast and unstable than median) percentile of the attributes for the lines in the set.
For this set of lines the 60.sup.th percentiles are Max Speed=75, Margin=3 and INP=1, this equates to a profile index number of 34.
Second Example Set of Risk Factors to Analyze:

(54) TABLE-US-00018 Time since last DSLAM CPE Loop fault Product CP Vendor ID Category Region (days) MI A X CPE1 Very A1 Never Long
Imagine there are 20 lines in this set and we now have the profile index for each line.
We know that when lines are very long the product has no impact on the profile chosen. So for this risk factor we now select lines to analyze from:

(55) TABLE-US-00019 Time since last DSLAM CPE Loop fault Product CP Vendor ID Category Region (days) All A X CPE1 Very A1 Never Long
This gives a further 20 lines to analyze, but there still are not enough lines to be statistically valid.

(56) To get enough lines to analyze would use the algorithm described below to expand the criteria.

(57) Algorithm Each of the discrete factors are indexed, for example CP. CPa=1, CPb=2, etc. Each of the continuous factors are reduced to a number of buckets, which may be of none uniform width and not necessarily contain a fixed number of customers. For example, line loss, the indexed bins could be from 0 to 10 dB (@300 kHz) in 0.5 dB steps and 10 to 20 dB in 1 dB steps. Let us assume there are 5 CPs, 3 stability policies, 2 product rates, 30 loss bins and 2 DSLAM types. Rather than focusing on profiles, of which BT's NGA network has >4000, the profile can be split and the process applied to the sub profiles. An obvious split would be to break the profile into direction, rate and error control level. The remainder of this description focuses on the downstream. Setup phase (repeated weekly, monthly, or other) For each of the 1800 combinations of risk factor select a set of lines to analyze. As a starting point this is the set of existing lines that have that risk factor combination. The lines to analyze for a factor combination may be drawn may include the lines in nearby bins, either a. Because it is know there is no impact within that region b. Because there aren't currently enough lines for the single factor combination to give a statistically valid sample. An example of a) would be that above a certain loss there is known to be no difference between the 2 product rates, so in that region the analysis set of lines are taken for both products. If there are not enough lines in a single bucket the then factor combinations that lines are drawn from for analysis from are increased until the number of lines is great enough. This is done by including more indexes from one or more factors. 1 For each risk factor consider the current index+ and i.sub.f+1 (where i.sub.f starts at zero) estimate the fraction f.sub.f of the range of the risk factor included. (e.g. at the start if loss=10 dB then the range=1.5/20=0.075) 2 Multiple by variability factor (chosen by operator), f.sub.f*v.sub.f (vf may equal infinity, to prevent expansion in that direction.) 3 For the risk factor with the smallest f.sub.f*v.sub.f set i.sub.f=i.sub.f+1. 4 Repeat if the number of lines include still not great enough. For some risk factors, e.g. CP, the expansion to include more lines must be to all CPs, unless CPs can be grouped as having similar impact on profile choice. There are a set of existing lines to analyze for this risk combination. The profiles on those lines have a rate distribution and an error control level. A function is used to select a single rate level and error correction level. For example the error control level could simply be the mode and the rate level the 70.sup.th percentile, (closer to fast.) Rate level and error control level are recorded. Different percentiles can be chosen depending on the operators priority, in high rates/low latency, fewest profile changes or something else. The expansion of the set of lines, may rest in a profile for a risk combination being outside of the product chosen by the customer. The rate levels and error control levels should be checked and brought within bounds. Now for each risk factor combination there are rate levels, error control levels both up and downstream. When a new line is provisioned or a line reset the risk factors can be analyzed and a single starting profile selected.

(58) Once enough lines have been selected selection of a single profile for this risk factor following the method described for the previous example risk factor could be used.

(59) After all combinations of risk factor have been evaluated at table like this should be complete.

(60) TABLE-US-00020 Time since last DSLAM CPE Loop fault Pro- Product CP Vendor ID Category Region (days) file SI B X CPE2 Medium A1 Never 12 MI A X CPE1 Short A1 Never 34 SI B Y CPE2 Short A2 Never 5 SI C Y CPE1 Very A2 30-100 3 Long SI C X CPE2 Medium A2 Never 4 MI B Y CPE1 Short A2 Never 1 MI C X CPE1 Very A2 Never 4 Short MI A X CPE1 Short A1 Never ? . . .
Selecting First Profile for Lines, when Commissioning a New Line or Resetting the Profile on a Line:
Look up the risk factors for this new line in the table above to select the profile. There is only one profile for each combination of risk factors.
For a line that had a fault 45 days ago and is on the SI product with CP c DSLAM Y CP1 and a very long line in region A2 apply starting profile 3.