JP2005354280A - Policy rule optimization method and apparatus thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To automate the reviewing work of a policy rule inevitable for the operation management of a network. <P>SOLUTION: In a policy server, a plurality of policy rules each comprising a condition for indicating the state of a network to be managed and an action for indicating management control to be executed in that state while both of them are network operation management policies. A history collection step collects application history information including information regarding an applied period and information for indicating an application result for each of a plurality of policy rules. A history analysis step analyzes the application history information, and collects statistical information related to the generation of conditions corresponding to the applied policy rule. A feedback step changes the condition and action for composing each policy rule based on the statistical information obtained in the history analysis step. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

The present invention relates to an operation management technology for an IP (Internet Protocol) network or an MPLS (Multi Protocol Label Switching) network, and in particular, a policy rule for optimizing a policy rule defined by a network administrator regarding network operation in accordance with a network situation. The present invention relates to an optimization method and apparatus.
With the rapid spread of broadband technologies such as ADSL (Asymmetric Digital Subscriber Line) and FTTH (Fiber to the Home), traffic in the network is rapidly increasing.

  For this reason, for an operation manager who operates an ISP (Internet Service Provider) that provides each broadband service, a carrier that provides a transmission path, an IDC (Internet Data Center) that provides a data center service via the Internet, and the like. Therefore, it is required to provide an appropriate service flexibly corresponding to a traffic fluctuation accompanying a rapid increase in traffic and a change in a trend of users targeted by each service.

Network operation managers request external requirements such as usage requirements based on contracts with each service user and operation requirements related to network service operations, performance and reliability of network equipment related to services, and each user's requirements Policy rules are created based on design goals such as service quality.
The policy rule is an operation guideline consisting of a situation (condition) indicating the occurrence of various events such as an increase in traffic and the occurrence of a line failure, and execution of appropriate network control (action) corresponding to the situation. By registering these policy rules in the policy server and performing network operation management via this policy server, the workload of the network operation manager can be greatly reduced.

The present applicant has already applied for a technique for applying an appropriate policy rule in such a policy server (Japanese Patent Application No. 2003-61996 (unpublished)).
FIG. 9 shows a configuration example of the policy server.
In the network monitoring processing unit 410 provided in the policy server shown in FIG. 9, the monitoring content setting unit 411 instructs the polling processing unit 412 to perform a polling operation in response to an instruction from the policy management processing unit 420, and receives a trap. A trap to be received is set in the unit 413. The response to polling by the polling processing unit 412 and the network management information obtained by the trap received by the trap receiving unit 413 are accumulated in a network management database (network management DB) 414. Based on the network management information collected in the network management database 414 in this way, the status notification processing unit 415 detects the occurrence of congestion, the occurrence of a line failure, and the like, and notifies the policy management processing unit 420 of the detected event. .

The policy management processing unit 420 shown in FIG. 9 includes a policy database 421, and policy rules input by the network operation manager via the user interface device 401 are stored in the policy database management unit (policy DB management unit) 423. Through the policy database 421.
As an example of the policy rule registered in the policy database 421, as shown in FIG. 10A, in response to the occurrence of a line failure (condition 1), the corresponding route is switched to the tunnel route corresponding to the route. As shown in FIGS. 10B and 10C, priority (DSCP: Differentiated Service) applied to each flow type with respect to traffic increase (condition 2) as shown in FIGS. Code-point) control (action 2-1) or control to limit the band according to each user's contract requirements (action 2-2) is instructed. In FIG. 10C, the flow related to the VoIP service of the user set with the low charge is shown as reference VoIP (2), and the flow related to the VoIP service of the user set with the normal charge is shown as reference VoIP. It distinguishes by showing (1).

Also, the policy analysis unit 424 shown in FIG. 9 analyzes the policy rules registered in the policy database 421, extracts network monitoring items to be monitored by the network monitoring processing unit 410, and monitors the above-described monitoring content setting unit. By passing the network monitoring item extracted to 411, network monitoring in accordance with the policy rule is realized.
In the technology according to Japanese Patent Application No. 2003-61996 filed by the present applicant, the network status analysis unit 425 of the policy management processing unit 420 is based on the notification from the status notification processing unit 415 provided in the network monitoring processing unit 410. The situation is analyzed, and based on the analysis result, the policy selection processing unit 426 selects the policy rule registered in the policy database 421 that best matches the analysis result.

  For example, the analysis result by the network status analysis unit 425 indicates that the traffic is increasing and that the flow type (for example, VoIP: Voice over IP) with a high priority occupies most of the traffic. In this case, the policy selection processing unit 426 replaces the policy rule including the action 2-1 applied at the time of normal congestion from the two policy rules corresponding to the condition 2 shown in FIG. Then, a policy rule that includes a bandwidth limitation according to the charge setting as an action (see action 2-2 in FIG. 10) is selected. The policy application instruction unit 427 reads out the policy rule selected by the policy selection processing unit 426 via the policy database management unit 423 and distributes a necessary instruction to the node specified by the policy rule. 428 is instructed. In this way, the action specified by the selected policy rule is executed at the appropriate node.

In this way, by selectively applying the policy rule that best suits the current state of the network from among a large number of policy rules, it is possible to promptly recover from failures such as line disconnection and congestion, and to ensure proper service. Maintenance is realized.
Conventionally, the policy rules registered in such a policy server are periodically reviewed by the network operation manager himself, and are adjusted to the actual operation of the network at that time.

At this time, the network administrator extracts packet delays and discards and fluctuations related to these events from the traffic history information, and deletes unnecessary policy rules based on this information or creates specific conditions. Work such as changing the content of the corresponding action or changing the determination threshold value for a specific condition was performed manually.
Japanese Patent Application No. 2003-61996 (unpublished) Standard technology collection (IP phone) database: Full VoIP network system control, [online], JPO, [H16.4.5], Internet <URL: www.jpo.go.jp/shiryou/s_sonota/hyoujun_gijutsu/ip_tel/0016. html> Ryuichi Horie, policy controlled network, [online], Fujitsu Limited, [H16.4.5], Internet <URL: http: //magazine.fujitsu.com/vol51-6/paper03.pdf>

  As described above, when the network operation manager manually reviews the policy rules, the network operation manager needs sufficient knowledge and experience, and thus the burden on the network operation manager is extremely large. In particular, in a policy server in which a policy rule in which a plurality of different actions are combined in one condition is registered, the number of policy rules registered in the policy database is enormous.

In addition, even for network operations managers with sufficient knowledge and experience, the conventional technology for manually building policy rules can be used to construct policy rules that are sufficiently effective in a large-scale network, It has become difficult to keep up with changes in the actual traffic on the network.
First, in an environment where various broadband services with various usage conditions and charges are being implemented, the cause of network congestion and the measures to be taken to solve it are complicated. Because. Second, as the number of users of services that require high service quality, such as VoIP (Voice over IP) services and VPN (Virtual Private Network) services, rapidly increases, the dynamic traffic increases as the overall traffic increases. This is because the changes are becoming more and more rapid, and relying on human resources will no longer be able to extract information relating to the increasing trend of traffic and temporal changes. Furthermore, since the actual state of the network is constantly changing even during the period of such review work, by the time the revised policy rule is registered in the policy server, it has already become incompatible with the actual state of the network. I can even think of it.

  In particular, the fact that policy rules aimed at preventing or resolving congestion do not match the current state of traffic in the network leads to an increase in processing load on the network equipment that constructs the network. It causes the destruction. For this reason, it is necessary to review policy rules corresponding to the increase in traffic particularly quickly.

  An object of the present invention is to provide a policy rule optimization method and apparatus capable of automating a policy rule review operation.

The first policy rule optimization method according to the present invention includes a history collection step, a history analysis step, and a feedback step.
The principle of the first policy rule optimization method according to the present invention is as follows.
Each of these is a network operation management policy, and in a policy server in which a plurality of policy rules including a condition indicating the state of a managed network and an action indicating management control to be performed in that state are registered, the history collection step is For each of the plurality of policy rules, application history information including information on the applied period and information indicating the application result is collected. The history analysis step analyzes the application history information and collects statistical information related to the occurrence of a condition corresponding to the applied policy rule. In the feedback step, the conditions and actions constituting each policy rule are changed based on the statistical information obtained in the history analysis step.

The operation of the first policy rule optimizing method configured as described above is as follows.
In the process of performing network operation management according to the policy rules registered by the policy server, the application history information is collected by the history collection step, and this collected application history information is regularly used as the history analysis step, for example. Passed. On the other hand, in the history analysis step, for example, a feature relating to a time-series distribution related to the generation period of the condition corresponding to each policy rule is extracted based on the application history information. Based on the analysis result thus obtained, the feedback step changes the policy rule by reflecting the extracted feature in the condition, for example. By registering the changed policy rule in the policy server in this way, the actual state of the network appearing as the application history information and the change thereof can be reflected in the set of policy rules.

The first policy rule optimizing device according to the present invention comprises a history collecting means, a history analyzing means, and a feedback means.
The principle of the first policy rule optimization device according to the present invention is as follows.
It is provided in a policy server with a policy database in which a plurality of policy rules consisting of conditions indicating the status of the network to be managed and actions indicating management control to be executed in that state are provided, and it matches the network to be managed. In the policy rule optimizing device for optimizing the policy rule registered in the policy database, the history collecting means, for each policy rule registered in the policy database, relates to the application of the policy rule and information regarding the occurrence timing of the condition and Collect application history information including information indicating application results. The history analysis means analyzes the application history information collected by the history collection means, and extracts statistical information related to the occurrence of a condition corresponding to the applied policy rule. The feedback means changes the conditions and actions constituting each policy rule based on the statistical information obtained by the history analysis means.

The operation of the first policy rule optimizing device configured as described above is as follows.
In the process in which the policy server executes the operation management of the network to be managed based on the policy rules registered in the policy database, the application history information on each policy rule is collected by the history collection means. To the history analysis means. On the other hand, the history analysis unit extracts, for example, features related to a time-series distribution regarding the generation period of the condition corresponding to each policy rule based on the application history information. Based on the analysis result thus obtained, the feedback means changes the policy rule by reflecting the extracted feature in the condition, for example. By registering the changed policy rule in the policy server in this way, the actual state of the network appearing as the application history information and the change thereof can be reflected in the set of policy rules.

A second policy rule optimizing apparatus according to the present invention is the above-described first policy rule optimizing apparatus, wherein the history analyzing means includes a distribution information collecting means and a distribution analyzing means, and the feedback means includes a discriminating means. Period setting means.
The principle of the second policy rule optimization device according to the present invention is as follows.
In the history analysis unit provided in the first policy rule optimizing device described above, the distribution information collection unit includes a time series relating to the application period from the occurrence time of the event related to the condition to the end time for each applied policy rule. Collect information that shows the distribution in. Also, in the history analysis means, the distribution analysis means analyzes the information collected by the distribution information collection means to obtain a bias regarding the time-series distribution of the application period of each policy rule, and statistically analyzes this bias. Is output as a feature obtained by. In the feedback means, based on the determination means and a predetermined criterion, the policy rule having a large bias regarding the time-series distribution of the application period corresponding to each policy rule is determined. Similarly, in the feedback means, the period setting means changes the policy rule determined by the determination means to a policy rule in which an action schedule is set based on the distribution range of the corresponding application period.

The operation of the second policy rule optimizing device configured as described above is as follows.
For example, when congestion occurs on a specific day of the week or a specific time zone, and policy rules that include actions to avoid the congestion state are applied repeatedly, the distribution provided in the history analysis means Analysis result indicating that the distribution of the policy rule application period is biased to, for example, a specific day of the week or time zone by analyzing the distribution of the policy rule application period collected by the information collecting means by the distribution analysis means Is obtained. The discriminating means provided in the feedback means discriminates such a policy rule by, for example, comparing the bias obtained by the distribution analyzing means with a predetermined threshold value, and provides it to the processing of the period setting means. This period setting means is, for example, a period including a time zone in which the above-described policy rule is frequently applied, that is, a period earlier than the start time of the distribution range of the application period and the end time of the distribution range. Change to a policy rule that stipulates a schedule for performing appropriate actions during the period until late. If the policy rule changed in this way is registered in the policy server, the action for avoiding the congestion is executed prior to the start time of the time when the congestion has repeatedly occurred, and the time when the congestion has repeatedly occurred After the band ends, execution of actions for avoiding congestion can be stopped.

A third policy rule optimizing apparatus according to the present invention is the above-described first policy rule optimizing apparatus, wherein the history collecting means includes success / failure information collecting means, and the history analyzing means includes success / failure information totaling means and evaluation means. And the feedback means includes priority order changing means.

The principle of the third policy rule optimization device according to the present invention is as follows.
In the history collecting unit provided in the first policy rule optimizing device described above, the success / failure information collecting unit generates an event for a policy rule that allows a plurality of actions to be selected for a condition indicating the occurrence of the event of interest. Success / failure information indicating whether or not the intended change has been obtained is collected for each action selected corresponding to. In the history analysis means, the success / failure information totaling means has a purpose by executing each of the selectable actions for the policy rule that can select a plurality of actions based on the success / failure information collected by the success / failure information collecting means. Aggregate the number of successes achieved. Further, in the history analysis means, the evaluation means evaluates the effectiveness of each action to achieve its purpose based on the number of successes collected for each action, and the evaluation result is the statistical feature analysis result. Output as. In the feedback means, the priority order changing means changes to a policy rule that selects an action evaluated as having a high degree of effectiveness in accordance with the condition based on the evaluation result by the evaluation means.

The operation of the third policy rule optimizing device configured as described above is as follows.
Based on the success / failure information corresponding to each action collected by the success / failure information collecting means provided in the history collection means, the success / failure count for each action is totalized by the success / failure information totaling means of the history analysis means, and further, Based on this, the effectiveness of each action is evaluated by the evaluation means.

  Thus, for example, for a policy rule that executes an action of selecting and switching one of a plurality of tunnel routes according to the occurrence of congestion, avoiding a congestion state by switching to the tunnel route for each switching destination tunnel route It is an index for objectively evaluating the success rate of congestion avoidance when each tunnel route is selected by collecting success / failure information indicating success or failure of the tunnel and performing the above-described analysis based on the success / failure information. A certain degree of effectiveness can be obtained.

  Based on the effectiveness obtained in this way, the priority change means, for example, changes the priority related to the selection of the switching destination tunnel route in order from the tunnel route with the highest effectiveness, thereby generating a congestion state. It is possible to automatically optimize the policy rule for switching to the tunnel route corresponding to the one that can be expected to avoid the congestion state with a high success rate.

A fourth policy rule optimizing device according to the present invention is the above-described first policy rule optimizing device, wherein the history collecting unit includes an individual period collecting unit, and the history analyzing unit includes a distribution investigation unit and a detecting unit. Provided, and the feedback means comprises a setting change means.
The principle of the fourth policy rule optimizing device according to the present invention is as follows.
In the history collecting means provided in the first policy rule optimizing device described above, the individual period collecting means includes a policy rule including a condition for detecting an excess of a request amount for resource allocation set for each of a plurality of service forms. For each service type, information on the occurrence period of the event related to the condition is collected.

  In the history analysis means, the distribution investigation means examines the distribution of event occurrence periods for each service form based on the information collected by the individual period collection means. Similarly, in the history analysis means, the detection means detects a combination of service forms in which the occurrence periods of events are exclusively distributed based on the investigation result by the distribution investigation means, and the detection result is statistically characterized. Output as analysis result. In the feedback means, the setting change means changes the policy rule so as to set a shared resource that can be assigned to both of the processes related to these service forms for the combination of service forms detected by the detecting means.

The operation of the fourth policy rule optimizing device configured as described above is as follows.
The individual period collection means provided in the history collection means has generated resource requests exceeding the allocation for each of a plurality of service forms including, for example, a VOD (: Video on Demand) service, a video conference service, and a live broadcast service. Information about the period is collected as history information and passed to the history analysis means.

  Based on such history information, a survey result indicating that a request for resources exceeding allocation has occurred in different time zones for the VOD service and the video conference service, for example, by the distribution survey means provided in the history analysis means. Is obtained, the detecting means detects the above-described VOD service and video conference service as a combination corresponding to the condition.

  Based on the detection result obtained in this manner, the setting changing unit can set a suitable shared resource for, for example, the VOD service and the video conference service, thereby enabling flexible resource allocation. As a result, policy rules that stipulate responses to requests that exceed allocated resources are automatically changed to enable resource allocation that matches the actual usage status of each service type, and policy rules can be optimized. Can be realized.

  According to the first policy rule optimizing method and the first policy rule optimizing apparatus according to the present invention, the characteristics of the network managed by the policy server are extracted by analyzing the accumulated application history. A set of policy rules reflecting features can be automatically constructed. As a result, an optimal policy rule can be formed in accordance with a change in the characteristics of the network to be managed, so that network traffic can be managed more efficiently.

  In particular, by extracting characteristics related to the bias in the distribution of the occurrence period of events such as congestion, and changing to policy rules that stipulate a schedule for performing appropriate actions during the period when events occur frequently based on this characteristic Therefore, it is possible to prevent the occurrence of an unfavorable event such as congestion in advance, and to reliably prevent the effects associated with such an event, that is, increase in packet delay and packet loss. Thereby, the reliability of service provision via the network can be greatly improved, and the customer satisfaction can be improved.

  Furthermore, for policy rules that select and execute multiple different actions according to the occurrence of events such as congestion, objectively evaluate the effectiveness of each action based on the accumulated history information, and based on this evaluation result By changing the action selection order, the policy server can preferentially select an action that is highly effective in the current state of the network, so that an event such as congestion can be quickly and reliably terminated.

  In addition, regarding policy rules that stipulate responses to requests that exceed allocated resources, based on accumulated history information, the degree of agreement between the actual status of requests from each service type and allocated resources, and the trend of requests from each service type By assigning an appropriate shared resource based on the evaluation result, it is possible to allocate resources that match the actual usage status of each service form. By optimizing such policy rules, it becomes possible to allocate and operate limited resources effectively with a plurality of service forms without waste.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 shows an embodiment of a policy server provided with a policy rule optimizing apparatus according to the present invention.
1 that are the same as those shown in FIG. 9 are denoted by the same reference numerals as those shown in FIG. 9, and description thereof is omitted.

The policy server shown in FIG. 1 includes a policy optimizing unit 210 corresponding to the policy rule optimizing device according to the present invention in addition to the units provided in the policy server shown in FIG.
In the policy optimization unit 210 illustrated in FIG. 1, the history information collection unit 212 receives information indicating a policy rule to be applied from the policy application instruction unit 211, and receives information on the network status from the network status analysis unit 425. The information is received and accumulated in the history information database 213. Further, the history information analysis unit 214 shown in FIG. 1 analyzes the history information accumulated in the history information database 213, and uses the analysis result for policy rule editing processing by the policy editing unit 215. The policy rule edited by the policy editing unit 215 is registered in the policy database 421 via the policy database management unit 423.

FIG. 2 shows an example of policy rules. 3 shows a detailed configuration of the policy optimization unit 210, and FIGS. 4 and 5 show examples of history information collected in the history information database.
First, the application history information collection function will be described.
In the history information collection unit 212 illustrated in FIG. 3, the application frequency counting unit 221 stores each policy rule in the history information database 213 every time the policy application instruction unit 211 instructs the policy application unit 428 to distribute the policy rule. The count value held correspondingly is added. As a result, as shown in FIG. 4A, the application frequency is counted for each of the policy rules indicated by numbers 1 to n in FIG. 2A, and the count result is stored in the history information database 213. Retained. Further, the period information collection unit 222 illustrated in FIG. 3 collects period information including the date and time when the application was started and the date and time when the application was canceled for each policy rule instructed to be applied by the policy application instruction unit 211. Then, the number indicating the applied policy rule and the period information collected corresponding to the policy rule are held in the history information database 213 as basic application history information (see FIG. 4B).

On the other hand, the flow information addition unit 223 and the action result collection unit 224 illustrated in FIG. 3 hold additional history information, which will be described later, in the history information database 213 in response to an instruction from the collection control unit 225.
For example, the collection control unit 225 applies a policy rule that defines an action for selecting and switching a plurality of tunnel routes in order from the candidate 1 like the policy rule indicated by number 11 in FIG. In this case, the option designated by the instruction from the policy application instruction unit 211 is passed to the action result collection unit 224 together with a number (for example, number 11) indicating the applied policy rule, and the option is selected. Directs collection of action results. In response to this, the action result collection unit 224 indicates, for example, whether or not the network status analysis unit 425 has been notified that the intended result such as the elimination of the congestion state has been obtained after the execution of the selected action. The action result is collected, and the combination of the action result and the above-described options is stored in the history information database 213 as additional history information regarding the action result of the corresponding application history information (see FIG. 5A). FIG. 5A shows an example in which “success” is held as an action result when a desired result is notified, and “failure” is held as an action result in other cases.

  In addition, the collection control unit 225 stipulates a countermeasure when a request from each of the plurality of service forms exceeds the allocated resource amount, such as the policy rule indicated by the number 12 in FIG. When the policy rule is applied, information indicating the service form that is the target of the action specified by the instruction from the policy application instruction unit 211 and the content of the executed action (for example, permitting or denying the request) are flowed The information is added to the information adding unit 223. In response to this, the flow information adding unit 223 adds these pieces of information to the application history of the corresponding policy rule held in the history information database 213 as additional history information regarding the usage distribution for each service form ( (Refer FIG.5 (b)).

Next, a function for analyzing the application history information collected in this way will be described.
In the history information analysis unit 214 shown in FIG. 3, the history information reading unit 226 reads the application history information stored in the history information database 213 once a month, for example, and applies the application frequency ( 4 (a)) to the application frequency evaluation unit 227, additional application history information (see FIG. 5 (b)) regarding basic application history information (see FIG. 4 (b)) and usage distribution for each service form. ) To the period distribution analysis unit 228. Further, the history information reading unit 226 passes application history information (see FIG. 5A) including additional information related to the action result to the effectiveness evaluation unit 229.

  The application frequency evaluation unit 227 illustrated in FIG. 3 sorts the application frequencies related to each policy rule received from the history information reading unit 226, for example, and determines whether the application frequency is greater than a preset threshold value. The policy rule application frequency is evaluated, the evaluation result is passed to the policy deletion unit 234 of the policy editing unit 215, and the policy deletion unit 234 determines the policy rule to be deleted.

  For example, the policy deletion unit 234 deletes a policy rule whose application frequency is determined to be smaller than a predetermined threshold by the application frequency evaluation unit 227 as an unnecessary policy rule with a very low application frequency, so that the policy database 421 is deleted. The set of policy rules registered in the policy database 421 can be optimized by optimizing the number of policy rules registered in the policy database 421.

  Further, the period distribution analysis unit 228 shown in FIG. 3 examines the correlation between the period and time when each policy rule is applied based on the basic application history information received from the history information reading unit 226. For example, by counting the number of days each policy rule is applied at each time of the day, as shown in FIG. 6A, the distribution of the period during which each policy rule is applied can be obtained. In FIG. 6A, there is a clear distribution bias between policy rule 2 (shown as policy 2 in FIG. 6) and policy rule 8 (shown as policy 8 in FIG. 6). For example, when the distribution range of the application period obtained as described above is smaller than a predetermined threshold value determined in advance, the period distribution analysis unit 228 is obviously biased and the application period is distributed. For the corresponding policy rule (for example, policy rules 2 and 8), information indicating the distribution range of the application period is output as an analysis result of the period distribution, and passed to the schedule setting unit 231 of the policy editing unit 215, The schedule setting unit 231 uses the schedule setting process.

  Based on the analysis result received from the period distribution analysis unit 228, the schedule setting unit 231 illustrated in FIG. 3 performs the purpose over a period corresponding to the distribution range of the application period, as illustrated in FIGS. The policy rule which prescribes | regulates the action containing the delivery schedule for delivering the instruction | indication required to the delivery destination is made. It is desirable that the schedule setting unit 231 sets the distribution schedule in anticipation of an appropriate grace period before and after the distribution range of the application period corresponding to each policy rule.

  In this way, the policy rule (eg, policy rules 2 and 8) is focused on the policy rule distribution schedule newly created based on the application history of each policy rule registered in the policy database 421. This reflects the trend of when certain events occur in the network, that is, the dynamic change trend of traffic in the network.

  Therefore, after these policy rules are registered in the policy database 421 via the policy database management unit 423, before the original policy rule is read and distributed by the policy application instruction unit 211, FIG. It can be expected that new policy rules as shown in (b) and (c) are read from the policy database 421 in accordance with the distribution schedule, distributed to an appropriate distribution destination, and applied. In other words, as in policy rule 2 and policy rule 8, priority control is not executed after an event to be avoided (that is, the traffic threshold value for each line is exceeded), but the change trend of traffic in the network is read ahead. Thus, priority control can be executed before congestion occurs, and the occurrence of events to be avoided can be suppressed.

  When the original policy rule is hardly applied with the registration of the new policy rule described above, the policy deletion unit 234 determines that the policy deletion unit 234 is based on the evaluation result by the application frequency evaluation unit 227 shown in FIG. The corresponding policy rule is deleted from the policy database 421. In this way, by deleting unnecessary policy rules that have lost their roles as new policy rules are registered, a set of policy rules registered in the policy database 421 can be optimized.

On the other hand, when the policy rules (for example, policy rules 2 and 8) for executing countermeasures according to the occurrence of an event are stored in the policy database 421, a change occurs in the actual traffic in the network. Can be dealt with flexibly.
For example, if the time zone during which traffic concentrates on the line of interest shifts, the policy rule that defines the distribution schedule as described above cannot be dealt with, and accordingly, the original policy rule 2 And policy rule 8 is applied. At this time, by analyzing the application history of these policy rules as described above, identifying the time zone when new traffic is concentrated, and creating a policy rule that defines a new distribution schedule Further, it is possible to further optimize the policy rules in response to changes in the actual traffic situation.

Next, a method for optimizing the policy rule will be described by paying attention to additional history information regarding the choice selected in the action and the action result.
Based on the information received from the history information reading unit 226 shown in FIG. 3, the effectiveness evaluation unit 229 selects, for each policy rule, the number of times each option is selected in the action when the policy rule is applied and the selection thereof. The action results obtained as a result of are aggregated. For example, when the application history information as shown in FIG. 5A is accumulated for the policy rule 11 shown in FIG. 2B, the effectiveness evaluation unit 229 shows the policy rule 11 shown in FIG. As described above, in the action, the switching routes Tunnel-2, 3 and 4 listed as candidates 1 to 3 are each selected, the number of times the switching avoids the congestion of the monitoring point tunnel route, and the switching route fails. The number of times is counted, and the effectiveness as an action is evaluated for switching to the switching route Tunnel-2, 3 and 4 based on the counting result. For example, the effectiveness evaluation unit 229 shown in FIG. 3 simply divides the number of successes S by the number N of times each option (switching path in this case) is selected, and this value is related to the effectiveness of each option. The evaluation result is output and passed to the priority order changing unit 232 of the policy editing unit 215, and the priority order changing unit 232 uses the priority order changing process for each option.

  As can be seen from FIG. 5A, when the policy route 11 is applied, when the switching route Tunnel-2 that is the highest priority candidate is selected, the tunnel route that is the monitoring point in most cases. The traffic amount of Tunnel-1 has failed to be reduced, while the action executed by selecting the switching route Tunnel-3 immediately after that has succeeded in reducing the traffic amount of the tunnel route Tunnel-1 without fail. . This tendency clearly appears in the evaluation result by the effectiveness evaluation unit 229 shown in FIG. 7A. This indicates that the tunnel route Tunnel-1 as a monitoring point in the network managed by the policy server. As the alternative route, the actual state of traffic in the network is shown in which the switching route Tunnel-3, which is the second candidate, is more appropriate than the switching route Tunnel-2.

  In the range of the application history information shown in FIG. 5A, the usefulness of the third candidate switching route Tunnel-4 is unknown (in FIG. 7, it is indicated by the symbol “-”). ) This switching path Tunnel-4 may be more useful than the switching path Tunnel-2, which is clearly known to be less useful. Therefore, when the usefulness cannot be calculated based on the application history information in the accumulated range as described above, the effectiveness evaluation unit 229, for example, selects the average value of the usefulness related to other options as the corresponding option. The priority level changing unit 232 may be notified of the usefulness of.

  Based on the evaluation result obtained in this way, the priority changing unit 232 defines a policy rule that defines an action in which the priority is set in descending order of usefulness to the options specified in the action of the corresponding policy rule. Is newly created and registered in the policy database 421 instead of the original policy rule (for example, policy rule 11) via the policy database management unit 423.

  Thus, based on the application history information accumulated in the history information database 213, a new policy for executing an action for selecting an option (for example, a switching route) that matches the actual state of the network rather than the original policy rule. Rules can be created. Such a new policy rule is registered in the policy database 421, and a policy rule that prescribes an action for selectively executing one of a plurality of options is accumulated as in the policy rule 11 shown in FIG. A set of policy rules registered in the policy database 421 is changed by changing the policy rule that defines an action for selecting a highly useful option in the actual state of the network reflected in the application history information to a high priority. Can be optimized.

Of course, the policy rule newly registered as described above defines an action for selecting a highly useful option with high priority in the actual state of the network reflected in the accumulated application history information. Depending on the event observed at the monitoring point, it can be expected that the most effective action is executed with a higher probability.
In addition, the policy rule optimization process as described above corresponds to the condition 2 shown in FIG. 10, and the action 2-1 for controlling the priority for each flow type or the action for limiting the bandwidth according to the charge setting It can also be used for optimization related to policy rules that specify that 2-2 is executed.

  For example, if there is a tendency for the traffic set at a monitoring point to increase because a user with a low charge uses a service with high priority such as VoIP, the priority of the flow related to each service is lowered. In action 2-1, there is a high possibility that the traffic volume at the monitoring point cannot be reduced. On the other hand, if the action 2-2 for limiting the bandwidth that can be used by the user associated with the contract for which the low fee is set is executed, the traffic volume of the monitoring point is surely reduced. Therefore, if the application history information accumulated as described above is analyzed by the effectiveness evaluation unit 229 for the policy rule in which the action 2-1 and the action 2-2 are defined, naturally, the action 2 -2 is considered to be evaluated as being more useful than Action 2-1.

  In response to this, the priority changing unit 232 creates a policy rule that preferentially selects the action 2-2 and registers it in the policy database 421 instead of the original policy rule. A policy that reduces the traffic volume by the flow control method by selecting according to the increase of traffic volume at the monitoring point in accordance with the fact that the flow related to the service with high priority by the set user is dominant Rules can be optimized.

  Applying policy rules optimized in this way minimizes the impact on communications related to users who should be prioritized based on the fee structure (that is, users with high usage charges). It is possible to reliably reduce the traffic at the monitoring point while keeping it at the same level. As a result, it is possible to provide the user with comfort in accordance with the fee structure, and in particular, it is possible to improve the customer satisfaction of the user who has a contract with a high usage fee set.

Next, a method for optimizing a policy rule will be described by paying attention to additional history information related to the service form that is the action target and the executed action.
Based on the additional application history information regarding the usage distribution for each service form received from the history information reading unit 226 (see FIG. 5B), the period distribution analyzing unit 228 determines the resource usage status according to each service form as time. Analyze by band. For example, based on the application history information accumulated for each service related to the policy rule 12 shown in FIG. 5B, the period distribution analysis unit 228 allocates resources for each service form for each day of the week and time period. The number of requests and the number of times the request is permitted are totaled (see FIG. 8A). In FIG. 8 (a), the service name indicating each service form and the number of times that the service form requested resource allocation are separated by a colon, and the number of times the request is permitted is enclosed in parentheses. Indicated.

  In the aggregated results obtained in this way, usage requests for video conferencing services are concentrated in the morning on weekdays, whereas usage requests for video on demand (VOD) services are in the afternoons and nights on Saturdays and Sundays. It is shown that it is concentrated (see FIG. 8A). For example, the period distribution analysis unit 228 obtains a correlation related to the distribution of usage requests for each service type for each combination of selecting two of the three service types for which resource allocation is defined in the policy rule 12. Based on the value indicating the correlation, the trend of the usage situation regarding the VOD service and the video conference service as described above can be grasped by detecting the combination of the service forms in which the time periods when the requests are generated are exclusively distributed. To do. The period distribution analysis unit shown in FIG. 3 shows the information indicating the combination of the service types in which the usage requests are exclusively distributed and the information indicating the rate at which the requests regarding each service type are permitted. 228 outputs the analysis result of the application history information related to the policy rule that stipulates how to deal with the excess of the requested amount with respect to the resource allocation amount, and passes it to the shared resource setting unit 233 of the policy editing unit 215, and this shared resource setting unit 233 It is used for resource allocation change processing by.

  For example, based on the result of aggregation shown in FIG. 8A, the resource request related to the VOD service and the resource request related to the video conference service are mutually exclusive based on the analysis result passed from the period distribution analysis unit 228 to the shared resource setting unit 233. It is shown that resource requests related to these services are both rejected with a high probability because the requested resource amount is exceeded.

  Based on such an analysis result, the shared resource setting unit 233 removes some of the resources that are uniquely allocated to the services in which the resource requests are exclusively distributed from the respective allocated resources, and removes these resources. As a shared resource that can be shared by each service, a new policy rule is created to reject the resource request when the resource request from each service exceeds the total of the allocated resource and the shared resource unique to each service.

  For example, based on the analysis result described above, the shared resource setting unit 233 removes 20 MB and 10 MB from the 40 MB and 30 MB allocated resources respectively allocated to the VOD service and the video conference service in the original policy rule 12. 30MB is set as a shared resource, and as shown in FIG. 8B, in addition to dealing with resource requests for resources allocated for each service type, resource requests for shared resources are permitted up to 30MB. Create a policy rule that specifies

In this way, by changing the resource allocation method based on the application history related to the policy rule that regulates the response to the resource request for the resource allocated for each service type, the policy rule as described above is optimized, It is possible to realize resource operation that matches the usage status of each service appearing in the application history.
If the policy rule created as described above (see FIG. 8B) is registered in the policy database 421 in place of the original policy rule, resource requests by each service will be handled in accordance with the new policy rule. Since countermeasures are taken, both the VOD service and the video conference service are permitted to request resources up to 50 MB.

  In other words, by using resources allocated to the VOD service that was not used at all on weekday mornings, a video conferencing service with a large amount of resource requests was realized, while on Saturday and Sunday afternoons and nights. By using the resources allocated to the video conference service that has not been used at all, a VOD service with a large amount of resource requests can be realized.

  As described above, according to the policy rule optimized as described above, it is possible to effectively use limited resources by interchanging between a plurality of services.

As described above, the policy rule optimization method and apparatus according to the present invention accumulates history information related to the application of policy rules, and based on the result of statistical analysis of the accumulated history information, Rules can be automatically and quickly optimized to match the current state of the network and changing trends.
As a result, an operation manager who operates an ISP that provides each broadband service, a carrier that provides a transmission path, an IDC that provides a data center service via the Internet, etc. creates work and policy rules for grasping the current state of the network. Since the burden on work can be greatly reduced, it is extremely useful in the field of network operation management.

  In particular, it is a fatal obstacle in various network services by analyzing the temporal change trend of traffic in the network and making it possible to create policy rules optimized according to the trend itself. In view of the fact that it is possible to prevent line congestion and associated packet loss, it can be said that the policy rule optimizing method and apparatus according to the present invention play a very important role in the field of network operation management. .

  In addition, by providing the accumulated history information and statistical analysis results to the operation manager, it is possible to assist the operation manager in formulating higher-level business development, and to It also contributes to information gathering in areas such as exploration.

It is a figure which shows embodiment of the policy rule optimization apparatus concerning this invention. It is a figure which shows the example of a policy rule. It is a figure which shows the detailed structure of a policy optimization part. It is a figure which shows the example of application history information. It is a figure which shows the example of application log information. It is a figure explaining the operation | movement which extracts the characteristic regarding distribution of an application period. It is a figure explaining the operation | movement which investigates the usefulness of an option. It is a figure explaining the feature extraction regarding the use time zone for every service classification. It is a figure which shows the structural example of a policy server. It is a figure which shows the example of a policy rule.

Explanation of symbols

210 Policy optimization unit 211, 427 Policy application instruction unit 212 History information collection unit 213 History information database 214 History information analysis unit 215 Policy editing unit 221 Application frequency counting unit 222 Period information collection unit 223 Flow information addition unit 224 Action result collection unit 225 Collection control unit 226 History information reading unit 227 Application frequency evaluation unit 228 Period distribution analysis unit 229 Effectiveness evaluation unit 231 Schedule setting unit 232 Priority change unit 233 Shared resource setting unit 234 Policy deletion unit 401 User interface device 410 Network monitoring processing Unit 411 monitoring content setting unit 412 polling processing unit 413 trap receiving unit 414 network management database (network management DB)
415 Status notification processing unit 420 Policy management processing unit 421 Policy database 423 Policy database management unit (policy DB management unit)
424 Policy analysis unit 425 Network status analysis unit 426 Policy selection processing unit 427 Policy application instruction unit 428 Policy application unit

Claims (5)

Each is a network operation management policy, and in a policy server in which a plurality of policy rules consisting of a condition indicating the state of the network to be managed and an action indicating management control to be executed in that state are registered,
For each of the plurality of policy rules, a history collection step for collecting application history information including information about an applied period and information indicating an application result;
A history analysis step of analyzing the application history information and collecting statistical information related to the occurrence of a condition corresponding to the applied policy rule;
A policy rule optimizing method comprising: a feedback step of changing a condition and an action constituting each policy rule based on the statistical information obtained in the history analysis step. Provided in a policy server having a policy database in which a plurality of policy rules including a condition indicating a state of a network to be managed and an action indicating management control to be performed in the state are registered. In addition, in the policy rule optimizing device that optimizes the policy rule registered in the policy database,
For each policy rule registered in the policy database, with respect to application of the policy rule, history collection means for collecting application history information including information on the occurrence time of the condition and information indicating the application result;
History analysis means for analyzing application history information collected by the history collection means and extracting statistical information relating to the occurrence of a condition corresponding to the applied policy rule;
A policy optimization apparatus comprising: feedback means for changing conditions and actions constituting each policy rule based on statistical information obtained by the history analysis means. The policy optimization device according to claim 2, wherein
The history analysis means includes
A distribution information collecting means for collecting information indicating a distribution in a time series regarding an application period from an occurrence time to an end time of an event related to a condition for each applied policy rule;
A distribution analysis that analyzes the information collected by the distribution information collecting means to obtain a bias regarding the time-series distribution of the application period of each policy rule, and outputs the bias as a feature obtained by statistical analysis Means and
The feedback means includes
A discriminating means for discriminating a policy rule having a large bias with respect to a time-series distribution of an application period corresponding to each policy rule based on a predetermined criterion;
A policy optimizing apparatus comprising: a period setting unit that changes the policy rule determined by the determining unit to a policy rule in which an action schedule is set based on a distribution range of a corresponding application period. The policy optimization device according to claim 2, wherein
The history collection means obtains a desired change for each action selected corresponding to the occurrence of the policy rule that allows a plurality of actions to be selected for the condition indicating the occurrence of the event of interest. A success / failure information collecting means for collecting success / failure information indicating whether or not
The history analysis means includes
Based on the success / failure information collected by the success / failure information collection means, the policy rules for which the plurality of actions can be selected, for each selectable action, the number of successes achieved by the execution of the action is aggregated. Success / failure information aggregation means;
Evaluation means for evaluating the effectiveness of each action to achieve a purpose based on the number of successes totaled for each action, and outputting the evaluation result as an analysis result of a statistical feature;
The feedback means includes priority order changing means for changing the action rule evaluated as having high effectiveness to a policy rule that selects the action with high priority based on the evaluation result by the evaluation means. Policy optimization device characterized by The policy optimization device according to claim 2, wherein
The history collecting means collects information on the occurrence period of an event related to a condition for each service form with respect to a policy rule including a condition for detecting an excess of a requested amount for resource allocation set for each of a plurality of service forms. With individual period collection means,
The history analysis means includes
Distribution investigation means for examining the distribution of the occurrence period of the event for each service form based on the information collected by the individual period collection means;
Detecting means for detecting a combination of service forms in which the occurrence periods of the events are exclusively distributed based on the investigation result by the distribution investigation means, and outputting the detection result as an analysis result of a statistical feature; With
The feedback means includes a setting change means for changing a policy rule so as to set a shared resource that can be assigned to both of the processes related to the service forms for the combination of service forms detected by the detecting means. Feature policy optimization device.

Images