Zhu Zesheng, Sun Ling
Telecommunication network management system (TNMS) is one of the most important tools to operate high efficiently a small telecommunication network. In the past years, how to build fast and reasonably TNMS has been being a very important and complex problem to be solved so as to improve the quality of network operation. Due to the vary kinds of new techniques and equipment used widely in the networks, the requirement for ability to analyze and process the spatial information about the network operation has largely and fast increased in TNMS. This paper discusses mainly the key technique to develop the real time and high quality TNMS based on PC ARC/INFO. Thus, this paper introduces a graphic model for the operation management of small telecommunication network, which is based on the computer network clustering construction theory and can be implemented in the environment of PC ARC/INFO. Further, this new theory transforms a complex problem of the operation management of network into a simple problem about the graphic processing in the graphic model. On the other hand, the excellent ability of PC ARC/INFO to analyze and process the spatial information, including the operation of various complex graphic information,is used to implementthisgraphic model.
1.INTRODUCTION Telecommunication network has very wide meaning, and is usually a complex system which may be a computer network, telecommunication network and ISDN network, etc.. A management system of telecommunication network is really a very important tool for operating effectively the network, and becomes also a very important part of today telecommunication networks (Cassel, 1989). While the complexity of telecommunication networks is increasing,the importance of management system for the operation of telecommunication network is also increasing. Obviously, this importance may be the same as that of common computer operation system for computer operation, and further, the former may be more important. In general, a telecommunication network system of high efficiency and performance must have a network management system of good performance. Otherwise, it is not possible that this telecommunication network is effectively operated. Further, because telecommunication network is usually also a computer communication system of high speed operation, the time performance of this operation management system is very important for the real time performance of telecommunication network. In the management of telecommunication network of the above discussion, the performance management of network is one of very important problems to be solved(Ball, 1992). The performance management includes usually the management of the time measurement, e.g.the average delay time, information processing speed, e.g. the information throughput, and the operation status of the network which relates to provide the various services for users of the network. In fact, the key task of the performance management of network is to adjust reasonably the information flow in the network so as to make the operation status of the network satisfactory and to increase the efficiency of operation of the network infrastructure as largely as possible. However, the performance management of network is still a very complex problem in the current time, and so far is not solved by any satisfactory method. The major reasons of production of this situation are that telecommunication network is a very complex computer system and that to describe the performance and status of the network by an uniform and global view is very difficult. Thus, it is very difficult to define a standard for specifing and determining the performance of telecommunication network. So far, we have really no any effective method for managing the performance of operation of telecommunication network (Tjaden, 1991). Besides the above reasons, the performance managements in the current management techniques of telecommunication network are only to diagnose the network faults, to recognize reason of the network faults and to recover them to the normal operation (Lewis,1993). Thus, this performance management of network is really the performance management of a passive-response mode, and can not provide the performance management including the information flow management in a uniform and global view. Especially, the ATM(asynchronous transfer mode) network is being widely used in local or wide area so that the importance of the performance management of network, including the information flow management, is increasing. Obviously, the ATM network without the support of this performance management is certainly low efficiency. Further, the performance management, including the information flow management, in future telecommunication networks will become a very important task of both the performance management of network and the network operation (Samani, 1993). In general, the above information flow management includes two main contents as the following: 1)Manage the information flow pattern of total network in order to make the network operation on the pattern as good as possible; 2)Recognize abnormal information flow pattern, find out the reason to result in this pattern, and adjust it to a normal pattern. This paper is divided into six major parts, and discusses a new method for the real time management of small telecommunication network. By this method, one uniform and global view can be built and used to define the operation status of small telecommunication network. Otherwise, an effective model of the performance management of network is used to manage this global view of small telecommunication network. In the second part, a general model for supporting the real time system for operating small telecommunication network and how to apply it to describe the operation status of network are discussed respectively in detail. In the third part, the PC ARC/INFO implementation of this model and some key techniques related to it are discussed. In the fourth part, some typical operation process of the real time management system based on this model is discussed in order to show the implementation of the above key technique. In the fifth part, a typical example is discussed for showing a general method of implementing this real time management system. In the sixth part, some further works related to this real time management system are discussed. 2.GRAPHIC MODEL FOR REAL TIME MANAGEMENT How to apply an uniform view to describe the operation status of telecommunication network is the first problem to be solved for building a model for operating the network (Snodgrass, 1988). We note that, for any telecommunication network, supposing it as N0, its design and implementation are usually completed in an environment E0 which is described by a ideal and the supposed parameters. Thus, we define a sequential pair of (N0, E0) as a status of the network N0. In fact, after the telecommunication network is operated, both N0 (here, N0 is also considered as a final design and implementation of the network) and E0 are variable. In general, the faults and changes of some components in N0 make the performance of N0 change. Further, we can use a weighted special topology to represent the status of the network N0. The weight is usually used to describe some physical parameters, including the information flow, delay time, throughput, cost and reliability, etc.. Obviously, the changes of N0 can be represented in the changes of the weighted parameters and the special network topology. Otherwise, all values of these parameters of N0 can be divided into the two spaces Ns1 and Ns2. The N0 in Ns1 represents that the status of N0 is acceptable, i.e., the operation status of N0 does not degrade the primary goal of the performance of design of N0. on the other band, the N0 in Ns2 represents that the status of N0 is not acceptable, i.e., the operation status of N0 degrades the primary goal of the performance of design of N0 . Because the E0 is also variable, in the same way, the all values of E0 can be also divided into the two spaces, the Es1 and Es2. Further, the values of E0 in Es1 represents that the status or values of E0 are accepted, i.e., the operation environment E0 of N0 satisfies the primary goal of design environment of N0. Otherwise, the values of E0 in Es2 represent that the status or values of E0 is not accepted, i.e., the operation status of N0 in E0 degrade the primary goal of the performance of design of N0. Thus, we can obtain the description of four combination spaces based on the Ns1, Ns2, Es1 and Es2, which is discussed as the following: XS11=(N0 in Ns1, E0 in Es1), XS11=(N0 in Ns1, E0 in Es1) XS21=(N0 in Ns2, E0 in Es1), XS22=(N0 in Ns2, E0 in Es2) Obviously, the set XS=(XS11, XS12, XS21, XS22) describes all possible operation status of telecommunication network, where the XS11 represents the normal operation status of N0, and the XS12, XS21 and XS22 represent the abnormal operation status of N0. Further, we can find some important relationships between the above one status and the other status. For example, the status XS12 may be transformed to the status XS11 by the adjustment of the performance management of N0. The status X21 represents that there are some faults in N0 and this status must be recovered to the normal status by the recovery management of faults of N0. In the most situations of application, the status XS22 is transformed into the status XS11 by the performance and fault managements of N0. The four status about the operation management of telecommunication network can be described as the relation XS22->XS21- >XS12->XS11 and XS22->XS12->XS21->XS11. Obviously, this relation describes the decision of operation management of telecommunication network, i.e., how to transform the worst abnormal status (for example, XS22) into the normal status. In general, there are the two paths in this relation, called the upper path and the lower path respectively, which can be used to implement the transformation between the XS22 and the XS11. Further, the two paths can be described as the path (XS22, XS21, XS12, XS11) and the path (XS22, XS12, XS21, XS11) or can be simply represented as the paths PXS1 and PXS2, respectively. It is possibly show that the decision PXS1 of operation management of N0 is more effective than the decision PXS2. For the simplicity, we show only the existence of the paths PXS1 and PXS2, here. Theorem:For any telecommunication network (design) N0, let its design environment be E0, then there are the two above management decisions PXS1 and PXS2. Proof:Since the design and implementation of N0 is based on the environment E0, the N0 can be represented as N0=f(E0), where f is a design function of wide meaning and E0 is variable. Further, for any telecommunication network N0 and the operation status (N0, E0) of N0, we have that (N0, E0) is in {XS11, XS12, XS21, XS22}. Let XS1={XS11}, XS2={XS12, XS21, XS22}. Because the goal of the performance management of telecommunication network N0 is to make the (N0, E0) equal to the XS11 or (N0, E0) be in XS1, the decision M={M1, M2, ..., Mr} of the operation management of telecommunication network N0 is to make M*{(N0, E0) in XS2}={(N0, E0) in XS1} while the (N0, E0) is in XS2, where the operation * represents the action of management decision. Further, let the (N0,E0)=XS22, we can show that there are the Mi and Mj in them, so as to make Mi*{(n0, E0)=XS22}={(N0, E0)=XS21} or Mj*{(N0, E0)=XS22}={(N0, E0)=XS12}. Because {(N0, E0)=XS12} is not equal to {(N0, E0)=XS11}, {(N0,E0)=XS21} is not equal to {(N0, E0)=XS11}, and there is Mi and Mj for Mi*{(N0, E0)=XS21}={(N0, E0)= XS11} or Mj*{(N0, E0)=XS12}={(N0, E0)=XS11}, thus, there are the management decision PXS1 and PXS2. The above theorem has clearly described the basic method for operating telecommunication network. However, how to represent the status (N0, E0) of the network N0 and how to determine the states XS11, XS12, XS21, XS22 are two key problems to be solved for implementing the above management decision. In fact, we can find that the above decision is really a decision based on the management of graphics. Thus, the operation management of telecommunication network N0 is really converted into a series of the management of graphics, where the graphics describe the various operation status of N0. This view shows also that we must find out some special graphics relevant to the various operation states of telecommunication network N0 for describing these states. Thus, we will introduce a new effective theory, called the computer network clustering construction theory (CNCCT), for showing how to construct the various graphics relevant to the status, e.g., (N0, E0), XS11, XS12, XS21 and XS22. The CNCCT thinks that, for any given design environment E0 which includes usually the user requirement and the basic infrastructure of telecommunication network, etc, we can construct a clustering tree of the network as the final design of topology of the network, and any other non-tree topology of the network can be generated from this clustering tree. Obviously, while the E0 varies, the relevant clustering tree varies certainly, so that this clustering tree becomes really an effective tool for describing the variation of the operation environment of telecommunication network. Further, the operation management or performance management of telecommunication network is really the management of various clustering trees (Note, a tree is an interesting and flexible data structure). Thus, the complex problem of the operation management of telecommunication network can be converted into the graphic management of the above special clustering trees. In general, this graphic management about the clustering tree requires a lot of operation about PC ARC/INFO, so that the ability of PC ARC/INFO to analyze the spatial information can be directly used to implement the real time management of operation of telecommunication network. For the convenience of the following discussion, we describe further a telecommunication network N0 as TN0, where TN0 represents a typical clustering tree about N0. In the same way, the operation environment of telecommunication network N0 is described as TE0, where TE0 represents also a clustering tree about E0. In fact, it is reasonable that the clustering tree is used for describing the implementation environment of N0, because we can use the interaction frequency among all users of N0 for constructing a clustering tree based on the frequency. Obviously, in the early stage of the operation of telecommunication network, the N0 is suitable for the E0, or TN0=TE0. However, we can use the same method as the above one of defining the status of telecommunication network for defining the status of the clustering tree as TXS1={TXS11} and TXS2={TXS12, TXS21, TXS22}. Further, we can also use the paths TPXS1 and TPXS2 for managing the telecommunication network N0. In the above description, we note that TN0 and TE0 are variable, so that the TN0 and TE0 can be represented as the following form: TN0={TN00, TN01, TN02, ...} TE0={TE00, TE01, TE02, ...} Where all elements in the TN0 and TE0 are the clustering trees. Thus, the goal of selection for the operation decisions is to make {TN0=TE0}={TN00=TE00, TN01=TE01, ...}, and the (TN0, TE0) is in TXS1 and the (TN0, TE0) is not in TXS2. 3.PC ARC/INFO IMPLEMENTATION OF GRAPHIC MODEL In the above discussion and analysis, the real time management of operation of telecommunication network is really transformed to the management of some special clustering trees. Thus, the complex problem of the real time management of telecommunication network is also transformed to the simpler problem of the graphic management in the graphic model, so that an uniform and global view is used to describe the problem of the operation management of telecommunication network with the help of a method of graphic operation. In order to implement the graphic model of the operation management of network in the environment of GIS(Bernhardsen, 1992), there are three problems to be solved, which are discussed as the following: 1)How to build a correspondent relationship between the abstract clustering tree and the infrastructure of small telecommunication network, for example, the relationship between the TN0 and the practical small telecommunication network; 2)How to manage the clustering tree with the help of the ability of PC ARC/INFO to analyze and process the spatial information; 3)How to apply the model based on the clustering tree for operating the small telecommunication network. The answer to the first problem is very simple. As the former discussion, for any telecommunication network, a relevant clustering tree can be constructed and be finally described on a graphic coverage about geographic information. With the database in PC ARC/INFO, the basic relationships between the nodes and arcs of the clustering tree and the infrastructure of small telecommunication network are easily built, so that the relationship between the TN0 and its practical small telecommunication network is also built. On the other hand, for the fault diagnosis and the performance optimization of small telecommunication network, this relationship is very important. For example, after the current clustering tree is compared with its standard clustering tree by a special algorithm, the abnormal situation of the current clustering tree, including the nodes and arcs, can be determined, so that the information about this situation can be used to determine the position of fault elements in small telecommunication network or to optimize the performance of the clustering tree as well as the small telecommunication network with the method of the performance optimization. However, the second problem is very complex, because it relates to the management of current clustering tree of small telecommunication network. This management includes two main contents, which are discussed as the following: 1)Determine the performance of current clustering tree or the abnormal situation in this clustering tree; 2)Recover the abnormal clustering tree to the clustering tree of the normal performance. There are two methods for determining the abnormal situation of clustering tree, which are called respectively the comparison method and the analysis method. The basic principle of comparison method is now discussed next. First, a base of the clustering trees for describing all normal states of operation of small telecommunication network must be built, i.e., each clustering tree in this base describes a normal status of operation of the network. Then, a special algorithm is used to operate the current clustering tree and the standard clustering tree at the same time for determining whether the current clustering tree is normal. On the other hand, the basic principle of analysis method is discussed as the following. In this method, each node and arc in the current clustering tree is checked for determining the abnormal node or arc so as to find out the abnormal tree. There are the various reasons to result in the abnormal clustering tree. However, they are usually divided into two classes, the hard reason and the soft reason. Further, the hard reason describes the fault of elements in the infrastructure of small telecommunication network. In general, this fault is physical, and can be not eliminated though changing the decision of operation of small telecommunication network. Otherwise, the soft reason is not physical, and can be eliminated though changing the operation decision. Further, our research shows that there are a number of the faults about the soft reason in the current and future ATM networks. In fact, the abnormal clustering trees generated from the two reasons are called respectively the hard abnormal clustering tree and the soft abnormal clustering tree. For the simplicity of this paper, some details of the abnormal clustering tree about the hard and soft reasons is not discussed here. However, this problem is easily solved by an overlay principle. In the operation of existing small telecommunication network, a number of problems are about the recovery of soft abnormal clustering tree. In general, the soft abnormal clustering tree relates close to the performance of the delay time and throughput of small telecommunication network. The method of transforming the soft abnormal clustering tree to the normal clustering tree is to adjust the distribution of information flow in the infrastructure of small telecommunication network. Further, this method is really to adjust the mode of resource use in small telecommunication network so that the abnormal clustering tree is transformed into the normal clustering tree. Because the status of information flow in small telecommunication network is fast variable, this adjustment has the obvious requirement for the performance of real time, i.e., the speed of adjustment must be very fast. On the other hand, the control action of this adjustment has obviously the character of delay. Otherwise, this adjustment is different from the other common control action, which is a pulse control of short time action. In the most situations, some standard recovery procedures can be constructed for recovering various soft abnormal clustering tree. Thus, a base of standard recovery procedures can be also constructed for recovering fast and simply the soft clustering tree. The management of clustering tree in the above discussion is really a feedback control method for managing the telecommunication network. However, a number of statistical analyses for the information flow in telecommunication network have shown that a feedforward control method for managing the clustering tree or small telecommunication network may be more satisfactory sometimes. The major principle of the feedforward control method can be simply discussed next. First, a number of the patterns of information flow in small telecommunication network can be built with the help of some statistical rules from a number of statistical analyses. Then, these patterns of information flow are used to design various effective control methods for managing the clustering tree of small telecommunication network, so that the real time control of the operation of small telecommunication network is effectively implemented. Otherwise, our research shows also that this feedforward control method is more suitable to future ATM networks. However, how to construct the statistic rule of information flow in small telecommunication network is a very complex problem. The key for solving this problem is how to recognize the patterns of information flow from collecting a lot of statistical data. Here, we suggest an effective method, called the multi-dimensions method, for describing and constructing the patterns of information flow. This method thinks that a four-dimensional space, which consists of time, date, month and year, can be effectively used to describe the patterns of information flow in small telecommunication network. Thus, a typical pattern of the information flow can be represented as the IFM=g(T, D, M, Y). Obviously, the g may be a very complex function. However, in the many situations, the g can be described as a linear function of combination of T, D, M, and Y, so that we can use the IFM=g(T, D, M, Y) to describe the information flow in the node and arc of clustering tree. On the other hand, the spatial operation about the g(T, D, M, Y) can easily implemented in PC ARC/INFO. In general, the implementation of some key techniques of the above discussion is very easy. Especially, the special macro language of PC ArcInfo can be used to design various program modules for implementing various operations, e.g., generating, comparing and diagnosing the clustering tree. Fortunately, there is a special network function module in PC ARC/INFO, so that the implementation of the above operations will be very easier. 4.TYPICAL ANALYSIS OF GRAPHIC MODEL In order to discuss deep some typical processes of operation of the graphic model for the operation management of small telecommunication network, we will introduce two examples for showing how to optimize the performance of small telecommunication network and how to recognize its faults. Otherwise, we will also discuss the problem how to design the method for optimizing the performance of small telecommunication network. 1)The performance optimization of small telecommunication network:In fact, the performance optimization is a very important task for operating small telecommunication network with the help of the real time management system of network. The goal of this optimization includes the improvement of the real time performance and the throughput performance of the network. Further, suppose the normal clustering tree of the current small telecommunication network be TN0 and the practical clustering tree be TN0'. The average delay time of N0 about the TN0 is t(TN0). The value of computation of the average delay time of TN0' is t(TN0') by fast algorithm. On the other hand, the throughput of TN0 is h(TN0), and the throughput of TN0 is h(TN0). Further, if the difference between t(TN0) and t(TN0) and the difference between h(TN0) and h(TN0') are larger than the given values a0 and b0, respectively, then the optimization of the clustering tree TN0 must be executed. In general, the methods for implementing the performance optimization include usually the physical optimization and the logical optimization, which are discussed respectively as the following. (1)Physical optimization:First, the TN0' is divided into several sub- clustering trees so as to recognize the minimum sub-trees which result in the abnormal TN0'. Then, these sub-trees are improved by the redundant capacity of nodes and links of N0, so that these sub-trees recover the normal status. Thus, the process of the physical optimization has been implemented. (2)Logical optimization:First, as the same as the above discussion, the TN0' is divided into several sub-clustering trees so as to recognize the minimum abnormal sub-trees. Then, some tables of resource about the sub-trees, including the tables of resource of various links and nodes of N0, are adjusted to recover the abnormal sub-trees and to implement the logical optimization. However, after the optimization is completed, the observation of these past abnormal sub-trees must be continuous, so that the assigned physical and logical resource must be collected when the abnormal status of these sub-trees elapses. Otherwise, if the abnormal status of these sub-trees exists forever, the substantial optimization or fault diagnosis of these sub-trees must be begun(Zesheng, 1990). 2)Recognition of the faults of network:From the above analysis, we can see that the recognition of the network faults in N0 is really implemented by the management of the clustering tree. The process relevant to this recognition is simply given as the following. The occurrence of network faults results in the variation of distribution of information flow in the network. Thus, after a series of sub- clustering trees about the abnormal clustering tree TN0' are analyzed and recognized, we can determine the sub-clustering trees which result in the abnormal clustering tree. Then, a special relationship between these clustering trees and the physical elements of the network can be used to determine the practical position of network fault. There are two typical network faults which are considered in our graphic model, i.e., the hard fault and the soft fault. Further, the former fault is mainly relevant to the hardware fault of the network, the latter is mainly to the operation decision and the software fault of the network. Obviously, the performance optimization and the fault diagnosis all are relevant to the determination of the minimum abnormal sub-clustering trees which result in the abnormal status of TN0. In order to determine whether the abnormal sub-trees are relevant to the performance optimization or the fault diagnosis, an algorithm for identification of these sub-trees must be designed to implement this goal. 3)Operation Decision for Optimization As the above discussion, the decision for optimizing the operation of small telecommunication network is really the decision for the performance optimization of the management of small telecommunication network TN0, and this process can be easily implemented by PC ArcInfo. The various decision for optimizing the operation of small telecommunication network are provided by the pattern generator of the network operation and are also affected by the feedback control for the operation status of small telecommunication network. The small telecommunication network N0 is really controlled by the double mechanisms, including the feedforward and feedback controls. Our experience on a prototype based on PC ARC/INFO has shown that the performance of application of this control method is very satisfactory. Especially, this control method will have the wider areas of application in future ATM networks. The monitor module of N0 supports the interaction between the network and its manager. The operation of all the above modules is supported by the environment of PC ARC/INFO. A better plan of research is that the various patterns about the operation management of small telecommunication network are included in a knowledge base with the AI technique, and the clustering trees are managed by the knowledge base and AI techniques. 4)Current Results on our Experiment In the current time, the special GIS software-Esri's PC ARC/INFO - is being used to develop various key techniques of the above discussion and to construct a prototype of the real time management system for operating the small telecommunication network. Some practical and simulation experiments are carried out. The application of GIS and the graphic model for the management makes the task of the real time management of the operation of small telecommunication network simpler and easier. All the experience shows that this graphic model has the better performance than the other common model. Further, some obvious advantages of the graphic model are discussed as the following. (1)Increase of the real time performance of the operation and management of small telecommunication network; (2)Easier evaluation of the operation quality of the network with the uniform view; (3)Easier visual operation for the operation management of the network in GIS environment; (4)Easier implementation of the performance optimization and the fault diagnosis of the network; (5)Easier construction of the user interface for the management; (6)Easier application of AI technique in the real time management of the network operation. 5.CONCLUSION In this paper, we have discussed a method for building and implementing the real time management system for operating the small telecommunication network. The some keys of this method includes the construction of management model based on the clustering tree, the graphic modelling of the problem of the operation management of network and the implementation of the graphic model in the environment of GIS. Obviously, this clustering tree model provides an uniform view to describe the operation management of network, and a powerful tool for managing the network. On the other hand, the graphic interface and spatial operation ability of GIS provide an excellent support with the implementation of this model. However, there are many problems to be researched in the future, including the deeper analysis and research of the relationship between the clustering tree and its special action of the network. On the other development, a special module in PC ARC/INFO for operating the tree data structure or graphic must be developed in the future so as to make the fast operation of the graphic model. Otherwise, some excellent tools of GIS, e.g., Esri's ARC View.2, will make the construction of our system more reasonable and the implementation of it easier. The other analysis and testing work on our prototype is doing in the current time. 6.ACKNOWLEDGE Many colleagues provide a lot of help with our research. Especially, we would thank telecommunication engineering department and the agricultural information research center for many helps and warm research environment. 7.REFERENCES Cassel, L. N.,Partridge, C. and Westcott, J. 1989, "Network Management Architectures and Protocols: Problemsand Approaches", IEEE J. Select. Areas Commun., 7(7): 100-110. Ball, L.1992, Cost-Efficient Network Management, New York:McGraw-Hill. Tjaden, G., et al. 1991,"Integrated Network Management for Real-Time Operations",IEEE Network, 5(4): 10-15. Lewis, L. and Dreo, G, 1993, "Extending Trouble Ticket Systems to Fault Diagnostics", IEEE Network, 7(5):44-51. Samani, M. M. and Sloman, M. 1993, "Monitoring Distributed System", IEEE Network, 7(5):20-30. Snodgrass, R. 1988, "A Relational Approach to Monitoring Complex Systems", ACM Trans. on Computer Systems, 6(2):157-196. Bernhardsen, T. 1992, Geographic Information Systems,VIAKIT A.S./Norwegian Mapping Authority. Zesheng, Z. 1990, "On Improving and Optimizing the Performance of Existing Packet-Switching Networks", Proc. of International Telecommun. Symposium, Brazil:25.5.1-25.5.5.
Sun Ling
A. Reseacher, JiangSu Academy of Agricultural Sciences
JiangSu Academy of Agricultural Sciences
Nanjing, JiangSu, 210014, P. R. China
Telephone:86-025-4431481-325
Fax:86-025-4432691