【Summary】
In recent years, under the powerful driver of application needs, my country's communications industry has made great progress. Many enterprise units in the existing communications industry, such as telecommunications companies or mobile groups, one of the main features of their information systems is that the real-time monitoring requirements of the line are high, the amount of data is huge, how to integrate real-time control and information systems. Become a key part of the system implementation.
After participating in a network management system of a communication company and the design of the decision support system, we analyzed the integration and application of the two, deeply felt that there is a good design strategy and paying attention to the tools selected. . This project is mainly for real-time monitoring of subnet subnets and related links in communication, implementation alarms, routing control, and authorization, etc., the key is to provide a real-time map interface, and the well will summarize data Organization, establish a data warehouse and further implement data mining analysis, thereby supporting business decision analysis. As one of the designers, I focus on discussing the policies of the control system and information system integration herein.
【text】
As we all know, the communication industry needs a means of monitoring the communication network. Its work is characterized by the distribution of each sub-station and base station in geographic location L, more need to provide different sub-link links in higher levels. means. Generally speaking, since the data is massive, how to deal with the data obtained throughout the network system, so as to combine the analysis of the decision-making department, it has become an important topic that urgently needs to be resolved. Briefly, distribution, real-time, and data massiveness are the core issues of the entire system design and integration.
First, let's discuss "NMS monitoring system". Since this system I have participated in design and development is not located in the basement of the base, it is positioned in the host communication data of each sub-station (including data traffic, link load, communication between the host, etc.) Collect, so the underlying operation requirements for specific communication transactions are not high.
Taking into account the above reasons, we use a geographic information system development platform MapInfo and use Delphi programming, with SQL Server database in the background (this is because of Microsoft's OLAP Service). Before analyzing and planning, we have made detailed discussions on ITU801 standards. This is just a standard for subnet and link definitions and layers. After listening to many sub-staff recommendations, MapInfo provides company A related MAP X Active X control is embedded in the Delphi program, using the rich classes and operations provided in Map X, such as Object, Layer, etc., well, and join the concept of subnets and links, The sub-station under the subsection becomes a different subnet, and the enlargement and narrowing function can be enrolled in a map of the entire region in one map, and can be displayed on the screen. At this time, it is only displayed Summary of each sub-station, small to the cabinet of a host, the cabinet is until the plugin board (because these must be monitored in real time). We use layered methods to achieve the above zoom. For some static data, such as sub-station, the position of the host, first use MapInfo's set of geographic information (MAP X is an Active X control provided to programming tools) to place a static hierarchical diagram. In the database.
Our new system is connected to the dedicated lines of each sub-station, can be used in real time, displayed on the map, reflecting the real-time state of each station, each subnet, and each link, and can control Command back the sub-station (such as forced link interrupt, routing conversion, etc.).
Now let's discuss the most critical issues, that is to integrate real-time control systems with enterprise information systems, our ideas and architecture can be generally used in a simple map (here.). In this architecture, detailed data is retained by each sub-station. The network management system will make a sum of data to the data in a certain time interval, and the contents that need to be saved are put into the database, such as traffic, one minute Station and other sub-station communication traffic, load a link in this sub-station (these links may be dynamically assigned, or may be a communication link between fixed slits). Despite this, the data is still massive, so if you want to send these data directly to each decision department, it is unrealistic to give the market sector. Therefore, we have established a data warehouse on the basis of the database, determine the topics of several data warehouses such as customers, time, traffic, billing, and faults, and prepare the original data in the database every other time. Handling work, establish a good data warehouse. The pretreatment here includes a number of aspects, such as data to establish computational time, but no data (accounting value zero), should be considered to establish failure invalid data, need to be rejected; some companies rent It is a special line based on monthly, and the intermediate communications therefore there is no some relevant records that should be excluded.
After pretreatment, the data fusion and summary will be used to use the OLAP Service. According to the needs of the decision-making department, provide corresponding data (such as: market sector requires every sub-station earning, customer distribution, customer cost, etc.). These can be pre-processed by the OLAP Service, and the processing over the process is logically existing in the form of cubes (Cube), and the storage space occupied is significantly reduced, such as 20 million in August 1999. Bar communication record, even if the text as a backup requires 4G space, only 200M left and right space is processed after the OLAP Service, therefore, the processed data is mainly stored in the machine of another relevant department, and cannot be placed with the primary server Together.
Finally, the system used by decision makers will be discussed. Since these departments are not scattered, we do not use OLAP Servce's Web Publishing Plan. Using Delphi, client software that accesses OLAP Service is used, and the Cube Browser control provided by OLAP Service is used, and various operations of data cubes are provided with interface-like interfaces, such as drilling (observation angles) and even the year ), Slice, rotation, etc. In order to facilitate the output print data, it also embeds Microsoft's Excel PivotTable, which can be converted to Excel's form, or converted into pie charts, column charts and graphs, etc. For example, you can observe the distribution graph of 24 hours a day, you can find that there is a reduction in communication flow after 12 o'clock in the night, and the decision-making department can develop certain offers or reduction measures to attract more customers to use the network after 12 o'clock.
In addition, when the data mining function in OLAP Service, the two types of algorithms provided are based on the classification of decision trees and the clustering of decision trees, and the clustering algorithm of the market sector will aggregate according to the cost of fees, and discovery The common characteristics of customers in the same consumer level, facilitating the development of policies, attracting customers. This effort will continue to continue to ensure sufficient massive data and discover the laws.
After running throughout the system, a series of work such as data acquisition and data processing is periodically automated, and the system application has been received for a lot of praise. Of course, there are also many problems, such as the capacity problem of the primary server database, the communication efficiency of the primary station and the sub-station, and the determination of the network fault in the network management system is not meticulous, and it is necessary to Confirmation is determined that there is still a lack of direct communication methods between decision systems and network management systems, which have further resolved and improvements. Real-time control system and enterprise information system integration is an important driving force for enterprises and institutions engaged in manufacturing, measurement and monitoring, etc., and improve work efficiency. If it is a large company, there is a need to have a complete set of systems, support web publishing, intelligent query, automatic identification such as fault prediction and data mining, etc., so that the underlying real-time monitoring with high-level decisions are better integrated together. Looking forward to their prospects, it is undoubtedly very beautiful, but we believe that the corresponding workload is very large, and it is still necessary to improve and break through.
Comment: The topic of integration can be combined, combined with practical discussion. The database and data warehouse technologies discussed in line with the direction of enterprise informationization. Examples of the problems encountered are not enough, and the discussion of real-time control can also be more detailed. (This article mainly refers to the papers of Guangdong Lin Jiayi et al.)
