一种被用于网络性能管理的模型及实现

高速通信网络（例如ＡＴＭ）技术能够架构一种灵活的支持宽带的多业务网络．诚然， ＣＯＲＢＡ（Ｃｏｍ ｍ ｏｎＯｂｊｅｃｔ Ｒｅｑｕｅｓｔ Ｂｒｏｋｅｒ Ａｒｃｈｉｔｅｃｔｕｒｅ）、移动计算（Ｔｒａｎｓｐｏｒｔａｂｌｅ ｃｏｍ ｐｕｔａｔｉｏｎ）、虚拟机（Ｖｉｒｔｕａｌｍ ａｃｈｉｎｅｓ）和积极网络管理技术的方法实现的网管系统具有灵活、健壮和容易修改的特征．但是，另一方面，它需要稳定、可靠和有效地提交ＱｏＳ（Ｑｕａｌｉｔｙ ｏｆＳｅｒｖｉｃｅ），而且其仅仅依靠网络协议本身是做不到的．为此，提出了一种用于网络性能管理的模型，该模型具有以下特点：第一，利用随机回应模型（Ｓｔｏｃｈａｓｔｉｃ Ｒｅｗ ａｒｄ Ｍｏｄｅｌ，ＳＲＭ）建立整体网络性能和可信度的模型；第二，能够实时地计算和预测整体网络端到端的性能可靠性（Ｐｅｒｆｏｒｍ ａｎｃｅ－ｒｅｌｉａｂｉｌｉｔｙ）、可用性（Ｐｅｒｆｏｒ－ｍ ａｎｃｅ－ａｖａｉｌａｂｉｌｉｔｙ）、该网络的平均无故障时间（ＭＴＴＦ）和可信度的值；第三，由于模型求真解十分复杂，对于高可靠性网络的管理，得到了该模型的近似模型及其它的数值解法，并在文章的最后，给出了一个例子．

A PERFORMANCE MANAGEMENT MODEL AND IMPLEMENTATION USED FOR NETWORK MANAGEMENT

High speed telecommunication network's (e.g. asynchronous transfer mode) technology promises a flexible, multiservice network that will support the broadband future. It will need to be robust and efficient in delivering quality of service, which will not be achieved by network protocols alone. A network performance management will be needed. If future networks are to be flexible allowing new services to be added easily, then their management systems will need to be equally flexible. Approaches derived from the developing technologies of distributed object oriented computing (common object request broker architecture, CORBA), transportable computation, virtual machine and proactive network management technology could allow management systems to be implemented as sets of small grain, cooperating distributed objects. Such systems should be more flexible, more robust, and easier to modify. On the other side, in order to provide efficient resource utilization and QoS guarantee in high speed networks, the performance and dependability analysis model used for performance management with three primary goals is proposed in this paper. First, use Stochastic Reward Models (SRM) to evaluate combined performance and dependability of network. Second, compute and forecast real timely end to end performance reliability (e.g. PR s0 (t) ), performance availability(e.g. PA s0 (t) ), mean time to failure (MTTF) and dependability(e.g. DT s T] ) for the overall network. Third, due to the arising complexity of exact SRMs, a complexity reduction approach is given. For highly reliable network's management, an approximation of model, its numerical approach and implementation are also presented in this paper.