The quality of large-volume streaming services can be assured only by effective network monitoring. Of particular importance, in multi-domain networks, monitoring must be conducted across domains and layers to fully grasp the details of overall network performance. However, when conducting monitoring, it takes too long to analyze the monitoring data because the monitoring system gathers massive sets of data from multiple locations. Furthermore, it is expensive to install the monitoring equipment in multiple locations. To realize faster analysis and lower cost, we propose adaptive monitoring system which troubleshoots service troubles automatically by creating and running monitoring scenarios; it can reconfigure the monitoring equipment and program dynamically at runtime. In this paper, we describe the function, design and implementation of this system. We adopt and extend existing technology in creating our system. By this extension, the network operator can manage the monitoring program on the monitoring sets placed in different domains in an integrated fashion. Furthermore, we discover and eliminate the bottleneck of a prototype system by measuring processing components in detail. The result is much faster overall processing. After showing the basic operations of the system, we confirm its effectiveness by conducting demonstration experiments on the high-speed test-bed network of Japan.
展开▼