Kazumine Matoba

Service oriented network architecture for scalable M2M and sensor network services

Network systems have evolved connecting various types of terminals and machines such as M2M system and sensor networks. On the other hand, the cloud data centers that have enormous processing capacity have been constructed by installing thousands of servers. New services will be created by connecting these resources. In this paper, three technical issues for M2M and cloud data center connected system will be shown: scalability, latency, and wasting system resources. We have developed service-oriented system architecture named “AAN” to solve these issues. The key components of AAN are add-on functions such as filter or cache functions on a messaging platform and a controller for them. Then, we also show our concept and technologies as a developed prototype system. Adaptive network cache achieved faster response time than without it.

Capacity dimensioning based on traffic measurement in the Internet

Capacity dimensioning in the Internet is becoming more important to provide stable and predictable quality of service (QoS) to the customers. However, capacity dimensioning is not an easy task because the network state changes frequently. We first propose a method to identify the bottleneck for improving the performance of the end users. We then focus on the measurement method for utilization of the bottleneck link between end-to-end hosts. To measure the utilization of the bottleneck link accurately, we propose a new measurement method that can eliminate several kinds of measurement errors. We collected measurement results from the actual operating Internet, and provide evidence to support the appropriateness of our approach. We finally propose a design framework in order to determine the adequate link capacity for offering a desired QoS level to end users.