Akiko Nakaniwa

File allocation in distributed multimedia information networks

We study allocation problems of multimedia files in distributed network systems. In these systems, the files are shared by users connected with different servers through high-speed communication networks. One of the most important problems in distributed systems is how to assign the files to servers in view of costs and delays. In these systems, it is obvious that there is a trading-off relationship between costs and delays. Our objective is to find the optimal file allocation such that the total cost is minimized subject to the total delay. In order to evaluate the optimization, we present a system model that can cover a wide range of multimedia network applications such as VoD (video on demand), corporation information networks, and so on. We introduce a 0-1 integer programming formulation for the optimization problem. We find the optimal file allocation by solving these formulae, and quantify the general tendency in distributed systems. Moreover, we make a comparison between the exhaustive search and the approximate method that we use for optimization.

A Cost-Effective Dynamic Content Migration Method in CDNs

Content Distribution Networks (CDNs) are highly advanced architectures for networks on the Internet, providing low latency, scalability, fault tolerance, and load balancing. One of the most important issues to realize these advantages of CDNs is dynamic content allocation to deal with temporal load fluctuation, which provides mirroring of content files in order to distribute user accesses. Since user accesses for content files change over time, the content files need to be reallocated appropriately. In this paper, we propose a cost-effective content migration method called the Step-by-Step (SxS) Migration Algorithm for CDNs, which can dynamically relocate content files while reducing transmission cost. We show that our method maintains sufficient performance while reducing cost in comparison to the conventional shortest-path migration method. Furthermore, we present six life cycle models of content to consider realistic traffic patterns in our simulation experiments. Finally, we evaluate the effectiveness of our SxS Migration Algorithm for dynamic content reconfiguration across time.

Reliability-based optimal allocation of mirror servers for Internet

We consider the optimal mirror allocation problems for the purpose of load balance in network servers. We focus on constructing high-reliability networks, and propose the optimal mirror allocation model such that the system reliability is maximized subject to costs and delays, in view of the trading-off relationship between the reliability and cost. This optimization model is capable of dealing with various kinds of network topology. We formulate this optimization problem as a 0-1 integer programming in this model, and we can find the place where each mirror server should be and the concrete assignment of files to mirror servers. Our objective is to find the optimal mirror allocation by solving this model, and to show the general characteristics about the load balance and the improvement of the system reliability by the distributed mirror allocation, quantitatively.

Server and route selection method for QoS-based anycast protocol

The anycast communication protocol is to select the best server and the best route for a particular client from a group of replicated servers with the same contents. As users have come to demand contents with high QoS, the necessity to support the anycast communications in QoS networks has grown. In this paper, we propose a server and route selection method with application-level QoS-based anycast protocol. The protocol we propose has the following advantages: 1) the latest server and route information can be constantly acquired through distributed resource management by E-BB; 2) the server and route selection algorithm enable us to consider both the server load and network load simultaneously; and 3) high reliability can be guaranteed by decentralized control.