[Japanese | Thesis | Researches in Minoh Lab | Minoh Lab]
Recently, some network systems for personal computers, AV devices and consumer appliances have been developed. When these appliances are connected to the home network, an appliance can make use of functions of other appliances over the network, and it extends their availability.
In the usual computer network, when we plug a computer to network, we have to assign an address and a name to the computer. Moreover, it's very troublesome to check whether the address is unique or not. To locate the troubles and cope with them, specialists are necessary. But when networks spread into homes widely, it will not be expected that users have expert knowledge. Therefore, appliances for the home network have to be configured automatically when a new appliance is joined to it or an appliance removed from it.
Most home network systems need special equipment as a server to manage information of appliances. However, in the case that the construction and maintenance of a network depend on the server, we have to prepare it. If the server has troubles or is removed from the network unexpectedly, the network will stop. It is inconvenient that a server is required to construct an ad hoc network outdoor. So, the network system that does not require a dedicated server is necessary for the home network.
In this paper, from a viewpoint that home network should be used without a dedicated server, we examine how to design a home network system that appliances can be attached/detached easily without any specialists. Especially, we consider the way that an appliance collects and stores information of other appliances if needed for appliance discovery service.
We classify management methods suitable for home network into models and compare them. Fixed Server Model has a dedicated server and the server collects and stores information of appliances. A client obtains needed information by asking the server. In this model, when the server has troubles or is removed from the network unexpectedly, we cannot use the network. Therefore, Double Server Model has two servers because we consider that the possibility of two appliances being removed at the same time is low. Even if the one server is removed, we can use the network continuously because the other server exists.
In Information Sharing Model, an appliance attached to the network sends own information to all other appliances periodically, and collects and stores information of all appliances. In this model, join time depends on the cycle of sending information. Therefore, in Represented Server Model, we provide a represented server to shorten join time and the represented server informs information of all appliances to new appliance.
By evaluating models based on time and traffic, we lead a rule to select a model for various networks according to environment or number of appliances which joining to.
We led the Double Server Model as the most suitable model for network in ordinary homes, and the Represented Server Model for ad hoc network constructed by portable appliances.
We can always construct the most suitable network for conditions by constructing network based on the led models. According to comparison with constructing the same network for various conditions, it's effective to construct network using the most suitable model for conditions.