Beyond device mobility

ABSTRACT

Users, application services, single files and files collections, computers, detachable memory, distributed network storage -- this is far from being the comprehensive list of thinkable today networked objects. More importantly, all of these objects can be mobile on their own or in any combination and independently of each other. The technology around information distribution, sharing, and retrieval drives towards new businesses, use cases and life styles whereas mobility plays an ever increasing role. Does current Internet architecture ready for that? Doubtfully not! This talk aims at discussing challenges around highly distributed and highly dynamic networking environment demanding generic support for mobility and session management.

Today's TCP/IP protocol suite offers quite limited functionality of managing sessions between applications. Basically it allows two operations: START a session and TERMINATE it. If applications require more sophisticated session control, such as SUSPEND, MOVE, RESUME, etc, then this is solved by each application in an application specific way. Obviously, it is very difficult to provide fully generic support for arbitrary session management and control desired by applications. On the other hand, there are quite many common functions, which can be delivered by communication substrate and used by many applications. Suspend / Resume communication model is one providing further steps towards generic support of mobility and session management.

By introducing the above model to communication it will be possible to approach a solution providing support for different types of mobile objects and at the same time broaden the scope of the notion of mobility by extending it onto non-instantaneous use cases caused by outages or deliberate user/system activities.

With regard to the above, there are a number of research issues which have to be solved in order to introduce session management support into the TCP/IP protocol suite.

Naming and addressing. This is one of the core issues, which affects all aspects of the design of generic session management and mobility support. The desire to address variety of different types of objects and support their continuous communications leads to the problem of creation of a new name space(s) providing correspondent name resolution mechanisms supporting correct, fast, and consistent updates of dynamic bindings between persistent names and changeable network dependent identifiers.

Session state management. This issue arises in conjunction with the introduced SUSPEND / RESUME communication model. The major challenge is how to preserve a communication and an application states across suspend - resume operations. The state should be persistent, i.e. entirely independent of underling protocol parameters, e.g. timeouts.

Dynamic bindings. This issue addresses, support for CHANGE operation across SUSPEND / RESUME functions. The subject for changes can be basically anything beginning from IP addresses, port numbers and going up to the mappings of file names to the directory structure and to device names were the content resides at the moment.

Use of cross-layer interaction. Each layer in the protocol stack contains quite much information, which can be useful for other layers and thus should be made available through a common mechanism of reaching this information base. This will provide support for various kinds of optimizations, which can be introduced in a simple way by knowing the entire communication state across all layers involved in the communication.

Interaction with applications. Disclosing partial information about communication state, accumulated inside the protocol stack, to applications will provide necessary support for adequate applications state management by applications themselves. This gives two obvious benefits: first, is that an application state can be preserved across any "undesirable" communication interruptions, and second, applications may delegate their state management to a communication substrate. Any of these approaches will benefit considering the possibility of longer communication disruptions.

Protocol support for session management. As changes in communication - mobility, disruptions, move of content and or applications, etc. taking place at one end, there is a need for synchronization with another communicating peer, there is a need for protocol providing consistent and secure updates for synchronizing affected bindings and preserving states during session SUSPEND time.

• Networks