Exploiting Logical Mobility in Mobile Computing Middleware

In the past few years, we have witnessed the growth in availability of mobile computing devices, such as mobile phones, personal digital assistants (PDAs), laptop computers and the like, combined with the advent of new wireless networking systems, such as 802.11b, Bluetooth, and GSM/GPRS. Users of such devices demand access to networked information and services at all times and locations, and, as such, encounter different networking structures and topologies. The introduction and use of mobile code techniques, from the acceptance of Java applets in web pages, to the creation of mobile agent platforms has demonstrated that it is possible to create dynamic systems that support code migration. We argue that logical mobility over physical mobility can offer the flexibility that traditional approaches cannot, helping to provide innovative solutions to mobility issues. We consider the following forms of mobile interactions, according to [1]: Client/Server interactions (CS), whereby the request of a client triggers the execution of a unit of code in a server and returns the results to the client, Remote Evaluation (REV), where a device can send code to another host, have it executed and retrieve the result, Code On Demand (COD), where a host can request a unit of code from another device to be retrieved and executed, and Mobile Agents (MA), where an agent is an autonomous unit of code that decides when and where to migrate. Moreover, we consider devices that can be nomadically connected to a fixed network (e.g., a laptop dialling up to an ISP), devices that are constantly connected to a fixed network over a wireless connection (e.g. a GPRSenabled mobile phone), devices that are connected to adhoc networks (e.g. Bluetooth piconets) and any combinations of the above. Related work on middleware supporting code mobility, geared for mobile environments include Sun Microsystems Jini[5], a technology based on Java which employs Code on Demand and Remote Evaluation techniques to deliver distributed services in high-speed and long lived networks. Jini allows clients to discover which services are currently offered and use them. Jini provides a centralised framework, which requires lookup services, functioning as indexes of services offered, to operate. Jini, and especially the JMatos mobile implementation, can be used by mobile devices to deliver context aware services to mobile devices, connected to a fixed network nomadically. Jini is not, on the other hand, particularly suitable for allowing mobile devices to offer services themselves, particularly in ad-hoc environments which lack a centralised lookup service. Lime[3] is another example, which provides application developers with a data-sharing middleware, geared for mobile agents and ad-hoc networks. With hosts acting as containers for mobile agents, Lime offers a Linda Tuplespace implementation for coordinating and sharing data, an implementation extended with location and mobility primitives. It does not currently provide any form of security and the fact that it only offers a flat tuple space as the only common data structure, limits the processing that can be made on the shared information. Moreover, the advantage of having mobile agents in physically mobile hosts is not made clear. Examples Logical over Physical Mobility