Mobile & Multi-tiered (M&M) - The Merger of Self Descriptive Objects, Mobile Agents, and Asynchrony

Sankar Virdhagriswaran, Crystaliz, Inc.

The current WWW and OMA has to evolve considerably in order to support cooperative problem solving systems that can span the world with many thousands of participants who autonomously decide to join and leave collaborative projects. The current WWW technology is best suited for a world wide information bureau (an original goal of the Web) and the current OMA is best suited for developing client/server systems that support connected, non-autonomous interactions, over high bandwidth networks with hundreds of participants. We at Crystaliz, Inc. have been developing mobile agent based systems that address the need for world wide, multi-site, multiple layer cooperative problem solving systems. In this position paper we describe our work and preview the key changes we hope to see in the OMA and WWW architectures in order to support multi-site, multiple layer cooperative problem solving systems.

Cooperative problem solving systems are characterized by several users modifying the same object while they participate in a project. The core problem when many users cooperate is consistency management. Consistency management involves replication, version management, configuration management, release management, workspace management, and cooperative transaction processing. To-date, cooperative problem solving systems for the most part have focused on version management for functional work groups spread over a LAN that perform one type of authoring activity (e.g., coding, ECAD, MCAD). However, the current WWW and the emerging wireless infrastructure enables a new kind of cooperative problem solving.

First, cross functional cooperation is not only enabled by the reach of the WWW (internet), but is also becoming common place as companies are empowering tiger teams to address productivity problems. These cross functional teams are often organized in a multi-tiered arrangement in order to allow players to work at different levels of abstraction. There are many tiers of abstraction between a car assembly team which works closely with suppliers to produce a car, and the dealer's sales team who is getting the customer preferences for the car from a customer. The WWW data dial tone allows these teams to be connected and supports multi-media communication, but a new technology is needed to support the multi-tiered arrangement of cooperating groups.

Second, while the WWW has become the multi-media data highway, its on-ramps, off-ramps, and some of the arteries are either clogged with traffic or have low through put. Furthermore, this problem is worsened when wireless networks are included in the picture. Wireless networks worsen the bandwidth problem and introduce new problems such as dis-connected operation, mobility of the hosts themselves, and ad-hoc workgroups. While existing one master - many slave approaches to data replication address bandwidth and disconnected operation problems, they do not address bandwidth and dis-connected problems that arise in the n to n to interactions found in a multi-tiered, multiple workgroup environments and do not address mobility of hosts at all.

Finally, in order to develop solutions to address the above problems, one needs to use an infrastructure that supports self descriptive objects, requests as objects, request/response asynchrony, and time asynchrony. If the objects can describe their characteristics consistently, then implementing consistency management becomes much easier. If the underlying distribution infrastructure allows applications to deal with request as objects instead of methods to an object, then the receivers can perform many optimizations including caching, dynamic source based routing, load balancing, look-ahead summarization, forwarding to another layer, context dependent operations, etc. Additionally, if the underlying infrastructure supports senders and receivers to deal with requests and responses as separate and independent entities, then optimizations such as pipelining, inter-leaving, etc. can be done on them. Finally, if the underlying infrastructure supports time asynchrony, then dealing with dis-connected operation is straight forward.

At Crystaliz, Inc. we are developing technologies that address the above problems. We would like to present our approaches to the above problems during the workshop. Sankar Virdhagriswaran Phone: (508) 287 4511 Crystaliz Inc. Fax: (508) 287 4512