HKU Sparkle Project

Prepared by C.L. Wang

Introduction

Computing has been an ever-changing paradigm since the beginning of its creation. With the millennium, there is an advent of a new computing environment. Computing is no longer limited to a "computer" per se. You see more and more different types of devices, such as personal digital assistants (PDAs) and mobile phones, taking advantage of wireless networks to connect to the Internet to provide services to the user. A wealth of effort going into the development of intelligent appliances and information appliances. More people are using mobile devices to access information or perhaps even just to communicate with each other. There is, no doubt, a trend towards more and more networked small devices with wireless access present in living and working spaces. The future, thus, will see great emphasis on pervasive computing. Pervasive computing can be summarized by 3 A's - having access to computing and information Anywhere, Anytime and from Any device. The computing environment can be characterized by: · 

• Heterogeneity: Computing will be carried out on a wide spectrum of client devices, each with different configurations and functionalities. 
• Prevalence of "Small" Devices: Many devices will be small, not only in size but also in computing power, memory size, etc. 
• Limited Network Capabilities: Most of the devices would have some form of connection. However, even with the new networking standards such as GPRS, Bluetooth, 802.11x, etc., the bandwidth is still relatively limited compared to wired network technologies. Besides, the connections are usually unstable. 
• High Mobility: Users can carry devices from one place to another without stopping the services. 
• User-Oriented: Services would be tailored for the user rather than a specific device, or specific location. 
• Highly Dynamic Environment: An environment in which users and devices keep moving in and out of a volatile network.

The whole environment can be seen as a huge ad-hoc distributed system, with a multitude of small devices moving from one place to another and cooperating with each other. With this new environment, new approaches have to be used to build applications. Current approaches to build distributed applications have been found to be flawed in a pervasive environment. The pervasive computing environment poses new requirements on the infrastructure. These requirements are: 

• Adaptation to Diversity: The infrastructure should provide the ability for applications to adapt their functionality according to the device requirements, networks, etc. 
• Increasing Interaction with Peers: Many of these devices will form ad-hoc networks among themselves in order to exchange information and to co-ordinate in order to provide services to the user. 
• Flexible Computation Model: In a pervasive computing environment, there are various ways of accessing different types of data according to different users' needs. A combination of code and data mobility should thus be enabled to construct a flexible computation model.

Our project aims to build a user-oriented infrastructure designed specially with the needs of the future in mind. Our goals include: 

• Client-Dependent Adaptability: Such a changing environment necessitates applications to be dynamic - to dynamically change according to user's device configuration. 
• User and Device Mobility: Users should be able to continue their work independent of their location or the device they are using. 
• Peer-to-Peer Co-operative Computing: With increased interaction between peer users, direct communication links within user peer groups can be established to support computations without any central control.

To achieve the above goals, a combination of software development techniques and infrastructural entities have to be used. The traditional software architecture for application development has to be changed. Instead of being huge monolithic chunks, software should be made up of smaller components. In addition, the network should have some intelligence to enable adaptability and provide for user-oriented services. Also, mobile code systems have to be incorporated in the software architecture, which utilize a combination of data and code mobility in order to achieve a flexible computation model. 

An application executing with "facets".

The Sparkle System Overview

In our architecture, all applications are built from small functional components, called facets. A facet fulfills a certain functionality. It is stateless and possesses single invoking interface. It is associated with a manifest called shadow, which contains a description of its resource requirements, run-time behaviour, context conditions, etc. Facets run in containers, which is an abstraction that  provides an application-like feeling to end-users. Each container is associated with a particular application and contains a set of functionalities which the application can offer. The container, while running, would request for facets which fulfill the required functionalities. Figure 1 (above) shows an overall picture of the Sparkle architecture that operates on the facet abstraction. It consists of several entities, which are briefly described below:

• Client subsystem. Client devices provide a platform for the applications to run. Since applications are made of facet components, the client devices are responsible for retrieving and executing the constituent components. They discover nearby peers and proxy servers from which to request for components through its network/discovery module. They also use facet caches for performance improvement. In case a client does not have sufficient resources to run a facet, as a last resort, it can decide to delegate the execution to a peer or the computational server through the use of a lightweight mobile agent system (LMAS) incorporated in it.
• Facet servers. Facet server resembles the current web servers in the web architecture. It is a place for the service providers to put their designed facets on, publish and release them to the general public. Also, in cooperation with the proxy server, it will advertise the newly published facets hosted on the server by sending their associating shadows to the proxy.
• Intelligent proxies. Intelligent proxy in our architecture is both a broker and a middleman sitting between client device and facet server. It receives request from the client, and returns the corresponding facets by either fetching them from its own cache or from the facet server. While serving a query, it also takes into account details such as the client configuration, the execution environment, the user profile and preferences. Thus, it can return a facet that is best suited for the particular execution environment (adaptation) and the particular user (user customization).
• Computational servers. The computational servers, which are resource rich, are responsible to carry out executions for the clients just in case the client devices do not have enough resources.

Under normal cases, the client request is sent to the proxy. Since the proxy possesses the overall view: it knows what facets are available in the facet servers and it knows the requirements of the client, it can return the facet that best matches the client’s requirement. If connection to the proxy is not available, the client device can send requests to nearby peers to see if they have a facet with the required functionality available or even migrate the whole execution through the proposed mobility system: the LMAS.

Previous Research Students:

• Nalini Moti Belaramani (M.Phil, 2000-2002)  (Thesis)
• Yuk Chow,(M.Phil, 2000-2004) (Thesis)
• Vivien Wai-Man Kwan (M.Phil, 2000-2004)  (Thesis)
• Siu Po Lam (M.Phil, 2002-2004) : Context-aware State Management for Pervasive Computing (Thesis)
• Kong Choi Yu (M.Phil, 2002-2004) : Effective Partial Ontology Mapping in a Pervasive Computing Environment (Thesis Draft)
• Chen Weisong (M.Phil, 2002) : A Pervasive Information Framework Based on Semantic Routing and Cooperative Caching (Thesis Draft)
• Zhang Xiolei (Ph.D, 9/2003)

Nalini Moti Belaramani	Chow Yuk	Vivien Wai-Man Kwan
(Left to right: Laurel, Pauline, Weisong, and Tianqi)

Publication

• C.L. Wang , X.L. Zhang, N. Belaramani, P.L. Siu, Y. Chow, and F.C.M. Lau, Software Infrastructure for Context-aware Mobile Computing, to appear in Enabling Technologies for Wireless e-Business Applications, Springer. (abs)

• Francis C.M. Lau , Nalini Belaramani, Vivien W.M. Kwan, Pauline P.L. Siu, W.K. Wing, and C.L. Wang, ``Code-on-demand and code adaptation for mobile computing,'' to appear in Mobile Middleware, CRC Press, 2005. (abs)

• Wai-Kwong Wing, Francis Chi-Moon Lau, and Cho-Li Wang, "Smart Retrieval and Sharing of Information Resources based on Contexts of User-Information Relationships", The First International Workshop on Ubiquitous Smart Worlds, Tamkang University, Taipei, March 28-30, 2005.

• Pauline P. L. Siu, C. L. Wang, and F. C. M. Lau, ``Context-aware State Management for Ubiquitous Applications,''  International Conference on Embedded and Ubiquitous Computing (EUC-04), pp. 776-785, Aizu, Japan, 26-28 August 2004. (pdf)(PPT)

•  Laurel C. Y. Kong, C. L. Wang, and F. C. M. Lau, ``Ontology Mapping in Pervasive Computing Environment,''  International Conference on Embedded and Ubiquitous Computing (EUC-04), pp. 1014-1023, Aizu, Japan, 26-28 August 2004. (pdf)(PPT)

•  Weisong Chen, C. L. Wang, and F. C. M. Lau, ``A Collaborative and Semantic Data Management Framework for Ubiquitous Computing Environment,'' International Conference on Embedded and Ubiquitous Computing (EUC-04), pp. 962-971, Aizu, Japan, 26-28 August 2004. (pdf) (PPT)

•  Yuk Chow, Wenzhang Zhu, Cho-Li Wang, Francis Chi-Moon Lau, ``The State-On-Demand Execution for Adaptive Component-based Mobile Agent Systems,'' The Tenth International Conference on Parallel and Distributed Systems (ICPADS 2004), pp. 46-53, Newport Beach, California, July 7 - 9, 2004. (paper)(PPT)

• Vivien Wai-Man Kwan, Francis C.M. Lau, and Cho-Li Wang, "Functionality Adaptation: A Context-Aware Service Code Adaptation for Pervasive Computing Environments", The 2003 IEEE/WIC International Conference on Web Intelligence, pp. 358-364, Halifax, Canada, October 13-17, 2003. (pdf)

• Nalini Moti Belaramani, Yuk Chow, Vivien Wai-Man Kwan, Cho-Li Wang, and Francis C.M. Lau, `` A Component-based Software Architecture for Pervasive Computing,’’ Chap. 10, pp.201-222, Intelligent Virtual World: Technologies and Applications in Distributed Virtual Environments, chapter 10, pp. 191-212, World Scientific Publishing Co., Release: 07/31/2004.  (PDF)  

• Nalini Belaramani, Cho-Li Wang and Francis C.M. Lau, ``Dynamic Component Composition for Functionality Adaptation in Pervasive Environments,’’ The 9th International Workshop on Future Trends of Distributed Computing Systems (FTDCS2003), San Juan, Puerto Rico, May 28 to 30, 2003. (PPT)

This is an ongoing research project conducted by The Systems Research Group. The project is supported by a grant from the Research Grants Council (RGC) of the Hong Kong government (Project No. HKU7146/04E). The team, together with Shanghai Jiaotong University and Zhejiang University,  are working on a new 4-year project "Studies on Core Theories and Techniques for Pervasive Computing" funded by National Natural Science Foundation of China (NSFC), from 01/2006 to 12/2009. If you consider to join this research, please send e-mail to Dr. C.L. Wang (clwang@cs.hku.hk). You may also get more information about our recent progress on Sparkle system in the Sparkle-II project webpage here.