Middleware enhancements for metropolitan area wireless Internet access
Introduction
Wireless data networks are rapidly evolving and are becoming pervasive, both in the local area and the metropolitan area domains. Commercial access to the latter is (currently or imminently) available in many forms, ranging from data-over-cellular to cellular digital packet data (CDPD) and radio-based systems (GPRS) and UMTS. This growth and evolution of technologies is simultaneously accompanied by new applications and increased adoption by end-users. Almost all of this usage is on mobile devices ranging from handhelds to notebook computers, and most applications fall into the category of wireless Internet software.
There are many important factors that may contribute to the success of wireless Internet applications and their use; including: device capabilities, networking support and middleware technologies. However, one major factor is support for emerging mobile and wireless applications. We believe that applications such as mobile commerce (m-commerce) will drive the widespread adoption of wireless Internet for both business and individual users. Mobile commerce applications include: mobile financial applications, user and location-specific mobile advertising, mobile inventory management, mobile proactive service and others. The components of a typical mobile commerce environment are shown in Fig. 1.
As can be seen in Fig. 1, mobile commerce encompasses applications that vary greatly in nature and criticality. To highlight the similarities and differences between them, the characteristics of several typical mobile commerce applications and the infrastructure support they require are shown below [1]:
- •
Mobile financial applications. Applications where mobile device becomes a powerful financial medium, e.g. banking, brokerage and payments for mobile users. Infrastructure requirements: location management, network dependability and roaming.
- •
Mobile advertising. Applications turning the wireless infrastructure and devices into a powerful marketing medium, i.e., user specific and location sensitive advertisements. Infrastructure requirements: location management, multicast communication and roaming.
- •
Mobile inventory management. Product locating and shopping. Applications attempting to reduce the amount of inventory needed by managing in-house and inventory-on-the-move. Applications helping to find the location of products and services that are needed. Location tracking of goods, boxes, troops, and people. Finding the location of a new/used car of certain model, color and condition. Infrastructure requirements: location management, network dependability, multicast communication and roaming.
- •
Proactive service management. Applications attempting to provide user information on services they will need imminently. Infrastructure requirements: location management, network reliability and roaming.
- •
Wireless re-engineering. Applications that focus on improving the quality of business services using mobile devices and wireless infrastructure. Instant claim-payments by insurance companies. Infrastructure requirements: roaming across multiple networks, network reliability and location management.
- •
Mobile auction or reverse auction. Applications allowing users to buy or sell certain items using multicast support of wireless infrastructure. Airlines competing to buy a landing time slot during runway congestion (a proposed solution to the air-traffic congestion problem). Infrastructure requirements: multicast location management, network reliability and roaming across multiple networks.
- •
Mobile entertainment services and games. Applications providing the entertainment services to users on per event or subscription basis. Video-on-demand, audio-on-demand, and interactive games. Infrastructure requirements: Quality of Service (QoS), roaming and location management.
- •
Mobile office. Applications providing a constantly available office environment to mobile users. Working from traffic jams, airport, and conferences. Infrastructure requirements: roaming, location management and network dependability.
- •
Mobile distance education. Applications extending distance/virtually ubiquitous educational support for mobile users. Participating in a class using streaming audio and video. Infrastructure requirements: QoS, multicast communication, location management and roaming.
- •
Wireless data center. Applications supporting a large amount of stored data to be made available to mobile users for making educated decisions. Detailed information on one or more products can be downloaded by vendors. Infrastructure requirements: QoS, roaming across multiple networks and location management.
It is evident from the above that almost all mobile commerce applications require network reliability, QoS, and dependability as mandatory requirements. While these terms encompass many definitions, they are all impacted upon by the intermittent connectivity characteristic inherent in wireless networks. In this paper, we focus on issues relating to connectivity and network quality that are intrinsic to wireless networks. Due to a multitude of environmental, structural, electromagnetic and atmospheric factors (as distinct from network related factors like congestion and link failure), all wireless networks are susceptible to regions in the coverage area that possess subnormal characteristics, usually in the form of lowered quality or loss of connectivity. Traditional point-to-point wireless applications such as cellular voice can tolerate such situations, but data applications, especially those that require a formal notion of group membership can be adversely affected by trouble spots. For example, in mobile commerce, the following applications are likely to be seriously affected by intermittent connectivity:
- •
Financial applications (including auctions).
- •
Collaborative decision making.
- •
Inventory management/product location tracking.
The other m-commerce applications listed above will also be affected, but possibly to a lesser extent, i.e., the semantic correctness of these applications is unlikely to be compromised. In group applications, where a specific set of members participate, the absence of response from a user (possibly due to intermittent connectivity) may lead to problem scenarios (see Fig. 2).
In this example, four different outcomes are possible. The first is the typical reaction of higher layer protocols, namely, to evict such users and allow them to rejoin later. In the second and third cases, the group is forced to wait until the user transmits an input, the difference being whether the group is prepared to wait indefinitely or for a bounded period of time. This duration may be determined by previous values and/or using any other information on brief disconnectivity. In the final scenario, the group waits for different periods of time based on the type of user: passive listener, active listener, and active participant. For the first type, the user can be removed from the session and can rejoin later without affecting the overall state of the group. In the second case, the group may wait for some time, but ultimately the user can be removed without further consequence. For the third type of user, the group would wait longer as the input from such users can affect the overall state of the application.
To handle such situations, the state of the application could be frozen until a timeout occurs or the user is connected again. However, if the user response could not have affected the outcome of the group application, it is not clear that: (i) whether the application should indeed wait for a timeout and (ii) what durations are suitable. If a small value is chosen, this may lead to the eviction of essentially live mobile users from the group. Evicted users, may then attempt to rejoin the group, only to find that the state of the application has changed significantly (such as bidding in a mobile auction). Conversely, a higher timeout value may force connected users to wait for hosts that have failed outright, or have been permanently disconnected. This strategy will also lower the performance of the application (perhaps even compromising its overall usability). It is clear that a more dynamic approach is required. Thus, to address these issues, we are investigating the notion of middleware that not only distinguishes temporary or brief disconnectivity from outright failure, but also provides a level of intelligence in reasoning about the characteristics of specific trouble spots.
Furthermore, the group membership algorithms used by applications can adaptively learn from a ‘system server’ about typical and worst-case trouble spots. Then an informed decision can be made as to whether to evict the user, suspend all group activity or permit the application to continue.
The results and recommendations presented in the rest of this paper will also be useful in suggesting methods to reduce the impact of brief disconnectivity on applications. The analysis is performed in the context of a wireless urban infrastructure, i.e., in a metropolitan area wireless network. The choice of such an arena was motivated by the observation that many wireless Internet users are likely to be in metropolitan environments, and so we postulate that technology to eliminate (or at least reduce) the effects of trouble spots is most required here.
Section snippets
Background
The advent of metropolitan area wireless networks has introduced a new dimension into communication support for distributed applications. Both commercial and academic infrastructures like SprintCDMA [2] and GUIDE [3] are becoming increasingly prevalent, but such networks are inherently subject to location dependent variability in terms of connectivity, latency, and bandwidth. As noted above, application domains such as mobile commerce may not be able to tolerate the coverage and quality
Motivations
Our original motivation for this project was derived from the collaborative computing frameworks (CCF) project. CCF is a suite of software systems that constructs a virtual work environment on Internet connected computer systems and supports application and data sharing, steering, audio conferencing, and other collaboration modalities. CCF may be viewed as high-level middleware for e-commerce applications. CCF is constructed on a group communications protocol layer, termed CCTL, that supports
Failure resilience in metropolitan wireless networks
We postulate that systems such as Ricochet are generally used for point-to-point communications (e.g. wireless email and web access). Emerging mobile commerce and other mobile applications, however, are likely to utilize multiway communications that offer multiple qualities of service. In these situations, simple recovery schemes that work well for two-way communication may not suffice. In order to support group communication in wireless environments, issues relevant in such contexts must be
Wireless service evaluation metrics
One outstanding issue is the pertinent question of how to ascertain the semantics relating to a specific trouble spot. When encountered, a trouble spot is initially detected by the occurrence of fault reports through the use of reliable multicast. At this stage, the system must determine (by executing a series of diagnostic tests) which category of trouble spot it has encountered so that it may invoke an appropriate failure handler. In order to achieve this, we propose the following metrics for
Metric evaluation
Performance evaluation of the proposed metrics is presented in relation to trouble spots of the third category, i.e., long disconnections. The experimental evaluation environment consists of a single laptop connected to the Internet through a Ricochet modem. The laptop is moved along an urban road for 1 mile at a speed of approximately 35 mph with the direction of travel alternating between north and south for successive experiments. The motivation behind this selection of environment was to
Response time
The results from the response time tests are shown in Fig. 4 and display a number of interesting features. Given that the environment is particularly noisy, the graphs clearly show two peaks (trouble spots) which occur at the same geographic locations. Both of these peaks are clearly in excess of the nominal response time (i.e., <1 s).
We postulate that the amplitude difference between the graphs can possibly be accounted for by the radio’s propagation model. As the frequency used by the modems
Quantification, categorization and transposition
Using the values suggested above, we tentatively refine the criteria for trouble spot categorization shown in Table 3 to that shown in Table 4.
Given that both trouble spots exhibit response times greater than 2 s, packet loss that reaches 100% and occur in areas with poor signal quality, Table 4 shows that both trouble spots are disconnections. Since the size of the trouble spots is not in the order of meters, they are not of the second category. Also, since the host recovers, they are not fatal
Conclusions
In this paper, we have presented a novel approach for detecting and adapting to network trouble spots in wireless networks. We believe that such schemes are of value to group communication protocols and facilitate adaptation by end applications, particularly those in the domain of mobile commerce. In m-commerce applications based on systems such as CCTL, the QoS can be dynamically changed when trouble spots are encountered, thereby compensating for network degradation at the (temporary) cost of
J.S. Pascoe is a final year PhD student at The University of Reading, UK. He received a first class BSc from Reading in 1999 and began a PhD studying group communications later that year. Pascoe has authored numerous papers in international conferences, produced several journal papers and is serving as a program committee member for a number of conferences. He has also acted as an industrial consultant and is active in the distributed systems community.
References (16)
- U. Varshney, R. Vetter, Mobile commerce: framework, applications and networking support, Special issue on Mobile...
- ...
- K. Cheverst, N. Davies, et al., The role of connectivity in supporting context-sensitive applications, in: H.W....
- Ricochet Technology Overview, Metricom, Inc., 1999....
- I. Rhee, S. Cheung, P. Hutto, A. Krantz, V. Sunderam, Group communication support for distributed collaboration...
- J.S. Pascoe, R.J. Loader, V.S. Sunderam, An election based approach to fault-tolerant group membership in collaborative...
- R.J. Loader, J.S. Pascoe, V.S. Sunderam, Introducing fault-tolerant group membership into the collaborative computing...
- E.A. Hall, The Core Internet Protocols: The Definitive Guide, O’Reilly & Associates,...
Cited by (16)
A review for mobile commerce research and applications
2007, Decision Support SystemsWhat drives mobile commerce? An empirical evaluation of the revised technology acceptance model
2005, Information and ManagementThe Influence of Characteristics of Mobile Live Commerce on Purchase Intention
2023, Sustainability (Switzerland)Perceptions of smartphone users' acceptance and adoption of mobile commerce (MC): The case of Jordan
2020, Journal of Computer ScienceSWOT analysis of M-commerce
2017, Mobile Commerce: Concepts, Methodologies, Tools, and Applications
J.S. Pascoe is a final year PhD student at The University of Reading, UK. He received a first class BSc from Reading in 1999 and began a PhD studying group communications later that year. Pascoe has authored numerous papers in international conferences, produced several journal papers and is serving as a program committee member for a number of conferences. He has also acted as an industrial consultant and is active in the distributed systems community.
V.S. Sunderam is a professor of Computer Science at Emory University. His research interests are in parallel and distributed processing systems, and infrastructures for collaborative computing. His prior and current research efforts have focused on system architectures and implementations for heterogeneous metacomputing, including the PVM system and several other frameworks such as IceT, CCF, and Harness. Professor Sunderam teaches computer science at the beginning, advanced, and graduate levels, and advises graduate theses in the area of computer systems.
U. Varshney is on the faculty of CIS at Georgia State University, Atlanta. He received his BE in electrical engineering in 1988 from the University of Roorkee, now Indian Institute of Technology-Roorkee (IIT-R). Then he pursued graduate work in electrical engineering at IIT-Bombay before receiving MS in computer science in 1992 and a PhD in telecommunications and computer networking in 1995, both from University of Missouri-Kansas City.
He has authored over 50 papers in dependable wireless networks, mobile commerce, wireless multicast, wireless ATM and several other topics in major journals and international conferences. He is credited with some of the early research in mobile commerce.
He has presented some very well received tutorials and workshops at major international conferences (ACM Mobicom, IEEE WCNC, IFIP HPN, HICSS to name just the few). He has delivered over 50 invited speeches, including several keynotes at conferences and workshops. He has also been involved with International Executive education programs at GSU. Since 1999, Upkar has received the highest teaching evaluations in 30+ member strong CIS department. In October 2000, Upkar was awarded the prestigious Myron T. Greene Outstanding Teaching Award.
He was a guest editor for ACM/Kluwer Journal on Mobile Networks and Applications (MONET)’s special issue on Mobile Commerce (with Ron Vetter). He is also on the editorial board of IEEE Computer. He has also served on the program committees of IEEE WCNC, IEEE LCN, ACM Workshop on Mobile Commerce, HICSS and several other international conferences.
R.J. Loader has just retired from being a Senior Lecturer in Computer Science at Reading University in order to pursue a more active research career. His research interests began in the early days of local area networks with the construction of fast network interfaces which was followed by work on protocol designs for reliable multicasting for cluster computing and protocols that adapt dynamically to changes in the network state. The current focus is in the design of protocols to support secure and fault tolerant collaborative systems using CCF as the exemplar. Over the years Dr Loader has taught just about every course in System Architecture at undergraduate and postgraduate level. He has also acted as a graduate supervisor.
- 1
This project was supported in part by NSF grant ACI-9872167 and DoE grant DE-FG02-99ER25379.
- 2
Supported, in part, by an RCB research grant at Georgia State University.