Partner Organizations:University of Missouri-Kansas City: Collaborative Research
Professor Vijay Kumar has collaborated on DAYS research. Specific contributions are discussed in the report details.
Activities and findings:
Research and Education Activities: Our objective in this project is to design a universal wireless information dissemination system, which can be accessed from any geographical location. Thus, a user (a company or an individual) from anywhere or at any time can pull the desired information from the space, process it, and push it to the desired place in the space for future use or to be shared. We call our system DAYS (DAta in Your Space). This investigation will promote an innovative use of wireless technology in managing information in the space. It will take the information processing from the confined environment to the open space, which will become a universally accessible reliable storage medium and a global information processing platform. Our data dissemination approach is unique. We view space around us as both a persistent storage medium and also as a data dissemination resource, while in physical terms it is actually a collection of wireless channels. We disseminate information on these channels, receive updates from users and install them in relevant databases. In the development of DAYS we do utilize the concept of broadcast disk and satellite radio system. DAYS also has full database and transaction management capability. For this reasons we have to provide necessary database services such as transaction serialization, database recovery, database access, database query, etc. Because it is constructed on wireless platform we also develop scheme for efficient channel management, data caching and indexing for faster access. The following list defines our major goals: 1. Design a broadcast server/mobile station coupled architecture which ensures a global broadcast coverage (as opposed to local). In this scheme DAYS can be accessed by any geographical location without the constraints of cellular architecture of PCS or GSM. 2. Development of a unique concurrency control technique which would allow DAYS to handle consistency preserving execution of user transactions. It will allow a wireless user the ability to perform a transaction by combining a read through the broadcast and write through a special mobile transaction. 3. DAYS will provide techniques to predict mobile user movement and to move a user's data to future predicted broadcast servers prior to his arrival in a broadcast cell. 4. In DAYS, a mobile user must be able to determine what local broadcast channel contains his desired data. We will investigate several approaches to satisfy this and recommend the best. 5. The contents of these broadcast channels will change depending upon the requirements of the users and also of DAYS. We will investigate efficient indexing schemes for managing the reorganization of broadcast contents and supporting finer granularity information access. 6. It may be necessary to duplicate or replicate broadcast contents to more than one dedicated channel. This may be decided on demand or for improving availability. We will investigate schemes for distribution of data in DAYS. This is collaborative research and the entire project is being investigated under a joint effort. However, specific tasks will be divided between University of Missouri-Kansas City (UMKC) and Southern Methodist University (SMU). We list below our progress in reaching the goals listed above. Note that the order of managing these tasks may differ with what we listed in the funded proposal. This kind of rearrangement becomes necessary for their efficient management. DAYS Architecture Validation: This task is crucial to all subsequent tasks, therefore this is being looked into by both groups (SMU and UMKC). In our proposal we presented a general architecture, which we are validating. Our validation goalis divided into subgoals and we have focused on the component that maps information categories into geographical locations, which forms a part of broadcast servers. The mapped information forms the broadcast (contents and format), which is pushed to the selected channel by the server. We have developed a mapping scheme, which identifies and select information content and maps to the relevant geographical location. In this mapping we have considered (a) correct association, (b) granularity match, (c) and termination condition. For example, weather could be associated with a country or a state or a city or a town. Thus, with a coarse granularity weather is associated with a country and in a finer granularity it is associated with a town. If a town is the finest granularity, then it defines the terminal condition for association for weather. We have also used a simple heuristic approach to develop a mapping mechanism. On DAYS architecture we have so far published two papers one of which was an invited paper. The original architectural design of DAYS was published in WECWIS 2001 and the second paper, which is on information mapping component of the architecture has been accepted for presentation and publication in the proceedings of MDDS 2003 workshop. The validation task is not complete and it will continue in parallel to other related tasks. UMKC is leading the task of information mapping. Concurrency Control Technique: Wehave previously proposed a concurrency control technique, BUC, which facilitates the update of data read from a broadcast. This approach is a combination of processing algorithms, concurrency control algorithms, and broadcast structure. In the past year we reported on the results of an analytic performance study which examined the overhead associated with BUC. Our study showed the superiority of the BUC technique by reducing the overhead requirements. Although the BUC approach and our DAYS architecture can be used independently, their use together provides a solid approach for the design of future wireless broadcast applications. Note that BUC in combination with DAYS provides an approach whereby all types of transactions from fixed and wireless users may be executed in parallel while still achieving serializability. In addition, through the use of a backchannel BUC allows users reading data through a broadcast the ability to perform consistent updates at content provider locations. SMU is leading this taskBroadcast Management: To manage the uninterrupted information access from channel by universally mobile users, broadcasts on channels may have to be dynamically reorganized. We have considered two approaches (a) static and (b) dynamic management. Static approach will serve the mobile local commuters and for international mobile users dynamic scheme will be necessary. UMKC is leading this task and so far we do not have any papers published. Channel Identification: UMKC is leading this task. It is important for a user to know what channel to tune to retrieve the desired information. We are looking into two ways (a) static and (b) location dependent. In the static approach there is 1:1 mapping between a channel and the broadcast contents. That is weather information will always be broadcast on frequency 100 for example and the user will be notified by a popup menu on his mobile device. This is similar to the channel allocation on normal radio broadcast. In the location dependent approach the information content may move to other channels or may be duplicated. Thus weather broadcast can be done on more than one channel covering the entire geographical locations of the world. The information also move to channel identified by a location. For example, weather for Kansas may be on channels tunable only from Kansas. This work is in progress and no paper has yet been submitted. Channel Indexing: When a client wishes to listen to a broadcast, he must know what channel to listen to and also from where on the channel the information starts. For example, if weather of multiple cities is on a single channel, then it is important for a user to know from where his weather information begins. This kind of identification will require complex indexing mechanism, which will be able to handle coarser and finer granularity. A simple approach was developed in our MDDS 2003 paper. We are investigating how to handle finer granularities of information. For example, weather for a country, weather for a state, weather for a city and so on. SMU is leading this task, but work is in the preliminary stages. Broadcast distribution: This problem resembles database distribution in classical distributed database systems. It is solved by defining (a) partition, (b) partial replication, and (c) full replication. IN DAYS this problem appears in a similar fashion but has to deal with data distribution on channels. This distribution affects data availability to mobile users, therefore, we are looking into user access behavior. For example, if stock quotes are hot data items, then it may be fully replicated to a set of channels. We do not have any papers on this task at present. Simulation: Our plan is to study the performance of individual schemes we develop and at the end the performance of the entire DAYS will be measured. Both institutions will be involved in this task. At UMKC we are in the process of investigating the performance of mapping scheme. SMU has already developed a simulation model used to examine the performance of BUC. We are currently writing a paper to be submitted for publication. Data Mining: Although data mining activities were not included in the original scope of the proposal, at SMU we have been investigating Web usage mining techniques targeted to expanding algorithms for frequent sequence and subtree mining. Web access is supported by the DAYS architecture. Due to the limited resources available on mobile devices, Web usage mining will be use to help determine cache and broadcast channel content. This year the PI's met two times. There we discussed our future plans and also exchanged useful ideas with other mobile researchers. Training and Development Activities: UMKC: Initially a Ph. D. student Nimisha Garg was working on this project. I have added one more Ph. D. student to start working on the project. SMU: Ahmad Al-Mogren successfully completed his PhD during the Fall 2002 semester. His research centered around initial design of DAYS and BUC. A new PhD student began working on DAYS in Spring 2003 semester. Web Sites: In addition to our publications, all our research findings and reports are available on our web sites, which are freely accessible. These sites are: http://www.engr.smu.edu/~mhd/mobile.htm http://www.cstp.umkc.edu/~kumar/idm01.htm
Findings: Major Findings: · Development of an information mapping scheme. · Verification of DAYS architecture and its components. · Development of concurrency control technique, BUC, for data broadcast. · Performance studies have shown the suitability of BUC to DAYS. · Developed caching algorithms and replacement strategies, so-called, FAR.
Journal Publications:
Other Specific Products:
http://www.engr.smu.edu/~mhd/mobile.htm
Contributions:
Contributions within Discipline:
We are the first to propose the universal use of wireless channels for universal data dissemination for information exchange. The completion of this project will offer a data dissemination framework where mobile or static users will not have to go through web for retrieving necessary information. Similarly organizations will not have to rely on web for broadcasting data for the use of consumers.
We are satisfied with our progress and we believe that our progress.
Special Requirements for Annual Project Report:
Unobligated funds: less than 20 percent of current funds