IISc Projects

IDEAS: Intelligent Digital Economy Algorithms Sponsoring Organization: Intel Technology India Private Limited, Brigade Champak, Infantry Road, Bangalore - 560 001 Mathematical modeling has a key role to play in the emerging digital economy. There is strong evidence to suggest that sophisticated mathematical modeling can impart significant efficiencies to electronic business processes in the digital economy such as e-procurement, e-selling, e-logistics, and e-supply chain in general. Our objective in this proposal is to create "visionary" intellectual property by developing superior algorithms for making e-business extremely efficient. The mission is to develop "breakthrough" e-biz algorithms which will form the "core" of "futuristic" e-business. The proposed research will be a highly cross-disciplinary effort involving areas such as combinatorial optimization, randomized algorithms, evolutionary computing, theory of auctions and negotiations, game theory, learning, and probability. Examples of such algorithms will include: A multi-attribute combinatorial auction algorithm for optimal selection of vendors in procurement of bundles of materials or services (A preliminary version of this was experimented with successfully in the PROMISE project) An intelligent decision analysis algorithm for selection of bids and product offerings in e-sourcing. Learning algorithms for modeling history and bidder behavior for optimal decision making in private marketplaces Demand-supply matching algorithms for complex products and services. Design of "sound" market mechanisms, that is, economic mechanisms with properties such as allocative efficiency, individual rationality, budget balance, and incentive compatibility (also called truth-revealing mechanisms), for deployment in e-business processes. Dynamic pricing algorithms for efficient yield management in e-supply chain processes. DESIGN AND ANALYSIS OF SIX SIGMA SUPPLY CHAINS @ GENERAL MOTORS University Prinicipal Investigators: Prof. Y. Narahari and Dr. N.R.S Raghavan General Motors Prinicpal Investigator: Dr. Jeffrey D. Tew, Enterprise Systems Lab, GM R & D Center, WARREN, MI 48090-9055 Concept Supply chain process is a complex, composite business process comprising a hierarchy of different levels of value-delivering business processes. Achieving outstanding delivery performance is the primary objective of any industry supply chain. In this research, we propose "six sigma supply chains" as a notion to quantify the level of delivery performance of a supply chain. We seek to evolve a set of "fundamental principles" based upon which we propose to develop a rigorous, systematic, and scalable methodology to design six sigma supply chains. The research is founded on connections between mechanical design tolerancing and supply chain management. The methodology is sought to be used in designing and reengineering General Motors supply chains so as to achieve high levels of delivery performance. Background A supply chain network can be viewed as a network of facilities in which a customer order will flow through internal business processes such as procurement, production, inventory management, and logistics, ultimately resulting in delivery of required products on time to customers. As one can imagine, when the number of resources, operations, and organizations increases, managing the supply chain can become very complex. An entire supply chain could exist within a single company or a supply chain can span multiple enterprises. An important design objective in such networks is to achieve a high probability of delivery of finished products to the customer in a customer specified delivery window. This entails perfect synchronization among supply chain elements and individual business processes embedded within the supply chain process. This in turn requires variability reduction all along the supply chain. Variability reduction is a central theme in mechanical design tolerancing where a complex assembly is built from numerous subassemblies and components. In this proposal, we first establish a connection between two different disciplines: mechanical design tolerancing and supply chain management. Next we explore the use of statistical tolerancing techniques and best practices, such as the theory of process capability indices, Motorola six sigma program, and Taguchi methods in designing what we call as "six sigma supply chains." The primary objective is to originate a rigorous, systematic, and scalable methodology for design of synchronized supply chains, with outstanding delivery performance. The methodology is sought to be used in designing and reengineering General Motors supply chains so as to achieve high levels of delivery performance. IMPROVE@GM: PROCUREMENT ALGORITHMS IMPROVE@GM: Intelligent Market for PROcurement from VEndors @ General Motors University Prinicipal Investigators: Prof. Y. Narahari and Dr. N.R.S Ragha van General Motors Prinicpal Investigator: Dr. Jeffrey D. Tew, Enterprise Syste ms Lab, GM R & D Center, WARREN, MI 48090-9055 Concept IMPROVE is an innovative, intelligent prototype e-marketplace to demonstrate the feasibility of effective use of innovative purchasing mechanisms and market algorithms to provide cost- effective procurement of direct materials at General Motors. The objective of this research is to develop intellectual property that can be used as the foundation for building highly efficient procurement marketplaces. The research will focus on issues such as procurement auctions and procurement algorithms and demonstrate the use of mathematical modeling in making e-procurement extremely efficient. Background Companies world-wide are continuously looking to improve supply chain efficiencies. A major way of obtaining these efficiencies is through the use of e-marketplaces. Despite slower economic growth in recent times, it is expected that e-marketplaces in general and private e-marketplaces in particular will provide tremendous value and efficiency to supply chain operations. A private e-marketplace is a one-to-many market mechanism which allows an enterprise to be connected to its own group of business partners using its application and network infrastructure to optimize various business processes. Private marketplaces provide high performance and tight integration with suppliers. Procurement (direct and indirect) is an extremely important business process that can benefit from private marketplaces. The objective of the current proposal is to architect an intelligent and innovative private marketplace for General Motors procurement. We call this IMPROVE@GM (Intelligent Market for PROcurement from VEndors at General Motors). PROMISE: PRIVATE E-MARKETPLACE PROMISE: Private Procurement Marketplace for Intel's E-Business and Services Duration: March 2001 - December 2001 Sponsoring Organization:Intel Technology India Private Limited, Brigade Champak, Infantry Road, Bangalore - 560 001 Scope of Work: PROMISE is Intel's Intelligent, private, proof of concept, private marketplace for procurement of indirect materials. PROMISE will have exactly one buyer (INTEL) and a multitude of sellers. PROMISE has two primary goals: Add value to INTEL's procurement by minimizing costs and using multiple attributes to drive procurement Attract a large number of sellers to participate in PROMISE The following are some desirable features of PROMISE: efficiency, ease of usage, ease of negotiations, demand aggregation, supplier aggregation, optimal vendor selection, robust design model, and extensible architecture. Activities: Specifically, our technical advice included the following activities: Mechanism Design A variety of business models are relevant for the procurement problem: reverse auction, sealed bid (single round and multiple round) auction, multi-issue auctions, and combinatorial auctions. Based on the specifics of the Intel procurement problem, we suggested a three pronged approach for carrying out e-procurement transactions: In respect of purchase requisitions for which preferred suppliers are already known, do a demand aggregation and send a purchase order to a standard vendor, taking into account volume discounts In respect of items for which volume discounts are not known but suppliers are known, do a bundling of purchase requisitions and send an RFQ (request for quotation) to the suppliers. From the bids received, select the final list of vendors. A two-attribute combinatorial auction algorithm was developed for optimal vendor selection. In respect of items for which not even the suppliers are known, conduct a dynamic, open-cry auction and determine winners according to standard protocols. Advice on Software Architecture The software architecture of PROMISE is required to be robust, scalable, and highly extensible. A five tier architecture was suggested for PROMISE: presentation layer, messaging layer, adapter layer, business objects layer, and persistence layer. Microsoft .Net technology was used in conjunction with Microsoft Commerce Server, Microsoft SQL server, and C## technology. The analysis and design of the tool was carried out, following best practices in object oriented software engineering. Extensive UML modeling was done and the models were reviewed. Several design patterns were used in the design of PROMISE. These include: Abstract Factory, Strategy, Chain of Responsibility, Composite, and Facade Accomplishments The project was completed in two phases. In the first phase, object oriented analysis and design were carried out and a working prototype was built. This took about four months. The prototype was reviewed thoroughly and in the second phase, a more comprehensive prototype was built. The project was successfully completed and the PROMISE prototype was demonstrated to several experts within and outside Intel. The following modules of PROMISE stand out for their innovativeness. The initial prototypes for some these had earlier been developed at IISc as part of earlier M.E. projects. Demand Aggregation Module Bundling module Two attribute combinatorial auction algorithm for selecting vendors Business rule engine Auction House EBIZCHEM: B2B PORTAL FOR TRADING OF CHEMICALS Sponsoring Organization: Ebizchem Private Limited, Norquest House, Ahmedabad, Gujarat Project Team Duration: April 1, 2000 - March 31, 2001 Scope of Work: Ebiz Chem is an Ahmedabad-based chemicals company, which had partially set up an E-Commerce portal that supports Internet-based buying/selling/decision making for typical products in the chemical industry. The potential user base of the site is intended to be anyone in the world buying or selling chemicals. The product range supported encompasses industries such as: dyestuff, pharmaceuticals, basic chemicals, pesticides, petrochemicals, and textiles. Ebiz Chem had already carried out the analysis and design of the client and server side software for setting up this vertical portal. Our technical advice to EbizChem covered the following issues: Evaluating the analysis and design of the software. This involves going through the analysis and design documents provided by Ebiz Chem and verify whether the designed system conforms to the original requirements set forth in the specification of the system. Carry out a quality control (QC) testing of the software from the perspectives of: load handling and scalability; data management; and security (*)Identify flaws in the implementation and communicate the test results Suggest short-term and long-term solutions for the technical flaws discovered Activities: Specifically, our technical advice included the following activities: Design Review: The software architecture and design of the portal was reviewed comprehensively. Our review covered all the tiers in the architecture: thick client web server (CGI scripts) database layer We found numerous gaps in the design, especially in the web server and the database layer. We prepared a detailed document describing the problems and suggested different ways in which to solve the problems. Advice on Software Architecture and Technologies The software architecture of the portal is required to be robust, scalable, and highly extensible. A three tier architecture had been frozen: presentation layer (thick client), web server and application server layer, and persistence layer. We carried out a detailed comparison of web server technologies, application servers, and databases. We finally recommended the use of a Linux web server with a weblogic application server and Oracle database. We also recommended that the database tables be thoroughly revamped and checked for normalization and optimality Load Testing of Web Server A suite of load testing software was developed for testing the scalability of Ebizchem's web site. The programs were developed in C and used several Linux system calls. Using the synthetic loads generated by the suite, the Ebizchem website was subjected to stress testing. Peak loads of about 100 requests per minute were found to crash the web site. This led to detailed analysis and diagnosis, after which several measures were suggested to improve the scalability of the web site. Accomplishments: The project was completed in three phases. In the first phase, the existing architecture of the website was studied in much detail and numerous design flaws were discovered. Concrete suggestions were provided to circumvent these problems. In Phase 2 of the project, we looked at the software architecture and found that the architecture was weak in terms of scalability and also database processing. Our principal findings were that: the web server is involved much too often in unnecessary processing seriously limiting its productivity; the CGI scripts developed are very unstructured and can be organized in a much superior way; the database tables are not even in first normal form and they need to be designed in a sound way, focusing on normalization and efficiency. These inputs were very useful to the Ebizchem design team. In Phase 3, we carried out detailed load testing of the Ebizchem website. This involved developing load generating programs and collecting statistics. The programs were developed based on state-of-the-art literature available in the area. An approach based on S-clients was used. The load testing experiments revealed that the web site had poor scalability. The experiments also revealed subtle reasons for the poor performance. Several short-term and long-term measures were suggested for improving the performance of the web site.

Intel Laboratory for Internet Technologies and E-Commerce Department of Computer Science and Automation Indian Institute of Science