Properly designed handover procedures in cellular wireless systems are essential for maintaining ... more Properly designed handover procedures in cellular wireless systems are essential for maintaining continuity of a call in progress and for minimizing the probability of forced call termination and signaling and switching load on the network. We model a channel exchange scheme that involves exchange of channels between two mobiles that are moving in opposite directions across the handover area of adjacent cells. The channel exchange method is interesting since it yields low values of handover failure probability as compared to the case with no channel exchange. We compare the performance of the cellular system with and without channel exchange through simulation and analysis. For realistic values of new call arrival rate, it is found that the handover failure probability obtained from analysis is in close agreement with that obtained through simulation
We develop an approach for resource management for stream sessions based on measurements at the s... more We develop an approach for resource management for stream sessions based on measurements at the sources. The results in this paper assume that (i) the sessions are carried in an edge-to-edge virtual path (VP) or label switched path (LSP), that (ii) weighted fair queueing (WFQ) is used at each hop of the LSP, with one queue being used for all the traffic from this LSP, and that (iii) the stream sources are statistically equivalent; e.g., they could be packet voice sessions between two locations of an enterprise.
Most on-line (i.e., not stored) Variable Bit Rate sources would find it difficult to a priori dec... more Most on-line (i.e., not stored) Variable Bit Rate sources would find it difficult to a priori declare the traffic parameters required by a connection admission control strategy. There is thus the problem of measurement based on-line estimation of source parameters. In this paper we address the problem of selection of source parameters based on minimising a bufferbandwidth cost function in the network, for a specified delay QoS Violation Probability. We consider the shaping delay plus first hop multiplexing delay; this is adequate, for example, for n statistically identical packet voice sources being multiplexed at a PBX, or in approaches where the end-to-end delay bound is broken into per hop delay bounds. Our approach yields a leaky bucket rate parameter p' , and the sum of the shaper buffer and leaky bucket depth (B, + a). We show that, for a fluid source model, for a linear buffer-bandwidth cost function, and for lossless multiplexing, a sustainable rate parameter of p* and burst parameter of 0 yields the minimum cost. We propose and study a stochastic approximation algorithm for on-line estimation of p*. We then use buffer-bandwidth cost considerations to arrive at an optimal leaky bucket depth 0' > 0 for lossy multiplexing of several statistically identical sources. The computation of o* must be done at the network node. We show, by an example, the improvement in cost that is possible by lossy multiplexing and a positive U*. ). determine a set of optimal leaky bucket parameters, and measurement based estimation of these parameters. network node access:'
Social networks refer to structures made of nodes that represent people or other entities embedde... more Social networks refer to structures made of nodes that represent people or other entities embedded in a social context, and whose edges represent interaction between entities. Typical examples of social networks are collaboration networks in a research community, networks arising out of interaction between colleagues of large organization etc. Social networks are highly dynamic objects that evolve quickly over time with addition and deletion of nodes and edges. Understanding the evolution of a social network is helpful in inferring trends and patterns of social contacts in a particular social context. In this paper, we consider social networks that are derived from telephone call records, i.e, graphs in which the individual phone numbers (and hence its users) are the nodes and the edges correspond to a telephonic contact between the two nodes they connect. We study the problem of extracting dense communities from such telecom call graphs. The problem studied here is set in the context of a larger project. We motivate the problem studied by describing the context in which it is set. Our analysis is based on suitable algorithmic engineering of an approximation algorithm for the densest subgraph problem by Charikar. We present empirical results on massive graphs with millions of nodes and edges. We also discuss many open problems that are important in the context of analyzing telecom call graphs.
Merchants often use marketing elements such as advertisements, coupons and product recommendation... more Merchants often use marketing elements such as advertisements, coupons and product recommendations, to attract customers and to convert visitors to buyers. We present a model for making a series of recommendations during a customer session. The model comprises of the customer’s probability of accepting a marketing element from a marketing spot and a reward for the marketing element. The probabilities can be estimated from customer history (such as traversals and purchases), while the reward values could be merchant specified. We propose several recommendation strategies for maximising the merchant’s reward and analyse their effectiveness. Our experiments indicate that strategies that are dynamic and consider multiple marketing spots simultaneously perform well.
In this paper, we introduce a new problem which we call as Online Probabilistic Weighted Bipartit... more In this paper, we introduce a new problem which we call as Online Probabilistic Weighted Bipartite Graph Matching. Consider a weighted bipartite graph with a source node set, a target node set, an edge set and a set of weights for the edges. The source nodes arrive in an online fashion based on a stochastic process. When a source node arrives, it needs to be matched with a target node before the arrival of next source node. Since the arrival process is stochastic, all the source nodes need not arrive and their order of arrival is also not known a priori. The objective is to match the arriving source node with a target node such that the expected sum of weights of the matching over the arrival process is maximized. We present some heuristics that perform well for this problem. We demonstrate the application of our formulation for session based recommendation [5]. Here the source nodes correspond to the web pages, the target nodes correspond to the advertisement that can be shown and the edge weights correspond to the revenue generated by showing the given advertisement on the given web page. The user traversal of web pages corresponds to the arrival process of the source nodes.
We present an architecture for and prototype of a system for quickly detecting software problem r... more We present an architecture for and prototype of a system for quickly detecting software problem recurrences. Re-discovery of the same problem is very common in many large software products and is a major cost component of product support. At run-time, when a problem occurs, the system collects the problem symptoms, including the program call-stack, and compares it against a database of symptoms to find the closest matches. The database is populated off-line using solved cases and indexed to allow for efficient matching. Thus problems that occur repeatedly can be easily and automatically resolved without requiring any human problem-solving expertise. We describe a prototype implementation of the system, including the matching algorithm, and present some experimental results demonstrating the value of automatically detecting re-occurrence of the same problem for a popular sofware product.
Re-occurrence of the same problem is very common in many large software products. By matching the... more Re-occurrence of the same problem is very common in many large software products. By matching the symptoms of a new problem to those in a database of known problems, automated diagnosis and even selfhealing for re-occurrences can be (partially) realized. This paper exploits function call stacks as highly structured symptoms of a certain class of problems, including crashes, hangs, and traps. We propose and evaluate algorithms for efficiently and accurately matching call stacks by a weighted metric of the similarity of their function names, after first removing redundant recursion and uninformative (poor discriminator) functions from those stacks. We also describe a new indexing scheme to speed queries to the repository of known problems, without compromising the quality of matches returned. Experiments conducted using call stacks from actual product problem reports demonstrate the improved accuracy (both precision and recall) resulting from our new stack-matching algorithms and removal of uninformative or redundant function names, as well as the performance and scalability improvements realized by indexing call stacks. We also discuss how call-stack matching can be used in both self-managing (or autonomic systems) and human "help desk" applications.
Properly designed handover procedures in cellular wireless systems are essential for maintaining ... more Properly designed handover procedures in cellular wireless systems are essential for maintaining continuity of a call in progress and for minimizing the probability of forced call termination and signaling and switching load on the network. We model a channel exchange scheme that involves exchange of channels between two mobiles that are moving in opposite directions across the handover area of adjacent cells. The channel exchange method is interesting since it yields low values of handover failure probability as compared to the case with no channel exchange. We compare the performance of the cellular system with and without channel exchange through simulation and analysis. For realistic values of new call arrival rate, it is found that the handover failure probability obtained from analysis is in close agreement with that obtained through simulation
We develop an approach for resource management for stream sessions based on measurements at the s... more We develop an approach for resource management for stream sessions based on measurements at the sources. The results in this paper assume that (i) the sessions are carried in an edge-to-edge virtual path (VP) or label switched path (LSP), that (ii) weighted fair queueing (WFQ) is used at each hop of the LSP, with one queue being used for all the traffic from this LSP, and that (iii) the stream sources are statistically equivalent; e.g., they could be packet voice sessions between two locations of an enterprise.
Most on-line (i.e., not stored) Variable Bit Rate sources would find it difficult to a priori dec... more Most on-line (i.e., not stored) Variable Bit Rate sources would find it difficult to a priori declare the traffic parameters required by a connection admission control strategy. There is thus the problem of measurement based on-line estimation of source parameters. In this paper we address the problem of selection of source parameters based on minimising a bufferbandwidth cost function in the network, for a specified delay QoS Violation Probability. We consider the shaping delay plus first hop multiplexing delay; this is adequate, for example, for n statistically identical packet voice sources being multiplexed at a PBX, or in approaches where the end-to-end delay bound is broken into per hop delay bounds. Our approach yields a leaky bucket rate parameter p' , and the sum of the shaper buffer and leaky bucket depth (B, + a). We show that, for a fluid source model, for a linear buffer-bandwidth cost function, and for lossless multiplexing, a sustainable rate parameter of p* and burst parameter of 0 yields the minimum cost. We propose and study a stochastic approximation algorithm for on-line estimation of p*. We then use buffer-bandwidth cost considerations to arrive at an optimal leaky bucket depth 0' > 0 for lossy multiplexing of several statistically identical sources. The computation of o* must be done at the network node. We show, by an example, the improvement in cost that is possible by lossy multiplexing and a positive U*. ). determine a set of optimal leaky bucket parameters, and measurement based estimation of these parameters. network node access:'
Social networks refer to structures made of nodes that represent people or other entities embedde... more Social networks refer to structures made of nodes that represent people or other entities embedded in a social context, and whose edges represent interaction between entities. Typical examples of social networks are collaboration networks in a research community, networks arising out of interaction between colleagues of large organization etc. Social networks are highly dynamic objects that evolve quickly over time with addition and deletion of nodes and edges. Understanding the evolution of a social network is helpful in inferring trends and patterns of social contacts in a particular social context. In this paper, we consider social networks that are derived from telephone call records, i.e, graphs in which the individual phone numbers (and hence its users) are the nodes and the edges correspond to a telephonic contact between the two nodes they connect. We study the problem of extracting dense communities from such telecom call graphs. The problem studied here is set in the context of a larger project. We motivate the problem studied by describing the context in which it is set. Our analysis is based on suitable algorithmic engineering of an approximation algorithm for the densest subgraph problem by Charikar. We present empirical results on massive graphs with millions of nodes and edges. We also discuss many open problems that are important in the context of analyzing telecom call graphs.
Merchants often use marketing elements such as advertisements, coupons and product recommendation... more Merchants often use marketing elements such as advertisements, coupons and product recommendations, to attract customers and to convert visitors to buyers. We present a model for making a series of recommendations during a customer session. The model comprises of the customer’s probability of accepting a marketing element from a marketing spot and a reward for the marketing element. The probabilities can be estimated from customer history (such as traversals and purchases), while the reward values could be merchant specified. We propose several recommendation strategies for maximising the merchant’s reward and analyse their effectiveness. Our experiments indicate that strategies that are dynamic and consider multiple marketing spots simultaneously perform well.
In this paper, we introduce a new problem which we call as Online Probabilistic Weighted Bipartit... more In this paper, we introduce a new problem which we call as Online Probabilistic Weighted Bipartite Graph Matching. Consider a weighted bipartite graph with a source node set, a target node set, an edge set and a set of weights for the edges. The source nodes arrive in an online fashion based on a stochastic process. When a source node arrives, it needs to be matched with a target node before the arrival of next source node. Since the arrival process is stochastic, all the source nodes need not arrive and their order of arrival is also not known a priori. The objective is to match the arriving source node with a target node such that the expected sum of weights of the matching over the arrival process is maximized. We present some heuristics that perform well for this problem. We demonstrate the application of our formulation for session based recommendation [5]. Here the source nodes correspond to the web pages, the target nodes correspond to the advertisement that can be shown and the edge weights correspond to the revenue generated by showing the given advertisement on the given web page. The user traversal of web pages corresponds to the arrival process of the source nodes.
We present an architecture for and prototype of a system for quickly detecting software problem r... more We present an architecture for and prototype of a system for quickly detecting software problem recurrences. Re-discovery of the same problem is very common in many large software products and is a major cost component of product support. At run-time, when a problem occurs, the system collects the problem symptoms, including the program call-stack, and compares it against a database of symptoms to find the closest matches. The database is populated off-line using solved cases and indexed to allow for efficient matching. Thus problems that occur repeatedly can be easily and automatically resolved without requiring any human problem-solving expertise. We describe a prototype implementation of the system, including the matching algorithm, and present some experimental results demonstrating the value of automatically detecting re-occurrence of the same problem for a popular sofware product.
Re-occurrence of the same problem is very common in many large software products. By matching the... more Re-occurrence of the same problem is very common in many large software products. By matching the symptoms of a new problem to those in a database of known problems, automated diagnosis and even selfhealing for re-occurrences can be (partially) realized. This paper exploits function call stacks as highly structured symptoms of a certain class of problems, including crashes, hangs, and traps. We propose and evaluate algorithms for efficiently and accurately matching call stacks by a weighted metric of the similarity of their function names, after first removing redundant recursion and uninformative (poor discriminator) functions from those stacks. We also describe a new indexing scheme to speed queries to the repository of known problems, without compromising the quality of matches returned. Experiments conducted using call stacks from actual product problem reports demonstrate the improved accuracy (both precision and recall) resulting from our new stack-matching algorithms and removal of uninformative or redundant function names, as well as the performance and scalability improvements realized by indexing call stacks. We also discuss how call-stack matching can be used in both self-managing (or autonomic systems) and human "help desk" applications.
Uploads
Papers by Natwar Modani