Daniel R. Dooly

On-line analysis of the TCP acknowledgment delay problem

We study an on-line problem that is motivated by the networking problem of dynamically adjusting of acknowledgments in the Transmission Control Protocol (TCP). We provide a theoretical model for this problem in which the goal is to send acks at a time that minimize a linear combination of the cost for the number of acknowledgments sent and the cost for the additional latency introduced by delaying acknowledgments. To study the usefulness of applying packet arrival time prediction to this problem, we assume there is an oracle that provides the algorithm with the times of the next L arrivals, for some L ≥ 0. We give two different objective functions for measuring the cost of a solution, each with its own measure of latency cost. For each objective function we first give an O(n2)-time dynamic programming algorithm for optimally solving the off-line problem. Then we describe an on-line algorithm that greedily acknowledges exactly when the cost for an acknowledgment is less than the latency cost incurred by not acknowledging. We show that for this algorithm there is a sequence of n packet arrivals for which it is &OHgr; (***)-competitive for the first objective function, 2-competitive for the second function for L = 0, and 1-competitivefor the second function for L = 1. Next we present a second on-line algorithm which is a slight modification of the first, and we prove that it is 2-competitive for both objective functions for all L. We also give lower bounds on the competitive ratio for any deterministic on-line algorithm. These results show that for each objective function, at least one of our algorithms is optimal. Finally, we give some initial empirical results using arrival sequences from real network traffic where we compare the two methods used in TCP for acknowledgment delay with our two on-line algorithms. In all cases we examine performance with L = 0 and L = 1.

TCP dynamic acknowledgment delay (extended abstract): theory and practice

WQ study an on-line problem that is motivated by the networking problem of dynamically adjusting delays of acknowlcdgmcnts in the Transmission Control Protocol (TCP). The thcoroticnl problem we study is the following. There is a scqucnco of n packet arrival times .4 = (or,. . . , an) and a look-nhoad coefficient L, The goal is to partition A into 1; subsequences orroe,. . . , a~: (where a subsequence end is deilncd by an acknowledgment) that minimizes a linear combination of the cost for the number of acknowledgments sent and the cost for the additional latency introduced by delaylng acknowledgments. At each arrival, an oracle provides the algorithm with the times of the next L arrivals. First we give an 0(n*) dynamic programming algorithm for optimally solving the off-line problem. Then we deacribc an on-line algorithm that greedily acknowledges exactly when the cost for an acknowledgment is less than the latency cost obtained by not acknowledging. We show that for this algorithm there ls a sequence of n packet arrivals for which it is S-l (fi) -competitive. Next we present a second on-line algorithm which is a slight modification of the first that we prove is 2-competitive. Let Copt be the cost of the optimal solution and let CA be the cost of the solution produced by algorithm A. We then show that for any on-line algorithm A with any constant look-ahead L, CA > SC+-c where c is a factor that can be made arbitrarily small with rcnpcct to Capt. Thus, in the worst case, our result for L = 0 IQ the hcst possible even for on-line algorithms that can use nny constant look-ahead. We then give some initial empirical results using arrival sequences from real network traffic where we compare the two methods used in TCP for acknowledgment delay with our two on-line algorithms. In all cases we examine performance mitli L = 0 and L = 1, Finally, we consider an alternate dcilnition for the latency cost in our objective function, *Supported in part by NSF NY1 Grant CC%9357707 with mntching funds provfdcd by WUTA.

TCP Dynamic Acknowledgment Delay: Theory and Practice

We study an on-line problem that is motivated by the networking problem of dynamically adjusting delays of acknowledgments in the Transmission Control Protocol (TCP). The theoretical problem we study is the following. There is a sequence of n packet arrival times A = and a look-ahead coefficient L. The goal is to partition A into k subsequences sigma1, sigma2, ...,sigmak (where a subsequence end is defined by an acknowledgment) that minimizes a linear combination of the cost for the number of acknowledgments sent and the cost for the additional latency introduced by delaying acknowledgments. At each arrival, an oracle provides the algorithm with the times of the next L arrivals. For all the results of our paper, we describe how to incorporate other contraints to better match the true acknowledgment delay problem. We first define two different objective functions for measuring the cost of a solution, each with its own measure of latency cost. For each objective function we first given an O (nsquared)-time dynamic programming algorithm for optimally solbing the off-line problem. Then... Read complete abstract on page 2.

Algorithms for the constrained maximum-weight connected graph problem

Given a positive integer R and a weight for each vertex in a graph, the maximum-weight connected graph (MCG) problem is to find a connected subgraph with R vertices that maximizes the sum of the weights. The MCG problem is strongly NP-complete, and we study a special case of it: the constrained MCG (CMCG) problem, which is the MCG problem with a constraint of having a predetermined vertex included in the solution. We first show that the Steiner tree problem is a special case of the CMCG problem. Then we present three optimization algorithms for the CMCG problem. The first two algorithms deal with special graphs (tree and layered graphs) and employ different dynamic programming techniques, solving the CMCG problem in polynomial times. The third one deals with a general graph and uses a variant of the Balas additive method with an imbedded connectivity test and a pruning method. We also present a heuristic algorithm for the CMCG problem with a general graph and its bound analysis. We combine the two algorithms, heuristic and optimization, and present a practical solution method to the CMCG problem. Computational results are reported and future research issues are discussed.

A shift-based sequencing method for twin-shuttle automated storage and retrieval systems

The continuing need for high-throughput Automated Storage and Retrieval Systems (AS/RS) has lead to the introduction of storage/retrieval machines that can carry more than one unit-load. However, this technology involves a large capital investment so careful operating methods are desired to make the most of its capabilities. In this paper, we study a shift-based sequencing problem for twin-shuttle AS/RS, where depletion (retrieval operations) and replenishment (storage operations) of items occur over different shifts. For example, certain warehouses or distribution depots deplete their items in stock during morning shifts and replenish during later shifts. We show that this problem can be transformed into the minimum-cost perfect matching problem and present an efficient polynomial-time optimum method that can achieve a large throughput gain over other methods. We also provide average-case and lower bound analyses for this problem.

Decomposition algorithms for the maximum‐weight connected graph problem

Given a positive integer R and a weight for each vertex in a graph, the maximum-weight connected graph problem (MCG) is to find a connected subgraph with R vertices that maximizes the sum of their weights. MCG has applications to communication network design and facility expansion. The constrained MCG (CMCG) is MCG with a constraint that one predetermined vertex must be included in the solution. In this paper, we introduce a class of decomposition algorithms for MCG. These algorithms decompose MCG into a number of small CMCGs by adding vertices one at a time and building a partial graph. They differ in the ordering of adding vertices. Proving that finding an ordering that gives the minimum number of CMCGs is NP-complete, we present three heuristic algorithms. Experimental results show that these heuristics are very effective in reducing computation and that different orderings can significantly affect the number of CMCGs to be solved.

Smartacking: Improving TCP Performance from the Receiving End

We present smartacking, a technique that improves performance of Transmission Control Protocol (TCP) via adap- tive generation of acknowledgments (ACKs) at the receiver. When the bottleneck link is underutilized, the receiver transmi ts an ACK for each delivered data segment and thereby allows the connection to acquire the available capacity promptly. When the bottleneck link is at its capacity, the smartacking receiver sends ACKs with a lower frequency reducing the control traffi c overhead and slowing down the congestion window growth to utilize the network capacity more effectively. To promote quick deployment of the technique, our primary implementation of smartacking modifies only the receiver. This implementatio n estimates the senders congestion window using a novel algo rithm of independent interest. We also consider different implemen- tations of smartacking where the receiver relies on explicit assistance from the sender or network. Experiments in a wide variety of settings show beneficial impacts of smartacking o n TCP performance, especially in environments with low levels of connection multiplexing on bottleneck links.

Real-Valued Multiple-Instance Learning with Queries

The multiple-instance model was motivated by the drug activity prediction problem where each example is a possible configuration for a molecule and each bag contains all likely configurations for the molecule. While there has been a significant amount of theoretical and empirical research directed towards this problem, most research performed under the multiple-instance model is for concept learning. However, binding affinity between molecules and receptors is quantitative and hence a real-valued classification is preferable.In this paper we initiate a theoretical study of real-valued multiple instance learning. We prove that the problem of finding a target point consistent with a set of labeled multiple-instance examples (or bags) is NP-complete. We also prove that the problem of learning from realvalued multiple-instance examples is as hard as learning DNF. Another contribution of our work is in defining and studying a multiple-instance membership query (MI-MQ). We give a positive result on exactly learning the target point for a multiple-instance problem in which the learner is provided with a MI-MQ oracle and a single adversarially selected bag.

Heuristic algorithms for the fiber optic network expansion problem

We consider a network expansion problem for fiber optic networks. Given the network topology, edge costs, and node profit per period, the network is built over T periods due to limited budget available per period. The problem is to find edges to be added at each period so that their edge cost does not exceed the budget limit and the resulting network is connected, with the objective of maximizing profit over T periods. We first consider some special graphs: lines, cycles, and trees. Then we show that NEP is strongly NP‐complete for general graphs, and present three heuristic algorithms and their performance analyses. Computational results show evidence that the simple greedy algorithm performs very effectively for finding a near optimal solution.

Real-valued multiple-instance learning with queries

While there has been a significant amount of theoretical and empirical research on the multiple-instance learning model, most of this research is for concept learning. However, for the important application area of drug discovery, a real-valued classification is preferable. In this paper we initiate a theoretical study of real-valued multiple-instance learning. We prove that the problem of finding a target point consistent with a set of labeled multiple-instance examples (or bags) is NP-complete, and that the problem of learning from real-valued multiple-instance examples is as hard as learning DNF. Another contribution of our work is in defining and studying a multiple-instance membership query (MI-MQ). We give a positive result on exactly learning the target point for a multiple-instance problem in which the learner is provided with a MI-MQ oracle and a single adversarially selected bag.