Informs Annual Meeting 2017

MB03B

INFORMS Houston – 2017

MA02

data are difficult to generate. In this tutorial, we provide an introduction to disaggregate demand models that are designed to capture in detail the underlying behavioral mechanisms at the foundation of the demand.

310B Military Decision Analysis Sponsored: Decision Analysis Sponsored Session Chair: Greg Parnell, University of Arkansas, gparnell@uark.edu 1 - Lessons Learned from Adaptive Red Team System Assessments Simon Goerger, PhD, ERDC, CEERD-IE, 3909 Halls Ferry Road, Vicksburg, MS, 39180-6199, United States, The Adaptive Red Team draws from a portfolio of methods to identify potential vulnerabilities in emerging technologies designed for Soldier use in contested environments, with interoperability challenges, adaptability needs, and training constraints. This presentation describes novel data collection and assessment system developed to support live experiment data collection. It shows the connection to resilient system designs by identifying hidden vulnerabilities introduced by the physical environment, congestion and frequency overload, and electronic warfare and computer network attacks. 2 - Considering System Resilience as Active and Passive Response Strategies to Perturbations Over Time Randy K. Buchanan, ERDC-Vicksburg, MS, 3909 Halls Ferry Road, Vicksburg, MS, 39180, United States, randy.k.buchanan@erdc.dren.mil, Adam M. Ross, Christina Rinaudo Despite no widely-accepted definition of resilience, the need for developing systems that can deal with the potential impact of uncertainties (i.e. perturbations) is a real one. In this research, we propose a practical approach toward framing, eliciting, and guiding analysis of tradeoffs for implementing aspects of resilience in military systems. The approach involves partitioning resilience into four related components: 1) assessing response coverage of perturbations, 2) assessing active strategies (e.g. flexibility), 3) assessing passive strategies (e.g. versatility), and 4) assessing timescale of responses. 3 - Using Process Improvement Techniques to Develop an Innovation Pipeline Procedure Niki Goerger, Business Development Director, U.S. Army Engineer Research and Development Center, 3909 Halls Ferry Road, An enterprise innovation pipeline process can provide an organized and structured methodology to ensure that new ideas are continuously generated, evaluated, and matured. Process improvement techniques were used to develop and refine a baseline innovation pipeline process. This presentation explores process improvement methods and techniques used to develop an enterprise innovation pipeline process at the Engineer Research and Development Center. 4 - Quantifying Resilience for Engineered Systems Gregory S. Parnell, University of Arkansas, 4207 Bell Engineering Center, Fayetteville, AR, 72701, United States, gparnell@uark.edu A resilient engineered system is able to successfully complete its planned mission(s) in the face of environmental and adversarial threats, and has capabilities allowing it to flexibly adapt to future missions with evolving threats. We use this definition to propose four methods to assess resilience for military systems: updated mission chain calculations for each performance measure; a Value-Focused Thinking Model with mission and platform resilience measures; a multiple objective value model with mission resilience value measures; and two separate value models: one including mission reliance and one for platform resilience. Simon.R.Goerger@usace.army.mil, Niki C.Goerger, Michael Rainey, Randy K.Buchanan, John R.Burt Vicksburg, MS, Postal code39180-619, United States, Maria.N.Goerger@usace.army.mil, Christina Rinaudo, Simon Goerger 310C Introduction to Disaggregate Demand Models Invited: Tutorial Invited Session Chair: Jiming Peng, University of Houston, Houston, TX, 77204, United States, jopeng@Central.uh.edu Co-Chair: Rajan Batta, University at Buffalo (SUNY), 410 Bell Hall, Buffalo, NY, 14260, United States, batta@buffalo.edu 1 - Introduction to Disaggregate Demand Models Michel Bierlaire, EPFL, Lausanne, Switzerland, michel.bierlaire@epfl.ch, Virginie Lurkin Demand information is an input for a great deal of operations research models. Assumed as given in many problem instances addressed in the literature, demand MA03

MA03A Grand Ballroom A Cloud Computing and Data Centers in Applied Probability II Sponsored: Applied Probability Sponsored Session Chair: Siva Theja Maguluri, IBM, Sleepy Hollow, NY, 10591, United States, siva.theja@gmail.com 1 - Dynamic Resource Provisioning in Data Centers via Robust Queueing Theory Chaithanya Bandi, Northwestern University, 2001 Sheridan Road, Suite 566, Evanston, IL, 60208, United States, c-bandi@kellogg.northwestern.edu, Vineet Goyal, Omar El Housni We consider the problem of dynamic resource provisioning and use tools from robust queueing theory and multi-stage optimization to present tractable solutions. 2 - Efficient Redundancy in Cloud Systems Gauri Joshi, Carnegie Mellon University, Pittsburgh, PA, United States, gaurij@andrew.cmu.edu The throughput (rate of task completion) of a multi-server system is typically the sum of the service rates of individual servers. Recent works show that task replication can boost the throughput, in particular if the service time distribution has high variability. This paper seeks to find the fundamental limit of the throughput boost achieved by task replication, which is an open problem. The optimal replication policy can be found using a Markov Decision Process framework, which is hard to solve in general. We obtain a converse bound on the service capacity, and propose a close-to-optimal replication policy based on it. 3 - Bounds on Coflow Scheduling in Datacenter Networks Javad Ghaderi, Columbia University, jghaderi@ee.columbia.edu In many data-parallel computing frameworks, intermediate parallel data is often produced which needs to be transferred among servers in the datacenter network. Coflow is a networking abstraction to capture such communication patterns. We consider the problem of efficiently scheduling coflows with release dates in a shared datacenter network so as to minimize the total weighted completion time of coflows. We present a deterministic algorithm with approximation ratio of 5, which improves the prior best known ratio of 12. For the special case when all coflows are released at time zero, we obtain a deterministic algorithm with approximation ratio of 4 which improves the prior best known ratio of 8. 4 - Joint Congestion Control and Scheduling for Optimal Delay in Stochastic Processing Networks Qiaomin Xie, Massachusetts Institute of Technology, 32-D771B, 32 Vassar Street, Cambridge, MA, 02139, United States, qxie@mit.edu, Devavrat Shah We consider the problem of designing and analyzing joint congestion control and scheduling mechanisms for large-scale stochastic systems, exemplified by datacenter networks, the Internet and wireless networks. We propose a scheme that guarantees a per-flow end-to-end delay bound of O(# of hops). Designing such a scheme involves two subproblems: rate control and packet scheduling. We design a flow-level rate control policy by using an insensitive rate allocation policy in so-called bandwidth-sharing networks. A packet-level scheduling scheme is obtained by adapting the Last-In-First-Out Preemptive-Resume policy.

MA03B Grand Ballroom B

Choice-based Revenue Management Sponsored: Revenue Management & Pricing Sponsored Session

Chair: Gustavo Jose Vulcano, Universidad Torcuato di Tella, Universidad Torcuato di Tella, Buenos Aires, na, Argentina, gvulcano@utdt.edu 1 - Demand Modeling in the Presence of Unobserved Lost Sales Pavithra Harsha, IBM.Research, 1101 Kitchawan Rd, Room 34-225, Yorktown Heights, NY, 10598, United States, pharsha@us.ibm.com, Shivaram Subramanian We present an integrated optimization approach to parameter estimation for discrete choice demand models where data for one or more choice alternatives are censored. We employ a mixed-integer program to jointly determine the prediction parameters associated with the customer arrival rate and their

131