Ram Ramesh

An Optimal Algorithm for the Orienteering Tour Problem

Orienteering is a sport in which a competitor selects a path from a start to a destination, visiting control points along the path. Each control point has an associated score, and the travel between control points involves a certain cost. The problem is to select a set of control points to visit, so that the total score is maximized subject to a budget constraint on total cost. Several versions of this problem exist. In the version considered in this research, the start and the destination are the same, and the problem is to construct a subtour of the set of control points. The orienteering problem is a variant of the traveling salesman problem, and arises in vehicle routing and production scheduling situations. This problem has been shown to be NP-hard in the literature. We develop an optimal algorithm to solve this problem, using Lagrangean relaxation within a branch-and-bound framework. The Lagrangean relaxation is solved by a degree-constrained spanning tree procedure. Characteristics of the Lagrangea...

Efficient lot-sizing under a differential transportation cost structure for serially distributed warehouses

Abstract A major cost element in the logistics of distributed warehousing is transportation cost. In most practical systems, the transportation costs are volume-dependent. The unit transportation costs are usually determined differentially among intervals of shipment volumes. While the unit cost is constant over an interval, it follows a stepwise declining pattern from an interval to the next higher interval of shipment volumes. This structure is analogous to that of quantity-discounted inventory systems. We consider a single-product, serial warehousing system operating under the differential transportation costs and the traditional holding and ordering costs. External demand occurs at the final stage at a constant continuous rate. Demand must be met on time over an infinite horizon under continuous review. The objective is to determine the ordering lot size for each warehouse such that the long-run average cost is minimized. We model the above problem and establish its structural properties. We first consider two-level differential transportation cost structures and develop efficient algorithms to obtain a best integer ratio and optimal powers-of-two policies. Next, we consider multilevel cost structures and develop a hypercube characterization of the solution space. Efficient algorithms to yield a best integer ratio and powers-of-two policies using the hypercube model and a relaxation approach using the two-level model are developed. Extensive computational studies with the algorithms reveal that the proposed modeling and algorithmic approach is both viable and efficient in solving practical distribution problems. Conclusions and direction for future research are presented.

An interactive method for bicriteria integer programming

An efficient interactive solution framework for bicriteria integer programming is developed. The proposed methodology follows the implicit utility maximization approach. The decision makers underlying utility function is assumed to be pseudoconcave and nondecreasing, and the problem is solved using an interactive branch-and-bound methodology. Several new concepts on bicriteria integer programming that offer great efficiency in the solution process are developed. The framework has been tested extensively, and results with problems having up to 80 variables and 40 constraints are presented. The results show that the methodology is an effective approach to solving practical bicriteria problems. >

Risk Management and Optimal Pricing in Online Storage Grids

Online storage service providers grant a way for companies to avoid spending resources on maintaining their own in-house storage infrastructure and thereby allowing them to focus on their core business activities. These providers, however, follow a fixed, posted pricing strategy that charges the same price in each time period and thus bear all the risk arising out of demand uncertainties faced by their client companies. We examine the effects of providing a spot market with dynamic prices and forward contracts to hedge against future revenue uncertainty. We derive revenue-maximizing spot and forward prices for a single seller facing a known set of buyers. We perform a simulation study using publicly available traffic data regarding Amazon S3 clients from Alexa.com to validate our analytical results. Our field study supports our analysis and indicates that spot markets alone can enhance revenues to Amazon, but this comes at the cost of increased risks due to the increased market share in the spot markets. Furthermore, adding a forward contract feature to the spot markets can reduce risks while still providing the benefits of enhanced revenues. Although the buyers incur an increase in costs in the spot market, adding a forward contract does not cause any additional cost increase while transferring the risk to the buyers. Thus, storage grid providers can greatly benefit by applying a forward contract alongside the spot market.

Efficient Risk Hedging by Dynamic Forward Pricing: A Study in Cloud Computing

Commodities such as cloud resources storage, computing, bandwidth are often sold to clients on a pay-as-you-go basis. Thus, resource providers absorb all risk arising from end users demand volatilities. We focus on the revenue risk management of commodities with highly volatile demand profiles using cloud computing as the application domain and bandwidth as the exemplar commodity. We extend the state of the art in risk hedging by introducing a new concept of dynamic forward contracts where a provider and a client flexibly interact through offers and responses over a set of time periods in a horizon. We develop an optimal pricing mechanism that takes into account the risk propensities of the provider and the client. The overall mechanism is modeled as a pair of nested dynamic programs denoting the offer-response interactions. The mechanism also incorporates two learning components: short-term learning on the clients demand and long-term learning on the clients risk propensity. We characterize two approaches for predicting the clients demand---a recursive demand prediction model and an aggregate demand prediction model. Detailed experimental studies of the proposed mechanism using real Web traffic data on the clients of Amazon Web Services have been carried out. The empirical results clearly demonstrate the superiority of the proposed mechanism over benchmark mechanisms such as the current industry practice of spot markets and static forward pricing mechanisms proposed in the literature in ex ante and ex post settings. The results also highlight key interaction effects among parameters controllable by a provider and the risk propensities of the market players, leading to valuable managerial implications for the practical adoption of the proposed mechanism.

Service Agreement Trifecta: Backup Resources, Price and Penalty in the Availability-Aware Cloud

Service Level Agreements (SLA) for cloud services entail complex trade-offs between interrelated variables such as price, penalty, and service availability (uptime) guarantee, with resource managem...

Resource Adjustment and Intervention Scheduling in the Availability-Aware Cloud

Server instances, also known as virtual machines (VMs), with various computing and storage capabilities (i.e. different combinations of CPU, memory, storage, and networking), are commonly offered by cloud computing service providers such as Amazon Web Services (AWS), Google, Microsoft, Rackspace, and SalesForce. The clients, ranging from individuals and small institutions to large companies, can rent and pay only for a set of VMs that are actually needed as a usagebased (pay-as-you-go) posted pricing model. This model alleviates the clients’ concerns about the risks of capacity under-utilization and demand non-fulfillment associated with in-house facilities. These VMs are mapped onto physical servers within cloud. Such datacenters are known to be susceptible to different types of failures from frequent small-scale failures (such as disk failures) to less frequent but more catastrophic failures (such as power distribution unit failures) (Dean 2009). Failures at the physical infrastructure level inevitably render the mapped VMs unavailable due to loss of connectivity, software bugs, human errors, etc.

Predicting Transient Downtime in Virtual Server Systems: An Efficient Sample Path Randomization Approach

A central challenge in developing cloud datacenters Service Level Agreements is the estimation of downtime distribution of a set of provisioned servers over a service window, which is compounded by three facts. First, while steady-state probabilities have been derived for birth-death processes involving server failures and repairs, they could be highly inaccurate under transience. Furthermore, steady-state cannot be assured under typical service windows. Therefore, estimation of transient distributions is essential. Second, the processes of failures and repairs may follow any distribution and hence need to be extracted using system log data and modeled using appropriate general distributions. Third, downtime distributions over service windows depend on the number of servers and their deployment structure for a contract. We develop an efficient and generalized sample path randomization approach to precisely estimate transient probabilities under three different checkpointing strategies and three flexible failure distribution models. The estimators are unbiased, consistent, efficient and sufficient. Their asymptotic convergence is established. The estimation algorithms are computationally efficient in solving practical problems and yield rich information on transient system behaviors. The methodology is general and extensible to various server failure and repair processes characterized using birth-death modeling.

Revenue and Lifecycle Management for Digital Media Networks

Lifecycle management of digital media objects such as online videos is a challenging problem. While media objects are strong sources of revenue to content providers they also have limited shelf-life, influence each other in terms of customer choices, are data-intensive and require large bandwidth for delivery. As the revenue priorities often change rapidly, tiered solutions emerge as necessary and viable tools for lifecycle management. We model digital media objects as a network capturing the inter-object impacts and utilize this network structure to optimally partition media objects into tiers. Addressing the context of large content providers like Amazon VOD that employ infrastructure platforms like Limelight for content storage and delivery, we develop a bilevel programming model to maximize the profits of a price-setting platform and a tiered allocation-setting content provider. We predict and observe two fundamental effects with digital media: direct access differential revenue effect of tiered services and traffic generating effect of media objects. Using a detailed longitudinal empirical study we demonstrate the effectiveness of the proposed pricing and provisioning strategy and illustrate the existence and impact of these effects in media markets.

Risk Management in Globally Distributed Call Center Networks

The emergence of globally distributed call-center networks, such as Dell International Support Service, has fundamentally increased the challenges to the management. This new trend of networking faces the risk of extensive demand fluctuation both over locations and over time. Existing literature in operation management uses pooling mechanisms to deal with demand fluctuations among different departments, while does not consider system-wide time-varying demand shifts. Queuing literature has developed algorithms to allow the number of servers to change in response to time-varying loads, however, a pooling configuration introduces additional business risk and is thus more complex to change. More sophisticated approaches are needed to accurately describe the reality of call-center operations. In this paper, we study a model generalized from the International Queue run in Dells globally distributed call-center network and identify its operation risks from both demand fluctuation and pooling management. Then we propose an economic framework to use a Real-Options Approach (ROA) to systematically analyze those risks to help the management to hedge the risks to a pre-chosen level and hence secure the associated service quality and the costs.