By sharing resources among different cloud providers, the paradigm of federated clouds exploits temporal availability of resources and geographical diversity of operational costs for efficient job service. While interoperability issues across different cloud platforms in a cloud federation have been extensively studied, fundamental questions on cloud economics remain: When and how should a cloud trade resources (e.g., virtual machines) with others, such that its net profit is maximized over the long run, while a close-to-optimal social welfare in the entire federation can also be guaranteed? To answer this question, a number of important, interrelated decisions, including job scheduling, server provisioning, and resource pricing, should be dynamically and jointly made, while the long-term profit optimality is pursued. In this work, we design efficient algorithms for intercloud virtual machine (VM) trading and scheduling in a cloud federation. For VM transactions among clouds, we design a double-auction-based mechanism that is strategy-proof, individual-rational, ex-post budget-balanced, and efficient to execute over time. Closely combined with the auction mechanism is a dynamic VM trading and scheduling algorithm, which carefully decides the true valuations of VMs in the auction, optimally schedules stochastic job arrivals with different service level agreements (SLAs) onto the VMs, and judiciously turns on and off servers based on the current electricity prices. Through rigorous analysis, we show that each individual cloud, by carrying out the dynamic algorithm in the online double auction, can achieve a time-averaged profit arbitrarily close to the offline optimum. Asymptotic optimality in social welfare is also achieved under homogeneous cloud settings. We carry out simulations to verify the effectiveness of our algorithms, and examine the achievable social welfare under heterogeneous cloud settings, as driven by the real-world Google cluster usage traces.

Spectrum auctions are efficient mechanisms for licensed users to relinquish their under-utilized spectrum to secondary links for monetary remuneration. Truthfulness and social welfare maximization are two natural goals in such auctions, but cannot be achieved simultaneously with polynomial-time complexity by existing methods, even in a static network with fixed parameters. The challenge escalates in practical systems with QoS requirements and volatile traffic demands for secondary communication. Online, dynamic decisions are required for rate control, channel evaluation/bidding, and packet dropping at each secondary link, as well as for winner determination and pricing at the primary user. This work proposes an online spectrum auction framework with cross-layer decision making and randomized winner determination on the fly. The framework is truthful-in-expectation, and achieves close-to-offline-optimal time-averaged social welfare and individual utilities with polynomial time complexity. A new method is introduced for online channel evaluation in a stochastic setting. Simulation studies further verify the efficacy of the proposed auction in practical scenarios.

With the recent advent of cloud computing technologies, a growing number of content distribution applications are contemplating a switch to cloud-based services, for better scalability and lower cost. Two key tasks are involved for such a move: to migrate the contents to cloud storage, and to distribute the Web service load to cloud-based Web services. The main issue is to best utilize the cloud as well as the application provider's existing private cloud, to serve volatile requests with service response time guarantee at all times, while incurring the minimum operational cost. While it may not be too difficult to design a simple heuristic, proposing one with guaranteed cost optimality over a long run of the system constitutes an intimidating challenge. Employing Lyapunov optimization techniques, we design a dynamic control algorithm to optimally place contents and dispatch requests in a hybrid cloud infrastructure spanning geo-distributed data centers, which minimizes overall operational cost overtime, subject to service response time constraints. Rigorous analysis shows that the algorithm nicely bounds the response times within the preset QoS target, and guarantees that the overall cost is within a small constant gap from the optimum achieved by a T-slot lookahead mechanism with known future information. We verify the performance of our dynamic algorithm with prototype-based evaluation.

More and more enterprises have set up their own private clouds by applying virtualization to their data centers, the benefit is flexible resource supply to different internal demands. Aiming to meet the peak demand in their resource provisioning, private clouds are often under-utilized. A new paradigm has emerged that advocates leasing the spare resources to external users, when and if adequate rental prices are offered. A broker is typically employed which pools the spare resources of multiple private clouds together and leases them to serve external users' jobs. Good mechanisms have yet to be derived for the broker to set the offered prices to buy spare resources from the private clouds, and to schedule jobs on the available resources, such that the economic benefits of both the broker and the private clouds are maximized. The design of the mechanism is especially challenging when we consider the dynamic arrival of users' jobs and volatile availability of spare resources at the private clouds, while aiming at long-term profit optimality. In this paper, we model the interaction between the broker and the private clouds as a two-stage Stackelberg game. As the leader in the game, the broker decides and offers prices for renting VMs of different types from each private cloud. As a follower, each private cloud responds with the number of VMs of each type that it is willing to lease. By combining with the Stackelberg game model we design online algorithms for the broker to set the prices and schedule jobs on the private clouds, and for the private cloud to decide the numbers of VMs to lease, based on the Lyapunov optimization theory. We prove that the broker achieves a time-averaged profit that is close to the offline optimum with complete information on future job arrivals and resource availability, while each private cloud makes their best earning. The proposed online algorithm is carefully evaluated based on usage traces of Google cluster and Amazon EC2.

Cloud providers often choose to operate datacenters over a large geographic span, in order that users may be served by resources in their proximity. Due to time and spatial diversities in utility prices and operational costs, different datacenters typically have disparate charges for the same services. Cloud users are free to choose the datacenters to run their jobs, based on a joint consideration of monetary charges and quality of service. A fundamental problem with significant economic implications is how the cloud should price its datacenter resources at different locations, such that its overall profit is maximized. The challenge escalates when dynamic resource pricing is allowed and long-term profit maximization is pursued. We design an efficient online algorithm for dynamic pricing of VM resources across datacenters in a geo-distributed cloud, together with job scheduling and server provisioning in each datacenter, to maximize the profit of the cloud provider over a long run. Theoretical analysis shows that our algorithm can schedule jobs within their respective deadlines, while achieving a time-average overall profit closely approaching the offline maximum, which is computed by assuming that perfect information on future job arrivals are freely available. Empirical studies further verify the efficacy of our online profit maximizing algorithm.

The emerging federated cloud paradigm advocates sharing of resources among cloud providers, to exploit temporal availability of resources and diversity of operational costs for job serving. While extensive studies exist on enabling interoperability across different cloud platforms, a fundamental question on cloud economics remains unanswered: When and how should a cloud trade VMs with others, such that its net profit is maximized over the long run? In order to answer this question by the federation, a number of important, correlated decisions, including job scheduling, server provisioning and resource pricing, need to be dynamically made, with long-term profit optimality being a goal. In this work, we design efficient algorithms for inter-cloud resource trading and scheduling in a federation of geo-distributed clouds. For VM trading among clouds, we apply a double auction-based mechanism that is strategy proof, individual rational, and ex-post budget balanced. Coupling with the auction mechanism is an efficient, dynamic resource trading and scheduling algorithm, which carefully decides the true valuations of VMs in the auction, optimally schedules stochastic job arrivals with different SLAs onto the VMs, and judiciously turns on and off servers based on the current electricity prices. Through rigorous analysis, we show that each individual cloud, by carrying out our dynamic algorithm, can achieve a time-averaged profit arbitrarily close to the offline optimum.

In a cognitive radio system, licensed primary users can lease idle spectrum to secondary users for monetary remuneration. Secondary users acquire available spectrum for their data delivery needs, with the goal of achieving high throughput and low spectrum charges. Maximizing such a net utility (throughput utility minus spectrum cost) is a central problem faced by a multihop secondary network. Optimal decision making is challenging, since it involves multiple data flows, cross-layer coordination, and economic constraints (budgets of sources). The picture is further complicated by the inter-play between secondary data communication and primary spectrum leasing mechanisms. This work is the first to investigate the full spectrum of socially optimal secondary user communication. We design a social welfare maximization framework for multi-session multi-hop secondary data dissemination based on Lyapunov optimization techniques. A salient feature of the framework is that it takes any given primary user mechanism as input, and produces correspondingly a dynamic, distributed rate control, routing, and spectrum allocation and pricing protocol that can achieve longterm maximization of the overall system utility. Through rigorous theoretical analysis, we prove that our online protocol can achieve a social welfare that is arbitrarily close to the offline optimum, with only finite buffer space requirement at each secondary user, and guarantee of no buffer overflow. Empirical studies are conducted to examine the performance of the protocol.

Minimizing latency and energy consumption is the prime objective of the design of data aggregation in battery-powered wireless networks. A tradeoff exists between the aggregation latency and the energy consumption, which has been widely studied under the protocol interference model. There has been, however, no investigation of the tradeoff under the physical interference model that is known to capture more accurately the characteristics of wireless interferences. When coupled with the technique of successive interference cancellation, by which a receiver may recover signals from multiple simultaneous senders, the model can lead to much reduced latency but increased energy usage. In this paper, we investigate the latency-energy tradeoff for data aggregation in wireless sensor networks under the physical interference model and using successive interference cancellation. We present theoretical lower bounds on both latency and energy as well as their tradeoff, and give an efficient approximation algorithm that can achieve the asymptotical optimum in both aggregation latency and latency-energy tradeoff. We show that our algorithm can significantly reduce the aggregation latency, for which the energy consumption is kept at its lowest possible level.

Multicast supporting non-uniform receiving rates is an effective means of data dissemination to receivers with diversified bandwidth availability. Designing efficient rate control, routing and capacity allocation to achieve optimal multirate multicast has been a difficult problem in fixed wireline networks, let alone wireless networks with random channel fading and volatile node mobility. The challenge escalates if we consider also the selfishness of users who prefer to relay data for others with strong social ties. Such social selfishness of users is a new constraint in network protocol design. Its impact on efficient multicast in wireless networks has yet to be explored especially when multiple receiving rates are allowed. In this paper, we design an efficient, social-aware multirate multicast scheme that can maximize the overall utility of socially selfish users in a wireless network, and its distributed implementation. We model social preferences of users as differentiated costs for packet relay, which are weighted by the strength of social tie between the relay and the destination. Stochastic Lyapunov optimization techniques are utilized to design optimal scheduling of multicast transmissions, which are combined with multi-resolution coding and random linear network coding. With rigorous theoretical analysis, we study the optimality, stability, and complexity of our algorithm, as well as the impact of social preferences. Empirical studies further confirm the superiority of our algorithm under different social selfishness patterns.

The recent advent of cloud computing technologies has enabled agile and scalable resource access for a variety of applications. Content distribution services are a major category of popular Internet applications. A growing number of content providers are contemplating a switch to cloud-based services, for better scalability and lower cost. Two key tasks are involved for such a move: to migrate their contents to cloud storage, and to distribute their web service load to cloud-based web services. The main challenge is to make the best use of the cloud as well as their existing on-premise server infrastructure, to serve volatile content requests with service response time guarantee at all times, while incurring the minimum operational cost. Employing Lyapunov optimization techniques, we present an optimization framework for dynamic, cost-minimizing migration of content distribution services into a hybrid cloud infrastructure that spans geographically distributed data centers. A dynamic control algorithm is designed, which optimally places contents and dispatches requests in different data centers to minimize overall operational cost over time, subject to service response time constraints. Rigorous analysis shows that the algorithm nicely bounds the response times within the preset QoS target in cases of arbitrary request arrival patterns, and guarantees that the overall cost is within a small constant gap from the optimum achieved by a T-slot lookahead mechanism with known information into the future.

In cognitive radio networks, the occupation patterns of the primary users can be very dynamic, which makes optimization (e.g., utility maximization) of data dissemination among secondary users difficult. Even under the assumption that all secondary users are fully collaborative, the optimization requires cross-layer decision making which is challenging. The challenge escalates if users are socially selfish, who prefer to relay data only to those other users with whom there are social ties. Such social selfishness of users translates into new constraints on network protocol design. There has been no study so far on the impact of social selfishness on data dissemination in cognitive radio networks. In this paper, we consider social selfishness of secondary users, and propose the design of a joint end-to-end rate control, routing, and channel allocation protocol which can maximize the overall throughput utility of multi-session unicast in cognitive radio networks. We give a distributed implementation of the protocol. Based on a Lyapunov optimization framework, we address social preferences of users using differentiated buffer sizes and relay rates for different data sessions, and apply back-pressure based transmission scheduling to achieve guaranteed utility optimality. A unique contribution of our Lyapunov optimization is that only a finite-sized buffer is required at each user node, which sets our design apart from other designs in existing literature where they assume infinite buffers. We investigate the the optimality of our protocol and the impact of user social selfishness using both theoretical analysis and extensive simulations.

Minimum-Latency Aggregation Scheduling (MLAS) is a problem of fundamental importance in wireless sensor networks. There however has been very little effort spent on designing algorithms to achieve sufficiently fast data aggregation under the physical interference model which is a more realistic model than traditional protocol interference model. In particular, a distributed solution to the problem under the physical interference model is challenging because of the need for global-scale information to compute the cumulative interference at any individual node. In this paper, we propose a distributed algorithm that solves the MLAS problem under the physical interference model in networks of arbitrary topology in O(K) time slots, where K is the logarithm of the ratio between the lengths of the longest and shortest links in the network. We also give a centralized algorithm to serve as a benchmark for comparison purposes, which aggregates data from all sources in O(log3n) time slots (where n is the total number of nodes). This is the current best algorithm for the problem in the literature. The distributed algorithm partitions the network into cells according to the value K, thus obviating the need for global information. The centralized algorithm strategically combines our aggregation tree construction algorithm with the non-linear power assignment strategy in [9]. We prove the correctness and efficiency of our algorithms, and conduct empirical studies under realistic settings to validate our analytical results.