Shiyao Chen

Queuing Analysis in Multichannel Cognitive Spectrum Access: A Large Deviation Approach

The queueing performance of a (secondary) cognitive user is investigated for a hierarchical network where there are N independent and identical primary users. Each primary user employs a slotted transmission protocol, and its channel usage forms a two-state (busy,idle) discrete-time Markov chain. The cognitive user employs the optimal policy to select which channel to sense (and use if found idle) at each slot. In the framework of effective bandwidths, the stationary queue tail distribution of the cognitive user is estimated using a large deviation approach for which closed-form expressions are obtained when N = 2. Upper and lower bounds are obtained for the general N primary user network. For positively correlated primary transmissions, the bounds are shown to be asymptotically tight. Monte Carlo simulations using importance sampling techniques are used to validate the obtained large deviation estimates.

Optimal deadline scheduling with commitment

We consider an online preemptive scheduling problem where jobs with deadlines arrive sporadically. A commitment requirement is imposed such that the scheduler has to either accept or decline a job immediately upon arrival. The schedulers decision to accept an arriving job constitutes a contract with the customer; if the accepted job is not completed by its deadline as promised, the scheduler loses the value of the corresponding job and has to pay an additional penalty depending on the amount of unfinished workload. The objective of the online scheduler is to maximize the overall profit, i.e., the total value of the admitted jobs completed before their deadlines less the penalty paid for the admitted jobs that miss their deadlines. We show that the maximum competitive ratio is 3 − √2 and propose a simple online algorithm to achieve this competitive ratio. The optimal scheduling includes a threshold admission and a greedy scheduling policies. The proposed algorithm has direct applications to the charging of plug-in hybrid electrical vehicles (PHEV) at garages or parking lots.

Large scale charging of Electric Vehicles

The problem of scheduling for large scale charging of Electric Vehicles (EVs) is considered. As part of the future EV infrastructure, a Large Scale Charging (LSC) facility is capable of charging hundreds of electric vehicles simultaneously. As an intelligent load in the future smart grid, LSC requires properly designed pricing and scheduling algorithms that take into account the electricity consumed, the arrival-departure characteristics, and overall charging capacity. The scheduling of LSC is formulated as a deadline scheduling problem. Utility functions that combine both amount of charge and tightness of the deadline are proposed. Under arbitrary (and deterministic) arrival, departure, and charging characteristics, a scheduling policy referred to as deadline scheduling with admission control is proposed. The proposed algorithm achieves the highest competitive ratio (against the best offline scheduling) for the utility function linear in charging level among all online scheduling algorithms. It also offers significant gain over benchmark scheduling algorithms such as the Earliest Deadline First (EDF) scheduling and the First Come First Serve (FCFS) scheduling in terms of average performance for general utility functions when tested with randomly generated charging requests.

Deadline scheduling for large scale charging of electric vehicles with renewable energy

The problem of scheduling for the large scale charging of electric vehicles with renewable sources is considered. A new online charging algorithm referred to as Threshold Admission with Greedy Scheduling (TAGS) is proposed by formulating the charging problem as one of deadline scheduling with admission control and variable charging capacities. TAGS has low computation cost and requires no prior knowledge on the distributions of arrival traffic, battery charging (service) time, and available energy from renewable sources. It has a reserve dispatch algorithm designed to compensate the intermittency of renewable sources. Performance of TAGS is compared with benchmark scheduling algorithms such as the Earliest Deadline First (EDF) and the First Come First Serve (FCFS) with aggressive and conservative reserve dispatch algorithms.

Maximum Throughput Region of Multiuser Cognitive Access of Continuous Time Markovian Channels

The problem of cognitive access of multiple primary channels by multiple cognitive users is considered. The primary transmission on each channel is modeled by a continuous time Markov on-off process. Cognitive access of the primary channels is realized via channel sensing. Each cognitive user adopts a slotted transmission structure, senses one channel in each slot and makes the transmission decision based on the sensing outcome. The cognitive transmissions in each channel are subject to collision constraints that limit their interference to the primary users. The maximum throughput region of this multiuser cognitive network is characterized by establishing inner and outer bounds. Under tight collision constraints, the inner bound is obtained by a simple orthogonalized periodic sensing with memoryless access policy and its generalizations. The outer bound, on the other hand, is obtained by relating the sum throughput with the interference limits. It is shown that when collision constraints are tight, the outer and inner bounds match. This maximum throughput region result is further extended by a generalized periodic sensing scheme with a mechanism of timing sharing. Under general collision constraints, another outer bound is obtained via Whittles relaxation and another inner bound obtained via Whittles index sensing policy with memoryless access. Packet level simulations are used to validate the analytical performance prediction.

Secondary Job Scheduling in the Cloud with Deadlines

The highly dynamic nature of the cloud environment leads to a time-varying resource utilization and the cloud provider can potentially accommodate secondary jobs with the remaining resource. To better implement the idea of resource reutilization in the cloud environment, the problem of secondary job scheduling with deadlines under time-varying resource capacity is considered in this paper. A transformation is proposed to reduce the offline problem with time-varying processor capacity to that with constant capacity. For online scheduling of under loaded system, it is shown that the earliest deadline first (EDF) scheduling algorithm achieves competitive ratio 1. For the overloaded system, an online scheduling algorithm V-Dover is proposed with asymptotically optimal competitive ratio when a certain admissibility condition holds. It is further shown that, in the absence of the admissibility condition, no online scheduling algorithm exists with a positive competitive ratio. Simulation results are presented to illustrate the performance advantage of the proposed V-Dover algorithm.

To migrate or to wait: Bandwidth-latency tradeoff in opportunistic scheduling of parallel tasks

We consider the problem of scheduling low-priority tasks onto resources already assigned to high-priority tasks. Due to burstiness of the high-priority workloads, the resources can be temporarily underutilized and made available to the low-priority tasks. The increased level of utilization comes at a cost to the low-priority tasks due to intermittent resource availability. Focusing on two major costs, bandwidth cost associated with migrating tasks and latency cost associated with suspending tasks, we aim at developing online scheduling policies achieving the optimal bandwidth-latency tradeoff for parallel low-priority tasks with synchronization requirements. Under Markovian resource availability models, we formulate the problem as a Markov Decision Process (MDP) whose solution gives the optimal scheduling policy. Furthermore, we discover structures of the problem in the special case of homogeneous availability patterns that enable a simple threshold-based policy that is provably optimal. We validate the efficacy of the proposed policies by trace-driven simulations.

Delay optimal multichannel opportunistic access

The problem of minimizing queueing delay of opportunistic access of multiple continuous time Markov channels is considered. A new access policy based on myopic sensing and adaptive transmission (MS-AT) is proposed. Under the framework of risk sensitive constrained Markov decision process with effective bandwidth as a measure of queueing delay, it is shown that MS-AT achieves simultaneously throughput and delay optimality. It is shown further that both the effective bandwidth and the throughput of MS-AT are two-segment piece-wise linear functions of the collision constraint (maximum allowable conditional collision probability) with the effective bandwidth and throughput coinciding in the regime of tight collision constraints. Analytical and simulation comparisons are conducted with the myopic sensing and memoryless transmission (MS-MT) policy which is throughput optimal but delay suboptimal in the regime of tight collision constraints.

Distributed Learning and Multiaccess of On-Off Channels

The problem of distributed access of a set of N on-off channels by K ≤ N users is considered. The channels are slotted and modeled as independent but not necessarily identical alternating renewal processes. Each user decides to either observe or transmit at the beginning of every slot. A transmission is successful only if the channel is at the on state and there is only one user transmitting. When a user observes, it identifies whether a transmission would have been successful had it decided to transmit. A distributed learning and access policy referred to as alternating sensing and access (ASA) is proposed. It is shown that ASA has finite expected regret when compared with the optimal centralized scheme with fixed channel allocation.