G. B. Mund

Performance analysis of cloud with queue-dependent virtual machines

Cloud computing provides a new way for industries to meet the emerging business need for agility. Many public clouds are available for developers to build web applications on cloud. The process of entering into the cloud is generally in the form of a queue, so that each user need to wait until the current user is being served. In the system, each Cloud Computing User (CCU) requests Cloud Computing Service Provider (CCSP) for use of resources. If CCU finds the server busy, then the user has to wait till the current user completes the job. This may result in increase of queue length as well as waiting time, which may lead to request drop. To handle this problem, CCSP needs to find ways to reduce waiting time. We propose a finite multiserver queueing model with queue dependent heterogeneous servers where the web applications are modeled as queues and the virtual machines are modeled as service providers. CCSPs can use multiple servers and the number of busy servers changes depending on the queue length for reducing queue length and waiting time. This helps us to dynamically create and remove virtual machines in order to scaling up and down. We develop a recursive method to obtain the system steady-state probabilities. Various performance measures of the proposed scheme have been described and evaluated. Computational experiences in the form of graphs are presented.

Analysis of discrete-time batch service renewal input queue with multiple working vacations

This paper investigates a discrete-time single server batch service queue with multiple working vacations wherein arrivals occur according to a discrete-time renewal process. The server works with a different service rate rather than completely stopping during the vacation period. The service is performed in batches and the server takes a vacation when the system does not have any waiting customers at a service completion epoch or a vacation completion epoch. We present a recursive method, using the supplementary variable technique to obtain the steady-state queue-length distributions at pre-arrival, arbitrary and outside observers observation epochs. The displacement operator method is used to solve simultaneous non-homogeneous difference equations. Some performance measures and waiting-time distribution in the system have also been discussed. Finally, numerical results showing the effect of model parameters on key performance measures are presented.

Performance Analysis of Cloud Computing Centers for Bulk Services

Cloud is a service oriented platform where all kinds of virtual resources are treated as services to users. Several cloud service providers have offered different capabilities for a variety of market segments over the past few years. The most important aspects of cloud computing are resource scheduling, performance measures, and user requests. Sluggish access to data, applications, and web pages spoils employees and customers alike, as well as cause application crashes and data losses. In this paper, the authors propose an analytical queuing model for performance evaluation of cloud server farms for processing bulk data. Some important performance measures such as mean number of tasks in the queue, blocking probability, and probability of immediate service, and waiting-time distribution in the system have also been discussed. Finally, a variety of numerical results showing the effect of model parameters on key performance measures are presented.

Dynamic Provisioning and Resource Management for Multi-Tier Cloud Based Applications

Abstract Dynamic capacity provisioning is a useful technique for handling the workload variations seen in cloud environment. In this paper, we propose a dynamic provisioning technique for multi-tier applications to allocate resources efficiently using queueing model. It dynamically increases the mean service rate of the virtual machines to avoid congestion in the multi-tier environments. An optimization model to minimize the total number of virtual machines for computing resources in each tier has been presented. Using the supplementary variable and the recursive techniques, we obtain the system-length distributions at pre-arrival and arbitrary epochs. Some important performance indicators such as blocking probability, request waiting time and number of tasks in the system and in the queue have also been investigated. Finally, computational results showing the effect of model parameters on key performance indicators are presented.

Multiserver bulk service discrete-time queue with finite buffer and renewal input

This paper analyzes a discrete-time finite-buffer multi-server bulk-service queueing system in which the interarrival- and service-times are, respectively, arbitrarily and geometrically distributed. Using the supplementary variable and the imbedded Markov-chain techniques, the queue is analyzed for the early arrival system. We obtain state probabilities at prearrival, arbitrary and outside observers observation epochs. Some performance measures, waiting-time distribution in the queue along with some numerical results, and special cases of the model have also been discussed. Finally, it is shown that in the limiting case the results obtained in this paper tend to the continuous-time counterpart.

Optimal Thresholds of an Infinite Buffer Discrete-Time Two-Server System with Triadic Policy

This paper analyzes a discrete-time infinite-buffer Geo/Geo/2 queue, in which the number of servers can be adjusted depending on the number of customers in the system one at a time at arrival or at service completion epoch. Analytical closed-form solutions of the infinite-buffer Geo/Geo/2 queueing system operating under the triadic (0, Q N, M) policy are derived. The total expected cost function is developed to obtain the optimal operating (0, Q N, M) policy and the optimal service rate at minimum cost using direct search method. Some performance measures and sensitivity analysis have been presented.

Computational analysis of multi-server discrete-time queueing system with balking, reneging and synchronous vacations

This paper proposes a discrete-time multi-server queue with multiple synchronous vacations under balking and reneging. Arriving customers decide whether to join the system or balk on the basis of some state-dependent joining/balking probabilities, and renege according to a geometric distribution when servers are busy. The servers take a vacation together if there are no customers in the system at a service completion instant. When the servers are on vacation, an arriving customer activates an impatience timer which is geometrically distributed. The inter-arrival times, service times and vacation times are assumed to be independent and geometrically distributed. We obtain closed-form expressions and develop a computational algorithm for calculating the steady-state probabilities. Specifically, we establish the application of the proposed framework in analyzing a multi-server queueing system with synchronous vacation under balking and reneging. Applications of such models can be found in a wide variety of real-time systems including call centers, computer and communication systems, cloud computing, quality control and maintenance in industrial establishments. We develop a cost model to determine the optimal service rate. Various performance measures and numerical examples are sketched out to demonstrate the impact of the proposed method. Some special cases of the model have also been discussed. Finally, we show that in the limiting case the results converge to the corresponding continuous-time counterparts.

Performance analysis and optimal resource usage in finite population cloud environment

Cloud computing provides a new paradigm for industries to meet the emerging business needs by accessing distributed computing resources such as infrastructure, hardware and software applications on-demand over the internet as services. As the technology and the need is growing very fast, in future there may be multiple vendors offering different services with different Quality of Services (QoS) and at various prices. This would lead to development of new methods and tools for the performance evaluation of the system to meet the offerings and requirements. In this paper, we present an analytical finite population model for performance evaluation of a private cloud computing system. Various performance measures of the cloud system for finite population environment indicate that the proposed provisioning technique helps the cloud operators in tuning the resources accordingly to improve the QoS targets.

Analysis of discrete-time queues with batch renewal input and multiple vacations

This paper analyzes a discrete-time multiple vacations finite-buffer queueing system with batch renewal input in which inter-arrival time of batches are arbitrarily distributed. Service and vacation times are mutually independent and geometrically distributed. The server takes vacations when the system does not have any waiting jobs at a service completion epoch or a vacation completion epoch. The system is analyzed under the assumptions of late arrival system with delayed access and early arrival system. Using the supplementary variable and the imbedded Markov chain techniques, the authors obtain the queue-length distributions at pre-arrival, arbitrary and outside observer’s observation epochs for partial-batch rejection policy. The blocking probability of the first-, an arbitraryand the last-job in a batch have been discussed. The analysis of actual waiting-time distributions measured in slots of the first-, an arbitrary- and the last-job in an accepted batch, and other performance measures along with some numerical results have also been investigated.

Optimal Management of Cloud Centers with Different Arrival Modes for Cloud Computing Environment

Cloud computing is a new computing paradigm in which information and computing services can be accessed from a Web browser by clients. Understanding of the characteristics of computer service performance has become critical for service applications in cloud computing. For the commercial success of this new computing paradigm, the ability to deliver guaranteed Quality of Services QoS is crucial. Based on the Service level agreement, the requests are processed in the cloud centers in different modes. This paper analyzes a finite-buffer multi-server queuing system where client requests have two arrival modes. It is assumed that each arrival mode is serviced by one or more Virtual machines, and both the modes have equal probabilities of receiving service. Various performance measures are obtained and optimal cost policy is presented with numerical results. The genetic algorithm is employed to search the optimal values of various parameters for the system.