Asymptotically Optimal Load Balancing in Large-scale Heterogeneous Systems with Multiple Dispatchers

Nowadays, the efficiency and even the feasibility of traditional load-balancing policies are challenged by the rapid growth of cloud infrastructure and the increasing levels of server heterogeneity. In such heterogeneous systems with many load-balancers, traditional solutions, such as $JSQ$ , incur a prohibitively large communication overhead and detrimental incast effects due to herd behavior. Alternative low-communication policies, such as $JSQ(d)$ and the recently proposed $JIQ$ , are either unstable or provide poor performance. We introduce the Local Shortest Queue ( $LSQ$ ) family of load balancing algorithms. In these algorithms, each dispatcher maintains its own, local, and possibly outdated view of the server queue lengths, and keeps using $JSQ$ on its local view. A small communication overhead is used infrequently to update this local view. We formally prove that as long as the error in these local estimates of the server queue lengths is bounded in expectation , the entire system is strongly stable. Finally, in simulations, we show how simple and stable $LSQ$ policies exhibit appealing performance and significantly outperform existing low-communication policies, while using an equivalent communication budget. In particular, our simple policies often outperform even $JSQ$ due to their reduction of herd behavior. We further show how, by relying on smart servers ( i.e., advanced pull-based communication), we can further improve performance and lower communication overhead.

In cloud computing, a lot of challenges like the server failures, loss of confidentiality, improper workloads, etc. are still bounding the efficiency of cloud systems in real-world scenarios. For this reason, many research works are being performed to overcome the shortcoming of existing systems. Among them, load balancing seems to be the most critical issue that worsen the performance of the cloud sector, and hence there necessitates the optimal load balancing with optimal task scheduling. With the intention of accomplishing optimal load balancing by effectual task deployment, this paper plans to develop an advanced load balancing model with the assistance acquired from the metaheuristic algorithms. Usually, handling of tasks in cloud system is an NP-hard problem and moreover, nonpreemptive independent tasks are crucial in cloud computing. This paper goes with the introduction of a new optimal load balancing model by considering three major objectives: minimum makespan, priority, and load balancing, respectively. Moreover, a new single-objective function is also defined that incorporates all the three objectives mentioned above. Furthermore, the deployment of tasks must be optimal and for this a new hybrid optimization algorithm referred as Firefly Movement insisted WOA (FM-WOA) is introduced. This FM-WOA is the conceptual amalgamation of standard Whale Optimization Algorithm (WOA) and Firefly (FF) algorithm. Finally, the performances of the proposed FM-WOA model is compared over the conventional models with the intention of proving its efficiency in terms of makespan, task completion (priority), and degree of imbalance as well.

In recent years, cloud computing provides a spectacular platform for numerous users with persistent and alternative varying requirements. In the cloud environment, security and service availability are the two most significant factors during the data encryption process. For providing optimal service availability, it is necessary to establish a load balancing technique that is capable of balancing the request from diverse nodes present in the cloud. This paper aims in establishing a dynamic load balancing technique using the APMG approach. Here in this paper, we integrated adaptive neuro-fuzzy interference system-polynomial neural network as well as memory-based grey wolf optimization algorithm for optimal load balancing. The memory-based grey wolf optimization algorithm is employed to enhance the precision of ANFIS-PNN and to maximize the locations of the membership functions respectively. Also, two significant factors namely the turnaround time and CPU utilization involved in optimal load balancing scheme are evaluated. Finally, the performance evaluation of the proposed MG-ANFIS based dynamic load balancing approach is compared with various other load balancing approaches to determine the system performances.

: A major attack can significantly reduce the capability to deliver services in large scale networked information systems. In this project, we have addressed the survivability of large-scale heterogeneous information systems which consist of various services provided over multiple interconnected networks with different technologies. The communications network portions of such systems are referred to as multi-networks. We specifically address the issue of survivability due to physical attacks that destroy links and nodes in multi-networks. The end goal is to support critical services in the face of a major attack by making optimum use of network resources while minimizing network congestion. This is an area which is little studied, especially for large-scale heterogeneous systems. In this report, we present an overview of our contribution in this area.

Software Defined Networking (SDN) has emerged as a new paradigm that offers the programmability required to dynamically configure and control a network. In this paper, we present a path assignment method based on load balancing(PA-LB) for a centralized controller to calculate optimal end-to-end virtual paths over the underlying network infrastructure, considering multiple requests and load balance simultaneously. In the paper, theoretical analysis shows that the path selection is NP-hard, and strongly polynomial time inapproximable under the assumption P not equal to NP. We also give a constant-factor polynomial time approximation algorithm for this problem under some bounded parameters. Extensive simulation results, over a wide range of underlying network topologies and input parameters, demonstrate that the proposed method outperforms traditional shortest path first (SPF) approaches not only in paths mapping number, but also in network load balance. Under the condition of limited resources, such as bandwidth, the proposed method has a better performance with the increase of the number of nodes. Moreover, optimal load balance can be kept in the network, in order to avoid network congestion and minimize the average packet transmission latency.

This article deals with randomized allocation processes placing sequentially n balls into n bins. We consider multiple-choice algorithms that choose d locations (bins) for each ball at random, inspect the content of these locations, and then place the ball into one of them, for example, in a location with minimum number of balls. The goal is to achieve a good load balancing. This objective is measured in terms of the maximum load, that is, the maximum number of balls in the same bin.Multiple-choice algorithms have been studied extensively in the past. Previous analyses typically assume that the d locations for each ball are drawn uniformly and independently from the set of all bins. We investigate whether a nonuniform or dependent selection of the d locations of a ball may lead to a better load balancing. Three types of selection, resulting in three classes of algorithms, are distinguished: (1) uniform and independent, (2) nonuniform and independent, and (3) nonuniform and dependent.Our first result shows that the well-studied uniform greedy algorithm (class 1) does not obtain the smallest possible maximum load. In particular, we introduce a nonuniform algorithm (class 2) that obtains a better load balancing. Surprisingly, this algorithm uses an unfair tie-breaking mechanism, called Always-Go-Left, resulting in an asymmetric assignment of the balls to the bins. Our second result is a lower bound showing that a dependent allocation (class 3) cannot yield significant further improvement.Our upper and lower bounds on the maximum load are tight up to additive constants, proving that the Always-Go-Left algorithm achieves an almost optimal load balancing among all sequential multiple-choice algorithm. Furthermore, we show that the results for the Always-Go-Left algorithm can be generalized to allocation processes with more balls than bins and even to infinite processes in which balls are inserted and deleted by an oblivious adversary.

Optimization of load balancing using fuzzy Q-Learning for next generation wireless networks

The issue of optimal load balancing has been raised in several important engineering fields such as the file management in a data base, the job control in a multiple processor system or the flow assignment in a packet switched data network. By modeling the subsystems as multiple GI/G/1 queues and applying known approximations and decomposition methods, the system mean response time is expressed as a function of the load allocation. To find the optimal load balancing, the convexity property is first studied with respect to the type of the arrival process of the system. We give then an efficient algorithm which finds for the convex case the global and for the non‐convex case a local minimum. Some important properties of the multiple Gl/G/1 queueing system are revealed with several numerical examples.

In parallel I/O systems, fast response to disk access and load balancing are two important performance objectives pursued by end users and applications. These performances are largely determined by the data allocation strategies, or file assignment algorithms. However, most existing algorithms can only obtain one of the performance objectives, including the well-known algorithms like Greedy, Sort Partition (SP) and Hybrid Partition (HP). New algorithms that can achieve both objectives are very necessary for parallel I/O systems. In this paper, we have proposed two new allocation algorithms for file assignment in parallel I/O systems: an offline Balanced Allocation with Sort (BAS) algorithm and an online Balanced Allocation with Sort for Batch (BASB) algorithm. Both algorithms aim to achieve the optimal mean response time and load balancing at the same time. The experiment results have shown that the BAS algorithm can get the optimal performance on response time among all compared algorithms and better performance on load balancing than SP. For online BASB algorithm, it can achieve the optimal performances on both response time and load balancing among all algorithms for comparison.

In this article, we present a new hybrid algorithm to enable and scale the high-performance conjugate gradients (HPCG) benchmark on large-scale heterogeneous systems such as the Tianhe-2. Based on an inner–outer subdomain partitioning strategy, the data distribution between host and device can be balanced adaptively. The overhead of data movement from both the MPI communication and the PCI-E transfer can be significantly reduced by carefully rearranging and fusing operations. A variety of parallelization and optimization techniques for performance-critical kernels are exploited and analyzed to maximize the performance gain on both host and device. We carry out experiments on both a small heterogeneous computer and the world’s largest one, the Tianhe-2. On the small system, a thorough comparison and analysis has been presented to select from different optimization choices. On Tianhe-2, the optimized implementation scales to the full-system level of 3.12 million heterogeneous cores, with an aggregated performance of 623 Tflop/s and a parallel efficiency of 81.2%.

Graphics Processing Units (GPUs) have been widely used as accelerators in large-scale heterogeneous computing systems. However, current programming models can only support the utilization of local GPUs. When using non-local GPUs, programmers need to explicitly call API functions for data communication across computing nodes. As such, programming GPUs in large-scale computing systems is more challenging than local GPUs since local and remote GPUs have to be dealt with separately. In this work, we propose a virtual OpenCL (VOCL) framework to support the transparent virtualization of GPUs. This framework, based on the OpenCL programming model, exposes physical GPUs as decoupled virtual resources that can be transparently managed independent of the application execution. To reduce the virtualization overhead, we optimize the GPU memory accesses and kernel launches. We also extend the VOCL framework to support live task migration across physical GPUs to achieve load balance and/or quick system maintenance. Our experiment results indicate that VOCL can greatly simplify the task of programming cluster-based GPUs at a reasonable virtualization cost.

In this paper, we consider the problem of power control and load balancing in the uplink of the cell-free (CF) multi-user (MU) massive multiple-input multiple-output system. The power control problem is solved using three different criteria: 1) power minimization; 2) maximize min quality of service (QoS); and 3) maximize sum spectral efficiency (SE) under imperfect channel state information. While power minimization and min-QoS maximization problems can be solved in polynomial time, sum SE maximization is NP-hard. Hence, we apply a successive approach to convert it into a geometric program and achieve near optimum solution. As the number of connections in CF-MU system is large, we minimize the number of base stations (BSs) serving each user while maintaining the optimized power consumption and QoS from the previous stage. We propose an iterative elimination (IE) algorithm to remove ineffective BSs for each user. The system analysis is performed under two common BS receivers: maximum ratio combining and zero forcing. The simulation shows that our method is better than both maximum signal-to-noise ratio association and full-set joint transmission, especially in the high QoS regime. Another notable result is that the selection of the linear receiver becomes more influential than the association method.

This paper reports on the results of a preliminary study in dynamic load balancing on an Intel Hypercube. The purpose of this research is to provide experimental data in how parallel algorithms should be constructed to obtain maximal utilization of a parallel architecture. This study is one aspect of an ongoing research project into the construction of an automated parallelization tool. This tool will take FORTRAN source as input, and construct a parallel algorithm that will produce the same results as the original serial input. The focus of this paper is on the load balancing aspect of that project. The basic idea is to reserve a certain percentage of the computation task, subdivide that percentage into arbitrarily fine tasks, and dole those small tasks out to nodes on request. Ij” the percentage is chosen correctly, then a minority of nodes should be involved in consuming the filler tasks, and the overall throughput of the job should increase as a result of the individual node efJciencies having increased. This paper will outline our approach to performing dynamic load balancing on an Intel iPSC/2. We take the view that the problem of load balancing is really a problem of dividing a “computational task” into smaller components, each of roughly equal complexity, and each an independent event. After this is done, the components of the task can be sent to a node for execution. The key to an optimally balanced load across all computational nodes is the ability to form a statistical profile of the individual components of each computational task. This statistical profile will determine an initial sequence of execution. Our experience indicates that a speedup on the order of 80% is achievable with the judicious use of profiled load balancing. During the process of execution, the initial profile will be altered according to the actual behavior exhibited by the nodes. The difference between the actual and expected performance will be used to determine how much additional time should be devoted to altering the current execution schedule. Currently, our work involves statically setting the load balancing parameters. Our load balancing system determines the execution schedule

Adaptive passenger-finding recommendation system for taxi drivers with load balancing problem

Nowadays, web servers have to respond to the huge amount of requests; therefore, balancing the load between server replicas becomes vital and critical. Conventional load balancers have been dedicated hardware that is susceptible being a single point of failure, vendor-dependent and not flexible. In this work, we propose a flexible and programmable scheduling algorithm that ponders different type of requests, various server processing capacity based upon their CPU, memory usage, I/O rate, and a number of connections. This algorithm uses optimization algorithm, considering not only load of servers but also network congestion. The proposed algorithm includes four main phases. The first two phases have the responsibility of creating sub-topology that causes to increase performance and scalability of the proposed algorithm by reduction of overall solution space. In the third phase, the joint optimal network and server-side load balancing approach are performed, and in the last phase, the post-processing is performed in order to select the server, creates a path, and injects appropriate flow table entries on the OpenFlow switches. The evaluation results show that the dynamic load balancing algorithm significantly decreases response-time whereas increases throughput, in comparison with the existing load balancing algorithms.