Distributed Load Balancing Research Articles

Распределенный алгоритм отображения распределенных многомерных данных на многомерный мультикомпьютер в системе фрагментированного программирования LuNA

Distributed algorithm with local interactions Patch is presented in the paper. Patch is intended for data distribution and dynamic load balancing in the LuNA fragmented programming system. LuNA system is used for creation of parallel implementations of large-scale numerical models for distributed memory systems. Execution of a fragmented program is controlled by LuNA run-time system, which uses different data and computation distribution algorithms to enable efficient use of resources and minimize total execution time of the program. Patch algorithm, developed to be used in the LuNA system, enables distribution of multidimensional data meshes on a multidimensional lattice of computational nodes of a supercomputer. The algorithm uses mapping of data to multidimensional lattice of cells (coordinates), which in their turn are mapped to computational nodes. That mapping makes it possible to account for data dependencies and preserve data locality during dynamic load balancing. Patch algorithm was compared with another LuNA data distribution algorithm Rope, fragmented realisation of a real numerical problem was used for experiments. Experiments showed that Patch algorithm provides a general reduction in total computational volume and distances, as compared to Rope algorithm.

A distributed load balancing and admission control algorithm based on Fuzzy type-2 and Game theory for large-scale SaaS cloud architectures

With the advent of cloud computing and virtualization technologies, business models are created to serve users with applications, platforms or processing resources. One of these models is providing software as a service (SaaS) in cloud environments. Due to its user-friendly nature, the demand for SaaS is increasing exponentially in recent years, so geographically distributed cloud infrastructures are required for hosting such services. On the other hand, with the world wide spread of applications, users need to access them all over the world. Common centralized architecture of resources is not responsive and clusters of large-scale distributed servers are required to host applications in different geographical locations. In this paper, a novel two-layer (regional–local) distributed resource management algorithm is proposed which can handle management tasks such as requests admission control and load balancing in large scales. The proposed algorithm is scalable, in which the control tasks are divided among application and proxy servers. Therefore, each component performs the task of managing resources independently. To control the admission of requests, a Self-Adaptive Fuzzy type-2 Controller containing two fuzzy controllers is introduced. Also, for distributing requests and balancing load in the proxy servers, a Game theory based algorithm proposed which converged to Nash equilibrium. The proposed algorithm is implemented and tested using four physical machines. The algorithm is evaluated using two standard dynamic workloads. By deploying this algorithm, the average percentage of rejected requests and LBM value are improved compared to rival algorithms about 15% and 4.5% respectively.

RALBA: a computation-aware load balancing scheduler for cloud computing

Cloud computing serves as a platform for remote users to utilize the heterogeneous resources in data-centers to compute High-Performance Computing jobs. The physical resources are virtualized in Cloud to entertain user services employing Virtual Machines (VMs). Job scheduling is deemed as a quintessential part of Cloud and efficient utilization of VMs by Cloud Service Providers demands an optimal job scheduling heuristic. An ideal scheduling heuristic should be efficient, fair, and starvation-free to produce a reduced makespan with improved resource utilization. However, static heuristics often lead to inefficient and poor resource utilization in the Cloud. An idle and underutilized host machine in Cloud still consumes up to 70% of the energy required by an active machine (Ray, in Indian J Comput Sci Eng 1(4):333–339, 2012). Consequently, it demands a load-balanced distribution of workload to achieve optimal resource utilization in Cloud. Existing Cloud scheduling heuristics such as Min–Min, Max–Min, and Sufferage distribute workloads among VMs based on minimum job completion time that ultimately causes a load imbalance. In this paper, a novel Resource-Aware Load Balancing Algorithm (RALBA) is presented to ensure a balanced distribution of workload based on computation capabilities of VMs. The RABLA framework comprises of two phases: (1) scheduling based on computing capabilities of VMs, and (2) the VM with earliest finish time is selected for jobs mapping. The outcomes of the RALBA have revealed that it provides substantial improvement against traditional heuristics regarding makespan, resource utilization, and throughput.

Load Balancing in Hypergraphs

Consider a simple locally finite hypergraph on a countable vertex set, where each edge represents one unit of load which should be distributed among the vertices defining the edge. An allocation of load is called balanced if load cannot be moved from a vertex to another that is carrying less load. We analyze the properties of balanced allocations of load. We extend the concept of balancedness from finite hypergraphs to their local weak limits in the sense of Benjamini and Schramm (2001) and Aldous and Steele (2004). To do this, we define a notion of unimodularity for hypergraphs which could be considered an extension of unimodularity in graphs. We give a variational formula for the balanced load distribution and, in particular, we characterize it in the special case of unimodular hypergraph Galton Watson processes. Moreover, we prove the convergence of the maximum load under some conditions. Our work is an extension to hypergraphs of Anantharam and Salez (2016), which considered load balancing in graphs, and is aimed at more comprehensively resolving conjectures of Hajek (1990).

Mimir: a term-distributed retrieval system for secret documents

In order to access sensitive documents shared over government, army and enterprise intranets, users rely on an indexing facility where they can quickly locate relevant documents they are allowed to access: 1) without leaking information about the remaining documents; 2) with a balanced load on the index servers. To address this problem, we propose Mimir, a distributed cipher retrieval system for sensitive documents. Mimir constructs the distributed indexes based on load balanced term distribution for better search efficiency and load balanced query. Mimir utilises encryption with random key, partial key update to protect sensitive data and improve query efficiency. Our experiments show that Mimir can effectively protect secret data and answer queries nearly as fast as an ordinary inverted index.

A bacterial foraging-based batch scheduling model for distributed systems

The problem of scheduling in the parallel and distributed environment is proven to be NP-complete and has been addressed by various heuristics. It is always desired from a scheduling scheme to distribute the load evenly on the available resources so as to have maximum resource utilisation while meeting the scheduling objective(s). Bio-inspired metaheuristics for job scheduling have gained immense popularity due to their effectiveness in providing near optimal solution in a reasonable time and computational complexity. This work proposes an evolutionary static scheduling technique based on bacterial foraging for a batch of independent jobs. This model generates the schedule minimising the node idle time and the makespan while exhibiting a balanced load distribution with minimum run time overhead. Simulation study proves the effectiveness of the proposed model in comparison with its peers.

QoS enhanced distributed load balancing and task scheduling framework for wireless networks using hybrid optimisation algorithm

Quality of service (QoS) is the main challenging issue in load balancing, task scheduling process in dynamic systems, such as grid system, peer-to-peer system, ad hoc networks, cloud computing system, pervasive computing system, and online social network system. The inefficient task scheduling process reduces QoS parameters, such as energy consumption, throughput, network lifetime, deadline missing ratio and schedule delay time. In this paper, we proposed QoS enhanced distributed load balancing, task scheduling frameworks for wireless networks (WNs) using hybrid optimisation algorithm (EDFHOA). We improve QoS of distributed frameworks by two phase system through an efficient clustering methodology using particle swarm optimisation (PSO) algorithm for load balancing in the first phase. A cuckoo search (CS) algorithm with load balanced for task scheduling in the second phase. The proposed two-phase hybrid algorithms enhance task scheduling, such as latency experienced by the balanced clustering, energy consumption, throughput, network lifetime, efficient resource utilisation and less processing time requirement and highly accurate than fault tolerant task allocation algorithm (FTAOA).

QoS enhanced distributed load balancing and task scheduling framework for wireless networks using hybrid optimisation algorithm

QoS enhanced distributed load balancing and task scheduling framework for wireless networks using hybrid optimisation algorithm

DFFR: A flow-based approach for distributed load balancing in Data Center Networks

With the increasing need to support high performance and distributed cloud-based computing applications, data centers are employing commodity switches to build multi-rooted trees. An effective distributed, adaptive flow scheduling algorithm is needed to realize the full potential of the multiple parallel paths provided by such networks. The overall aim of this work is to design a load balancer that can maximize the aggregate network utilization. In this paper, the Distributed Flow-by-Flow Fair Routing (DFFR) algorithm is proposed for flow balancing in Data Center Networks. It is a scalable, distributed, and adaptive algorithm designed for maximizing network resources. Our analysis shows that the algorithm has proven theoretical performance bounds, which gives a low variance for the aggregate bandwidth utilization. A simulation study was conducted to compare the performance of the DFFR algorithm with other load balancing algorithms. Our simulation results reveal that the DFFR algorithm outperforms a static routing assignment protocol. It is also compared to Distributed Dynamic Flow Scheduling (DDFS), which is chosen because it is a distributed algorithm like DFFR. DFFR is shown to perform better for random traffic patterns than DDFS, but worse for patterns where hosts always send to the same receiver. The evaluation concludes that the DFFR is an effective load balancer for Data Center Networks with random traffic patterns.

High-resolution multi-code implementation of unsteady Navier–Stokes flow solver based on paralleled overset adaptive mesh refinement and high-order low-dissipation hybrid schemes

ABSTRACTThe low-dissipation high-order accurate hybrid up-winding/central scheme based on fifth-order weighted essentially non-oscillatory (WENO) and sixth-order central schemes, along with the Spalart--Allmaras (SA)-based delayed detached eddy simulation (DDES) turbulence model, and the flow feature-based adaptive mesh refinement (AMR), are implemented into a dual-mesh overset grid infrastructure with parallel computing capabilities, for the purpose of simulating vortex-dominated unsteady detached wake flows with high spatial resolutions. The overset grid assembly (OGA) process based on collection detection theory and implicit hole-cutting algorithm achieves an automatic coupling for the near-body and off-body solvers, and the error-and-try method is used for obtaining a globally balanced load distribution among the composed multiple codes. The results of flows over high Reynolds cylinder and two-bladed helicopter rotor show that the combination of high-order hybrid scheme, advanced turbulence model, and overset adaptive mesh refinement can effectively enhance the spatial resolution for the simulation of turbulent wake eddies.

Load Balancing in Cloud-based Flow Security System

In the context of load balancing [1], Lu et al. proposed a novel class of algorithms called Join-Idle-Queue (JIQ) for distributed load balancing in large systems, where a group of dispatchers distribute jobs to a cluster of parallel servers. Each dispatcher maintains a queue called Idle-Queue for idle servers. When a job arrives to a dispatcher at random, the dispatcher sends it to a server which in the head of the Idle-Queue, or to a random server if the Idle-Queue is empty. Meanwhile conversely, when a server has no jobs, it actively requests to be placed on an Idle-Queue. Although this algorithm was shown to be quite effective [2,3], it is defective in some respects. This paper presents several new improvements, which makes the algorithm more effective, especially in Cloud-based Flow Security System (CFS). These improvements focus primarily on these aspects, such as the definition of idle server, data concurrency while choosing an Idle-Queue, etc. The experimental results show that the improved load balancing algorithm can improve the utilization of resources in a certain extent.

Sistem Load Balancing Menggunakan Least Time First Byte dan Multi Agent System

Network activity has increased in every year due to rapid growth of internet users. This phenomenon eventually increases the load server. The high load on server makes server down. The proposed system to handle the issue is using Least Time First-Byte algorithm and multi-agent system in distributed load balancing. The agent collects information resources on the backend servers and communicates with the agents. The Least Time First-Byte algorithm is then combined with the information resources from the agent, called as Least Time First Byte with Multi Agent System (LFB-MAS). The simulation results show that LFB-MAS performs load balance efficiently to all server and provides better performance. The LFB-MAS can process 100% from the 1,800 requests, whereas WLC algorithm is only capable of processing 74.50% from 1,800 requests and LFB without agent is only capable of processing 75.61% from 1,800 requests. The results prove that LFB-MAS can handle high tasks or requests and is reliable.

A New Effort for Distributed Load Balancing

A New Effort for Distributed Load Balancing

Disc Rehydration after Dynamic Stabilization: A Report of 59 Cases.

Study DesignA retrospective study investigating decrease in the nucleus pulposus signal intensity or disc height on magnetic resonance imaging (MRI) and disc degeneration.PurposeAlthough a degenerated disc cannot self-regenerate, distraction or stabilization may provide suitable conditions for rehydration and possible regeneration. This study aimed to evaluate clinical outcomes and disc regeneration via MRI in a series of patients with degenerative disc disease (DDD) who underwent lumbar stabilization with a dynamic stabilization system (DSS).Overview of LiteratureA dynamic system provides rehydration during early DDD.MethodsFifty-nine patients (mean age, 46.5 years) who undedwent stabilization with DSS for segmental instability (painful black disc) between 2004 and 2014 were retrospectively evaluated. All patients underwent MRI preoperatively and 12 months postoperatively. Intervertebral disc (IVD) degeneration grades at the implanted segment were categorized using the Pfirrmann classification system. Patients were followed for a mean of 6.4 years, and clinical outcomes were based on visual analog scale (VAS) and Oswestry disability index (ODI) scores.ResultsSignificant improvements in back pain VAS and ODI scores from before surgery (7 and 68%, respectively) were reported at 6 (2.85 and 27.4%, respectively) and 12 months postoperatively (1.8 and 16.3%, respectively). Postoperative IVD changes were observed in 28 patients. Improvement was observed in 20 patients (34%), whereas progressive degeneration was observed in eight patients (13.5%). Thirty-one patients (52.5%) exhibited neither improvement nor progression. Single Pfirrmann grade improvements were observed in 29% of the patients and two-grade improvements were observed in 5%.ConclusionsOur observations support the theory that physiological movement and a balanced load distribution are necessary for disc regeneration. We conclude that DSS may decelerate the degeneration process and appears to facilitate regeneration.

Towards Long-View Computing Load Balancing in Cluster Storage Systems

In large-scale computing clusters, when the server storing a task's requested data does not have sufficient computing capacity for the task, current job schedulers either schedule the task to the closest server and transmit to it the requested data, or let the task wait until the server has sufficient computing capacity. The former solution generates network load while the latter solution increases task delay. To handle this problem, load balancing methods are needed to reduce the number of overloaded servers due to computing workloads. However, current load balancing methods do not aim to balance the computing load for the long term. Through trace analysis, we demonstrate the diversity of computing workloads of different tasks and the necessity of balancing the computing workloads among servers. Then, we propose a cost-efficient Computing load Aware and Long-View load balancing approach ( CALV ). CALV is novel in that it achieves long-term computing load balance by migrating out an overloaded server data blocks contributing more computing workloads when the server is more overloaded and contribute less computing workloads when the server is more underloaded at different epochs during a time period. Based upon the task schedules, we further propose a task reassignment algorithm that reassigns tasks from an overloaded server to other data servers of the tasks to make it non-overloaded before CALV is conducted. The above methods are for the tasks whose submission times and execution latencies can be predicted. To handle unexpected tasks or insufficiently accurate predictions, we propose a dynamic load balancing method, in which an overloaded server dynamically redirects tasks to other data servers of the tasks, or replicates the tasks’ requested data to other servers and redirects the tasks to those servers in order to become non-overloaded. Finally, we propose a proximity-aware tree based distributed load balancing method to reduce the reallocation cost and improve the scalability of CALV . Trace-driven experiments in simulation and a real computing cluster show that CALV outperforms other methods in terms of balancing the computing workloads and cost efficiency.

Non-cooperative power and latency aware load balancing in distributed data centers

In this paper we propose an algorithm for load balancing in distributed data centers based on game theory. We model the load balancing problem as a non-cooperative game among the front-end proxy servers. We model the operating cost associated with a data center as a weighted linear combination of the energy cost and the latency cost. We propose a non-cooperative load balancing game with the objective of minimizing the operating cost and obtain the structure of Nash equilibrium. Based on this structure, a distributed load balancing algorithm is designed. We compare the performance of the proposed algorithm with the existing approaches. Numerical results demonstrate that the solution achieved by the proposed algorithm approximates the global optimal solution in terms of the cost and it also ensures fairness among the users.

Model of load balancing using reliable algorithm with multi-agent system

Massive technology development is linear with the growth of internet users which increase network traffic activity. It also increases load of the system. The usage of reliable algorithm and mobile agent in distributed load balancing is a viable solution to handle the load issue on a large-scale system. Mobile agent works to collect resource information and can migrate according to given task. We propose reliable load balancing algorithm using least time first byte (LFB) combined with information from the mobile agent. In system overview, the methodology consisted of defining identification system, specification requirements, network topology and design system infrastructure. The simulation method for simulated system was using 1800 request for 10 s from the user to the server and taking the data for analysis. Software simulation was based on Apache Jmeter by observing response time and reliability of each server and then compared it with existing method. Results of performed simulation show that the LFB method with mobile agent can perform load balancing with efficient systems to all backend server without bottleneck, low risk of server overload, and reliable.

Grid Self-Load-Balancing: the Agent Process Paradigm

Load balancing aims to exploit networked resources equitably in such a way that no resources are overloaded while others are under-loaded or idle. Many approaches have been proposed and implemented, but as new infrastructures emerge like grids and Global Computing (GC), new challenges are raised with regard to network latency. The location policy, as one of the main fundamentals of load balancing solutions, aims to locate overloaded and under-loaded nodes in a network. To do so, multiple communication messages are sent across the network. This technique wastes network resources and causes remarkable network delays in environments like GC, which makes it impractical. In this paper, we propose a new paradigm for adaptive distributed load balancing inspired by swarm intelligence and multi-agent systems. In such a paradigm, no load balancing service would be required. In fact, work tasks are self-aware and capable of self-load-balancing over an unknown-load network. By its nature and based on stigmergy mechanisms, communication frequency of the proposed paradigm is significantly reduced compared to existing solutions. The present work explains the fundamentals of this paradigm, coined the Agent Process Paradigm (APP), as well as its underlying algorithms. Results of performance evaluation are presented and discussed at the end of this paper.

On load balancing strategies for baggage screening at airports

We study load balancing policies for an airport baggage handling system (BHS). The performance of the screening subsystem is important as it can potentially cause traffic congestion of the whole BHS during peak period. Currently, the round-robin (RR) and the first-available (FA) polices are implemented at airports. This paper presents a simulative approach to evaluate the impact of load balancing policies on the system performance. Using discrete-event simulations, airport practitioners can assess the effectiveness of the policies according to the actual layout of a particular passenger terminal. In addition to the RR and FA, a join-shortest-queue (JSQ) policy is introduced. This scheme can be applied jointly with the RR for daily operations. A case study is presented to illustrate the characteristics of this scheme. Simulation results demonstrate that the RR-JSQ can replace the existing RR-FA for a better balanced load distribution and improve the overall system performance.

Flow distribution-aware load balancing for the datacenter

Leveraging centralized and distributed load balancing algorithm to jointly schedule long and short flows attracts much attention since by classifying flows, they could reduce the completion times of flows without losing the scalability. However, existing centralized-distributed joint algorithms generally classify flows according to a static threshold and are unable to adapt to traffic dynamics. More importantly, through over investigation, the overheads of the controller caused by classifying flows and handling flows can be further reduced.In this paper, we present a Flow Distribution-Aware Load Balancing (FDALB) mechanism to reduce flow completion times and achieve high scalability. In FDALB, flows are split into short flows and long flows according to a threshold. The traffic of short flows and long flows are balanced by distributed and centralized algorithms respectively. To adapt to traffic dynamics, we propose a simple yet effective scheme adaptively adjusting the splitting threshold. To reduce the overheads of classifying flows, FDALB leverages end-hosts to tag long flows, which requires no changes in networking hardware. To further reduce the overheads of handling flows, we proposed a new centralized algorithm without requiring flow rates information. Using realistic datacenter workloads, we show that FDALB reduces the average FCT of flows by up to 47% over ECMP while achieves higher scalability than the state-of-art load balancing mechanism Mahout.