Heterogeneous Server Clusters Research Articles

Exploiting Data Locality to Improve Performance of Heterogeneous Server Clusters

We consider load balancing in large-scale heterogeneous server systems in the presence of data locality that imposes constraints on which tasks can be assigned to which servers. The constraints are naturally captured by a bipartite graph between the servers and the dispatchers handling assignments of various arrival flows. When a task arrives, the corresponding dispatcher assigns it to a server with the shortest queue among [Formula: see text] randomly selected servers obeying these constraints. Server processing speeds are heterogeneous, and they depend on the server type. For a broad class of bipartite graphs, we characterize the limit of the appropriately scaled occupancy process, both on the process level and in steady state, as the system size becomes large. Using such a characterization, we show that imposing data locality constraints can significantly improve the performance of heterogeneous systems. This is in stark contrast to either heterogeneous servers in a full flexible system or data locality constraints in systems with homogeneous servers, both of which have been observed to degrade the system performance. Extensive numerical experiments corroborate the theoretical results. Funding: This work was partially supported by the National Science Foundation [CCF. 07/2021–06/2024].

Stalling in Queuing Systems with Heterogeneous Channels

The present paper considers a model of stalling in a queueing system (QS) with any number of different capacities of heterogeneous servers. The state graph and the corresponding linear system for steady-state probabilities are derived using the standard Markov chain technique. The obtained solution of the steady-state probabilities model is numerically stable; the complexity of the corresponding expressions does not depend on the number of QS states; thus, they enable us to analytically study the QS characteristics. Optimization of a stalling buffer is considered as well, and it is shown that stalling helps us to solve the slow server problem under an appropriate choice of stalling buffer size, making the slow servers usable under various values of system load. The asymptotic conditions of optimal query distribution in channels, when the ratio of the capacities of the fast and slow servers is increasing, are also established. Moreover, some applications of the developed model in heterogeneous server clusters and in work productivity modelling for forest harvesting applications are discussed.

The Redundant Energy Consumption Laxity Based Algorithm to Perform Computation Processes for IoT Services

There are various kinds of Internet of Things (IoT) services like home management services. In the IoT services, a large volume of data is gathered from various types of smart devices and is manipulated in scalable, high performance, and fault-tolerant computing systems like cloud computing systems. Reliable and available IoT services can be provided by replicating each application process on multiple virtual machines in a server cluster system. On the other hand, a server cluster system with process replicas consumes a larger amount of electric energy than non-replication approaches. Hence, not only fault-tolerant but also energy-efficient server cluster systems by using virtual machine technologies are required to realize IoT services. In this paper, a redundant energy consumption laxity based (RECLB) algorithm to redundantly and energy-efficiently perform each application process on multiple virtual machines is newly proposed. The RECLB algorithm is evaluated in terms of the total electric energy consumption of a server cluster and the average computation time of each process in both homogeneous and heterogeneous server clusters.

A Virtual Machine Placement Algorithm with Energy-Efficiency in Cloud Computing

Energy efficient virtual machines (VM) management and distribution on cloud platforms is an important research subject. Mapping VMs into PMs (Physical Machines) requires knowing the capacity of each PM and the resource requirements of the VMs. It should also take into accounts of VM operation overheads, the reliability of PMs, Quality of Service (QoS) in addition to energy efficiency. In this article, the authors propose an energy efficient statistical live VM placement scheme in a heterogeneous server cluster. Their scheme supports VM requests scheduling and live migration to minimize the number of active servers in order to save the overall energy in a virtualized server cluster. Specifically, the proposed VM placement scheme incorporates all VM operation overheads in the dynamic migration process. In addition, it considers other important factors in relation to energy consumption and is ready to be extended with more considerations on user demands. The authors conducted extensive evaluations based on HPC jobs in a simulated environment. The results prove the effectiveness of the proposed scheme.

Analysis of Cloud Network Management Using Resource Allocation and Task Scheduling Services

Network failure in cloud datacenter could result from inefficient resource allocation; scheduling and logical segmentation of physical machines (network constraints). This is highly undesirable in Distributed Cloud Computing Networks (DCCNs) running mission critical services. Such failure has been identified in the University of Nigeria datacenter network situated in the south eastern part of Nigeria. In this paper, the architectural decomposition of a proposed DCCN was carried out while exploring its functionalities for grid performance. Virtualization services such as resource allocation and task scheduling were employed in heterogeneous server clusters. The validation of the DCCN performance was carried out using trace files from Riverbed Modeller 17.5 in order to ascertain the influence of virtualization on server resource pool. The QoS metrics considered in the analysis are: the service delay time, resource availability, throughput and utilization. From the validation analysis of the DCCN, the following results were obtained: average throughput (bytes/Sec) for DCCN = 40.00%, DCell = 33.33% and BCube = 26.67%. Average resource availability response for DCCN = 38.46%, DCell = 33.33%, and BCube = 28.21%. DCCN density on resource utilization = 40% (when logically isolated) and 60% (when not logically isolated). From the results, it was concluded that using virtualization in a cloud DataCenter servers will result in enhanced server performance offering lower average wait time even with a higher request rate and longer duration of resource use (service availability). By evaluating these recursive architectural designs for network operations, enterprises ready for Spine and leaf model could further develop their network resource management schemes for optimal performance.

A column generation approach for power-aware optimization of virtualized heterogeneous server clusters

Increasingly, clusters of servers have been deployed in large data centers to support the development and implementation of many kinds of services, having distinct workload demands that vary over time, in a scalable and efficient computing environment. Emerging trends are utility/cloud computing platforms, where many network services, implemented and supported using server virtualization techniques, are hosted on a shared cluster infrastructure of physical servers. The energy consumed to maintain these large server clusters became a very important concern, which in turn, requires major investigation of optimization techniques to improve the energy efficiency of their computing infrastructure.In this work, we propose an efficient approach to solve a relevant cluster optimization problem which, in practice, can be used as an embedded module to implement an integrated power and performance management solution in a real server cluster. The optimization approach simultaneously deals with (i) CPU power-saving techniques combined with server switching on/off mechanisms, (ii) the case of server heterogeneity, (iii) virtualized server environments, (iv) an efficient optimization method, which is based on column generation techniques. The key aspects of our approach are the basis on rigorous and robust optimization techniques, given by high quality solutions in short amount of processing time, and experimental results on the cluster configuration problem for large-scale heterogeneous server clusters that can make use of virtualization techniques.

Power and energy management for server systems

This survey shows that heterogeneous server clusters can be made more efficient by conserving power and energy while exploiting information from the service level, such as request priorities established by service-level agreements.