Towards realistic benchmarks for virtual infrastructure resource allocators

Cloud computing has emerged as a prevailing platform for internet service hosting. To best utilize Cloud resources for profit making, Cloud providers rely on intelligent resource allocation algorithms when provisioning the virtualized environments for tenant service hosting. Conventional resource allocation proposals mainly focus on efficient allocation of the computing and storage resources, with little effort on ensuring the network performance of tenant services. To address this issue, a number of recent efforts abstract tenant services in the form of virtual infrastructures for resource allocation. A virtual infrastructure specifies the tenant's demand of both the computing resources for hosting virtual servers, and the network bandwidth for inter-virtual server communications. With the problem of resource allocation for virtual infrastructures being NP-hard in general networks, heuristic algorithms have been proposed for this problem. In this paper, we propose a novel optimization technique, named sequential rounding, to tackle the resource allocation problem for virtual infrastructures. The proposed technique extends the rounding technique used for the traditional virtual network embedding problem, while minimizing mapping conflicts introduced by the virtual infrastructure embed- ding problem. Experiments show that our proposed algorithm outperforms existing algorithms regarding both the acceptance ratio and average embedding cost of virtual requests.

In the context of 5G network slicing, in order to solve the problem of low service reliability, a high-reliability virtual network resource allocation algorithm based on service priority is proposed in this paper. The algorithm includes four steps: priority ranking of virtual network request and virtual nodes, reliability ranking of underlying nodes, resource allocation for virtual nodes, and resource allocation for virtual links. The priority ranking of virtual network request and virtual nodes mainly include two processes of ordering the virtual network based on service priority and ordering the virtual node based on centrality. In the step of allocating resources to virtual nodes, the main work is to allocate resources to virtual nodes based on the resources and reliability of the underlying nodes. In the simulation experiment, the algorithm in this paper is compared with the traditional algorithm, which verifies that the algorithm in this paper has achieved good results in terms of key business reliability indicators.

In the network slicing environment, how to improve the success rate of virtual network resource allocation has become an urgent problem to be solved. To solve this problem, this paper proposes a virtual network resource allocation algorithm based on the coefficient of variation and the order relationship under network slicing. The algorithm includes four steps: sorting the importance of the attributes of the underlying nodes, sorting the importance of the attributes of the virtual nodes, calculating the comprehensive attribute values of the underlying nodes and sorting them in descending order, fetching virtual network requests one by one and performing resource allocation. To improve the success rate of resource allocation, first allocate resources for important virtual nodes. When calculating the importance of a virtual node, first calculate the attribute value of the virtual node. After the attribute value is normalized by the min-max method, the comprehensive attribute value of each virtual node is calculated according to the weight and arranged in descending order. When allocating resources to virtual nodes one by one, nodes with large attribute values are preferentially allocated. In the performance analysis part, the algorithm of this paper is compared with the traditional algorithm, and it is verified that the algorithm of this paper has achieved good results in the two dimensions of the success rate of virtual network resource allocation and the utilization of underlying network resources.KeywordsNetwork slicingResource allocationCoefficient of variationOrder relationship

In the 5G era, heterogeneous networks (Het-Nets) are designed for achieving data rates and customized service demands. To realize this, virtualization technologies are widely accepted as enablers for implementing 5G HetNets, aimed at managing and scheduling virtualized physical resources in a flexible manner. However, the major focus of the existing research lies in the effective allocation of virtual resources and maximizing the number of implemented network services, ignoring the virtual resource security issues. However, the security threats and vulnerabilities due to the complexity of virtualization can lead to major performance outbreaks and information leakage. Therefore, this article attempts to tackle the security issues in 5G HetNets virtual resource allocation. The article starts from modeling the major security attacks for virtual resource allocation through comprehensive discussion on the typical types of security attacks. Following the attack model, a novel secure framework (VRA-RL-SecAwa) based on the emerging reinforcement learning approach, is presented. The proposed VRA-RL-SecAwa framework works in different phases, 1. Reinforcement-learning-based preliminary security preparation 2. Greedy-approach-based secure virtual node resource allocation embedding 3. Secure and shortest path virtual link resource allocation scheme 4. Network reconfiguration and update The proposed VRA-RL-SecAwa framework is evaluated through extensive simulations in order to demonstrate its efficiency and effectiveness. The results obtained validate the superiority of the proposed framework in contrast to existing variants of its category.

In this work, two novel virtual user association and resource allocation algorithms are introduced for a wireless virtualized heterogeneous network with hybrid energy supply. In the considered system, macro base stations (MBSs) are supplied by the grid power and small base stations (SBSs) have the energy harvesting capability in addition to the grid power supplement. Multiple infrastructure providers (InPs) own the physical resources, i.e., BSs and radio resources. The Mobile Virtual Network Operators (MVNOs) are able to recent these resources from the InPs and operate the virtualized resources for providing services to different users. In particular, aiming to maximize the overall utility for the MVNOs, a joint resource (spectrum and power) allocation and user association problem is presented. First, we present an alternating direction method of multipliers (ADMM)-based algorithm solution to find the near-optimal solution in a static manner. Moreover, we also utilize deep reinforcement learning to design the optimal policy without knowing a priori knowledge of the dynamic nature of networks. We have conducted extensive simulation and the performance evaluation demonstrate the advantages and effectiveness of the proposed schemes.

This paper focuses on the design of multiple dependable virtual infrastructures (VIs) over converged optical network and IT resources supporting uncertain elastic cloud services. VI dependability is defined as the VI ability to offer resilience in the case of IT resource failures and security guarantees in terms of isolation of VIs that share a common physical infrastructure. The VI planning optimization objective is to minimize the overall energy consumption. To address these issues together with the time variability and uncertainty of elastic cloud services, a novel VI planning scheme based on stochastic linear programming (SLP) is proposed. To cope with the increasing computational complexity inherently imposed in SLP formulations, the sample average approximation technique and Lagrangian relaxation are used to solve the SLP problem in an efficient manner. Our modeling results illustrate interesting trade-offs between the overall energy consumption of the VIs and the dependability impact on the energy consumption and the utilization of the infrastructure resources.

In order to solve the problem of cross-domain resource allocation in IoT in Smart Grid, this paper proposes a virtual network resource allocation mechanism based on particle swarm algorithm. Its goal is to minimize mapping overhead under planning request. In this paper, VN request is divided into multiple virtual subnets according to matching set of virtual resource matching phase, virtual library resource type price information and border node information. We also propose a cross-domain virtual network mapping algorithm based on particle swarm optimization. It can be used to improve the efficiency of cross-domain virtual network mapping. Finally, the execution time, mapping cost and performance of the algorithm in different environments are tested by simulation experiment, which verifies its efficiency and stability performance in virtual network partitioning.

An online auction mechanism for cloud computing resource allocation and pricing based on user evaluation and cost

The cloud computing paradigm consolidated the on-demand provisioning of virtual resources. However, the diversity of services, prices, data centers, and geographical footprints, have turned the clouds into a complex and heterogeneous environment. There are several IaaS providers differentiated by the provisioning costs and availability figures. Due to management complexity, the survivability and reliability aspects are often disregarded by tenants, eventually resulting in heavy losses due to unavailability of services that are hosted on Virtual Infrastructures (VIs). We present an alternative to improve VIs survivability and reliability, which considers the use of replicas and the spreading of virtual resources atop providers, regions, and zones. Replicas are used to achieve a user-defined reliability level while the controlled spreading of VI components decrease the probability of full outages. In addition, our proposal performs a cost-effective allocation. We formulate the VI allocation with survivability and reliability requirements as a Mixed Integer Program (MIP). Following, binary constraints are relaxed to obtain a Linear Program (LP). The LP solution is given as input for the simulated annealing technique, composing the Reliable and Survivable Virtual Infrastructure Allocation (RS-VIA) mechanism. Simulation results with different reliability requests indicate an increasing in survivability without inflating costs.

Improved resource utilization and fault tolerance of large-scale HPC systems can be achieved through fine-grained, intelligent, and dynamic resource (re)allocation. We explore components and enabling technologies applicable to creating a system to provide this capability: specifically 1) Scalable fine-grained monitoring and analysis to inform resource allocation decisions, 2) Virtualization to enable dynamic reconfiguration, 3) Resource management for the combined physical and virtual resources and 4) Orchestration of the allocation, evaluation, and balancing of resources in a dynamic environment. We discuss both general and HPC-centric issues that impact the design of such a system. Finally, we present our prototype system, giving both design details and examples of its application in real-world scenarios.

Network virtualization introduces two concepts: slice (i.e., virtual network), which consists of virtual nodes and links, and slice developer, which is the third role in networks. Slice developers can introduce new network services by using slices. A method for introducing new types of virtual nodes and links for new services into the slice-definition language and the virtualization infrastructure by evolving physical nodes (i.e., “virtualization nodes” or VNodes) is proposed. This evolution consists of two stages: the experimental stage and the operational stage. In the experimental stage, data and control plug-ins are developed and tested by the operator or vendor by using experimental sliver definitions. In the operational stage, which is focused on in this study, the plug-ins are integrated into original components in the infrastructure and are available for slice development by using normal sliver definitions. By mapping type names to plug-in identifiers and parameters, the proposed method enables abstract and simple definitions of slices by slice developers and authorization of plug-ins by the operator, but it remains the loose integration of the new function, i.e., the plug-in architecture used in the experimental stage. Prototyping and evaluation demonstrates that this method greatly simplifies both slice developers' tasks and operators' tasks.

Basic computing resources are compute, storage and network. Virtualisation in data centres has significantly reduced the provisioning time to deploy computing resources from weeks to hours or even minutes [1]. However, data centre virtualisation in itself is not enough for building a cloud-based infrastructure. Without software to manage virtual infrastructure, it will result in virtual resource sprawl that can quickly use up the underlying physical infrastructure. Moreover, cloud-based infrastructure is built not only on virtual infrastructure, but also on physical infrastructure which also needs to be managed and automated. Therefore, there is a need to have software for managing and automating cloud-based infrastructure both virtual and physical resources. In this paper, the main modules of cloud management and automation software and the interaction between those modules are explained and discussed.

In the few last years, network virtualization concept has attracted a great deal of interest from both industry and research communities as an important enabler for designing the future Internet architecture. It has emerged as a powerful way to deploy different network architectures and protocols over a shared physical infrastructure. A major challenge associated with the deployment of network virtualization in operator infrastructure is how to describe efficiently virtual resources needed for one or many virtualization services. This paper proposes a virtual resources provisioning schema (VN-SLA) that specifies virtual resources properties and associations. Besides, it regulates agreements and interactions between virtualization network actors. We give the specification model of the proposed VN-SLA schema and we illustrate it with an example.

Lab-grown diamond heat spreaders are becoming attractive solutions compared to traditional copper heat spreaders due to their high thermal conductivity, the ability to directly bond them on silicon, and allow for an ultra-thin silicon layer. Researchers have developed various thermal models and prototypes of lab-grown diamond heat spreaders to evaluate their cooling performance and heat spreading ability. The majority of existing thermal models are built using finite-element method (FEM) based simulators such as COMSOL and ANSYS. However, such commercial simulators are computationally expensive and lead to long solution times along with large memory requirements. These limitations make commercial simulators unsuitable for evaluating numerous design alternatives or runtime scenarios for real-world high-performance processors. Because of this modeling challenge, none of the existing works have evaluated the thermal behavior of lab-grown diamond heat spreaders on real-world high-performance processors running realistic application benchmarks. Recently, we have developed a parallel compact thermal simulator, PACT, that is able to carry out fast and accurate steady-state and transient thermal simulations and can be extended to support emerging integration and cooling technologies. In this paper, we use PACT to evaluate the steady-state and transient cooling performance of lab-grown diamond heat spreaders against traditional copper heat spreaders on various real-world high-performance processors (e.g., Intel i7 6950X, IBM Power9, and PicoSoC). By using PACT with architectural performance and power simulators such as Sniper and McPAT, we are able to run transient simulations with realistic benchmarks. Simulation results show that lab-grown diamond heat spreaders achieve maximum temperature and thermal gradient reductions of up to 26.73 °C and 13.75 °C when compared to traditional copper heat spreaders, respectively. The maximum steady-state and transient simulation times of PACT for the real-world high-performance chips and realistic applications used in our experiments are 259 s and 22 min, respectively.

In recent cloud computing research, the allocation of virtual resources has become a major concern. The Virtual Machine (VM) dynamic deployment technique is one of the key virtual resource management techniques. Virtual resource management predicts the resources on the virtual machine and manages accordingly. The study of the Virtual Machine resource dynamic deployment focuses on the fine-grained resource adjustment strategy. So, there is a need for predicting the usage of resources for the virtual machine to avoid delays, overprovisioning and under provisioning. Time series forecasting models are used because predictions of cloud resources are dependent on time. In this paper, different time series models and ensemble techniques have been fitted to predict resource usage for successive hours. Different time series analysis techniques were implemented to study the dataset and to choose the appropriate model. Various types of resources such as CPU usage, disk read times, disk write times and memory usage are predicted using the above models.