Optimal Provisioning and Fair Allocation of Resources in Cloud Computing Data Centers

云计算数据中心资源的优化配置和公平分配

• 批准号: RGPIN-2020-07158
• 负责人: Lambadaris, Ioannis
• 金额: $ 2.04万
• 依托单位: Carleton University
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=765863
• 关键词: Optimal; Provisioning; Fair; Allocation; Resources

As computation power, disk storage, and high-end network communication costs plummet, cloud computing has emerged to provide intuitive, utility-style access to vast pools of resources (compute, storage, networking, and software services). Such processing power is relatively affordable and readily available. We can confidently say that cloud computing is defining the Internet! Software Defined Networking (SDN) is a computer network architecture that offers a flexible means for configuring raw hardware networking infrastructures to support a wide range of dynamic service demands with fluctuating resource requirements. Virtualization is at the heart of SDN and is the key technology to move away from dedicated and expensive middleboxes to softwarized solutions. In particular, important network functions (e.g., firewalls, NAT, DNS cashing etc. ) can run on cheap off the shelve middleware (e.g., generic server grade computer motherboards) where they are implemented on virtual machines; such network functions are known as Virtual Network Functions(VNF). They can be orchestrated in a dynamic fashion and  data flows can be processed  through interconnected VNFs called Service Chains (SC).  Network slicing is a technique for resource provisioning in networks that has been recently applied to the mobile cloud; empowered by SDN and network function virtualization (NFV) gives rise to a  dynamic and robust approach for establishing and reconfiguring customized end-to-end logic networks that encapsulate dedicated and shared resources using raw networking infrastructure. Finally, data analytics techniques for processing and identifying trends  of traffic workloads will be crucial for next generation agile and self configurable clouds.  Currently the industry is witnessing a major migration towards SDN enabled cloud networking, which will impact a broad range of sectors including consumer communications/social networking, eCommerce, eHealth and Internet of Things (IoT) applications. As such, the impact of the SDN technology will have a long lasting effect on network users in the global market.    Getting services from the cloud, requires the designers/providers to have significant expertise, experience, and time to customise and configure resources that are assigned to a multitude of (not necessarily co-located) virtual machines/VNFs. Efficient and fair resource allocation and management in such a shared computing system is particularly challenging due to diverse types of available resources, highly fluctuating workload/user needs, heterogeneity in the resource capacities of servers, and placement constraints regarding which servers may be accessible by a pool of users. The goal of the proposed research program is to investigate the fair and efficient resource allocation/management and dynamic network slicing, in conjunction with data analytics for maintaining the quality of offered services in the emerging SDN enabled (mobile) cloud architectures.

随着计算能力、磁盘存储和高端网络通信成本的直线下降，云计算的出现提供了对大量资源（计算、存储、网络和软件服务）的直观、实用的访问。这种处理能力相对便宜并且容易获得。我们可以自信地说，云计算正在定义互联网！软件定义网络 (SDN) 是一种计算机网络架构，它提供了一种灵活的方法来配置原始硬件网络基础设施，以支持各种动态服务需求和不断变化的资源需求。 虚拟化是 SDN 的核心，也是从专用且昂贵的中间盒转向软件化解决方案的关键技术。特别是，重要的网络功能（例如防火墙、NAT、DNS 缓存等）可以在廉价的现成中间件（例如通用服务器级计算机主板）上运行，并在虚拟机上实现；此类网络功能称为虚拟网络功能（VNF）。它们可以以动态方式进行编排，并且可以通过称为服务链 (SC) 的互连 VNF 处理数据流。  网络切片是一种网络资源配置技术，最近已应用于移动云；在 SDN 和网络功能虚拟化 (NFV) 的支持下，产生了一种动态且强大的方法，用于建立和重新配置定制的端到端逻辑网络，该网络使用原始网络基础设施封装专用和共享资源。最后，用于处理和识别流量工作负载趋势的数据分析技术对于下一代敏捷和自配置云至关重要。  目前，业界正在见证向支持 SDN 的云网络的重大迁移，这将影响广泛的行业，包括消费者通信/社交网络、电子商务、电子医疗和物联网 (IoT) 应用。因此，SDN 技术的影响将对全球市场的网络用户产生长期持续的影响。    从云获取服务需要设计者/提供商拥有丰富的专业知识、经验和时间来自定义和配置分配给多个（不一定位于同一位置）虚拟机/VNF 的资源。由于不同类型的可用资源、高度波动的工作负载/用户需求、服务器资源容量的异质性以及关于哪些服务器可以被用户池访问的放置限制，在这种共享计算系统中高效且公平的资源分配和管理尤其具有挑战性。 拟议研究计划的目标是调查公平高效的资源分配/管理和动态网络切片，并结合数据分析来维持新兴 SDN 支持（移动）云架构中所提供服务的质量。