基于预测技术的动态云服务配置和自动扩展工作负载模式

56 浏览量更新于2024-08-26 收藏 438KB PDF 举报

"基于预测的动态云服务配置和自动扩展的工作负载模式" 在云计算领域中，云服务提供商需要根据性能的可靠性和其较低级别平台基础结构的可用性来协商其提供的客户服务的SLA。然而，性能管理不太可靠，迫切需要一种准确而有效的解决方案来支持初始静态基础结构配置以及动态重新配置和自动缩放的迭代方法。为了解决这个问题，我们提出了一种基于预测的技术，该技术将模式匹配方法与传统的协作过滤解决方案相结合，可以满足准确性和效率要求。这种技术可以抽象出常见的基础结构工作量，并在考虑更复杂的传统方法之前充当第一阶段高性能配置机制的一部分。该技术的主要思想是使用模式匹配方法来抽象出服务工作量模式，从监视日志中提取出常见的基础结构工作量，并将其与传统的协作过滤解决方案相结合，以满足准确性和效率要求。这种方法可以增强当前基于React式规则的可伸缩性方法和基于例如指数平滑的基本预测技术。在实现中，我们可以使用机器学习算法来对监视日志进行分析，并将其与模式匹配方法相结合，以抽象出服务工作量模式。然后，我们可以使用协作过滤解决方案来对服务工作量模式进行优化，以满足准确性和效率要求。本文提出了一种基于预测的技术，用于支持动态云服务配置和自动扩展的工作负载模式。这种技术可以满足准确性和效率要求，并增强当前基于React式规则的可伸缩性方法和基于例如指数平滑的基本预测技术。知识点： 1. 云服务提供商根据性能的可靠性和其较低级别平台基础结构的可用性来协商其提供的客户服务的SLA。 2. 性能管理不太可靠，迫切需要一种准确而有效的解决方案来支持初始静态基础结构配置以及动态重新配置和自动缩放的迭代方法。 3. 基于预测的技术可以满足准确性和效率要求，并增强当前基于React式规则的可伸缩性方法和基于例如指数平滑的基本预测技术。 4. 模式匹配方法可以抽象出服务工作量模式，从监视日志中提取出常见的基础结构工作量，并将其与传统的协作过滤解决方案相结合。 5. 机器学习算法可以用于对监视日志进行分析，并将其与模式匹配方法相结合，以抽象出服务工作量模式。 6. 协作过滤解决方案可以用于对服务工作量模式进行优化，以满足准确性和效率要求。标签： Quality of Service; Cloud Configuration; Autoscaling

Workload Patterns for Quality-driven Dynamic Cloud Service Configuration and

Auto-Scaling

Li Zhang, Yichuan Zhang

Software College

Northeastern University

Shenyang, China

{zhangl,zhangyc}@swc.neu.edu.cn

Pooyan Jamshidi, Lei Xu, Claus Pahl

IC4 / School of Computing,

Dublin City University

Dublin, Ireland

{pjamshidi,lxu,cpahl}@computing.dcu.ie

Abstract— Cloud service providers negotiate SLAs for cus-

tomer services they offer based on the reliability of performance

and availability of their lower-level platform infrastructure.

While availability management is more mature, performance

management is less reliable. In order to support an iterative

approach that supports the initial static infrastructure configura-

tion as well as dynamic reconfiguration and auto-scaling, an

accurate and efficient solution is required. We propose a predic-

tion-based technique that combines a pattern matching approach

with a traditional collaborative filtering solution to meet the

accuracy and efficiency requirements. Service workload patterns

abstract common infrastructure workloads from monitoring logs

and act as a part of a first-stage high-performant configuration

mechanism before more complex traditional methods are consid-

ered. This enhances current reactive rule-based scalability ap-

proaches and basic prediction techniques based on for example

exponential smoothing.

Keywords-Quality of Service, Cloud Configuration, Auto-

scaling, Web and Cloud Services, QoS Prediction, Workload Pat-

tern Mining, Collaborative Filtering.

I. INTRODUCTION

Quality of Service (QoS) is the basis of web and cloud

service configuration management and deployment [1,2].

Cloud service providers (CSPs) – whether at infrastructure,

platform or software level – provide quality guarantees usually

in terms of availability and performance to their customers in

the form of service-level agreements (SLAs) [4]. Internally, the

respective service configuration in terms of available resources

then needs to make sure that the SLA obligations are met [10].

To facilitate SLA conformance, virtual machines (VMs) can be

configured and scaled up/down in terms of CPU cores and

memory, deployed with storage and network capabilities.

Some current cloud infrastructure solutions allow users to de-

fine rules manually to scale up or down to maintain perfor-

mance levels.

QoS like service performance in terms of response time or

availability may vary depending on network, service execution

environment and user requirements, making it hard for

providers to choose an initial configuration and scale this

up/down to maintain the SLA guarantees, but also optimising

resource utilisation at the same time. We utilise QoS prediction

techniques here, but rather than bottom-up predicting QoS

from monitored infrastructure metrics [12,13,25], we reverse

the idea, resulting in a novel technique for pattern-based

resource configuration. We extract service workload patterns

(SWPs) that correspond to typical workloads of the

infrastructure and map these to QoS values. A pattern consists

of narrow range of metrics measured for each infrastructure

concern such as compute, storage and network under which the

QoS concern is stable. In a top-down approach, we then take a

QoS requirement and determine suitable workload-oriented

configurations that maintain required values. Furthermore, we

enhance this with a cost-based selection function, applicable if

many candidate configurations emerge.

We specifically look at performance as the QoS concern

here since dealing with availability in cloud environments is

considered as easier to achieve, but performance is currently

neglected in practice due to less mature resource management

techniques [10]. We introduce pattern detection mechanisms

and, based on a QoS-SWP matrix, we define SWP workload

configurations for required QoS. The accuracy of the solution

to guarantee that the chosen (initially predicted) resource

configurations meet the QoS requirements is of utmost

importance. An appropriate scaling approach is required in

order to allow this to be utilised in dynamic environments. In

this paper, we show that the pattern-based approach improves

the efficiency of the solution in comparison with traditional

prediction approaches, e.g. based on collaborative filtering.

This enhance existing solutions by automating current manual

rule-based reactive scalability mechanisms and also advances

prediction approaches for QoS, making them applicable in the

cloud with its accuracy and performance requirements.

Section II outlines the solution and justifies its practical

relevance. Section III introduces SWPs and how they can be

derived. Section IV discusses the selection of patterns as

workload specifications for resource configuration. The

application of the solution for SLA-compliant cloud resource

configuration is described in Section V. Section VI contains an

evaluation in terms of accuracy and performance of the

solution and Section VII contains a discussion of related work.

II. A

PPROACH OUTLINE -QUALITY-DRIVEN

CONFIGURATION AND SCALING

We now briefly discuss the state-of-the-art in cloud

resource configuration and its relevance to the solution . An

SLA is typically defined based on availability. Customers

expect that the services they acquire will be always available.

Thus, providers usually make extensive claims here. The

consensus in the industry is that cloud computing providers

generally have solutions to manage availability. Response time

2014 IEEE/ACM 7th International Conference on Utility and Cloud Computing

DOI

156

下载后可阅读完整内容，剩余9页未读，立即下载

weixin_38733367

粉丝: 3
资源: 906

基于预测技术的动态云服务配置和自动扩展工作负载模式

云资源管理：QoS驱动的工作负载模式预测与优化

IBM Tivoli工作负载自动化在云计算环境中的应用

公有云服务：塑造企业战略转型的驱动力

Spring云服务货币

eventest:AWS上的事件驱动的无服务器负载测试解决方案

2021中国基础云服务行业发展洞察

Windows Azure 动态模板：灵活的工作负载与虚拟机迁移

KNN算法驱动的云计算集群智能负载均衡策略

BloodPortal：MS课程云服务开源项目

Java云服务中的消息队列与事件驱动架构

最新资源