Managing Service Dependency for Cloud Reliability The Industrial Practice Tianyi Yang Baitong Li Jiacheng Shen Yuxin Suy Yongqiang Yangz and Michael R. Lyu

2025-04-27 0 0 235.74KB 2 页 10玖币

侵权投诉

Managing Service Dependency for Cloud

Reliability: The Industrial Practice

Tianyi Yang∗, Baitong Li∗, Jiacheng Shen∗, Yuxin Su†, Yongqiang Yang‡, and Michael R. Lyu∗

∗Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong SAR.

Email: {tyyang, btli, jcshen, lyu}@cse.cuhk.edu.hk

†School of Software Engineering, Sun Yat-Sen Univeristy, Zhuhai, China. Email: suyx35@mail.sysu.edu.cn

‡Computing and Networking Innovation Lab, Huawei Cloud, Shenzhen, China. Email: yangyongqiang@huawei.com

Abstract—Interactions between cloud services result in service

dependencies. Evaluating and managing the cascading impacts

caused by service dependencies is critical to the reliability of

cloud systems. This paper summarizes the dependency types in

cloud systems and demonstrates the design of the Dependency

Management System (DMS), a platform for managing the service

dependencies in the production cloud system. DMS features the

full-lifecycle support for service reliability (i.e., initial service de-

ployment, service upgrade, proactive architectural optimization,

and reactive failure mitigation) and reﬁned characterization of

the intensity of dependencies.

Index Terms—cloud computing, software reliability, AIOps,

service dependency

I. BACKGROUND AND MOTIVATION

Modern cloud systems, including Huawei Cloud, are often

constructed from a complex and large-scale hierarchy of dis-

tributed software modules.The common practice is to develop

and deploy these software modules as cloud microservices

that collectively comprise multiple cloud services [1], e.g.,

resource allocation, virtual network management, and virtual

machine management. Different microservices serve different

functionalities. The microservices communicate through well-

deﬁned APIs and respond to external requests as a whole

through service invocations.

Such an architecture beneﬁts scalability, robustness, and

agility but also complicates system reliability engineering.

However, the interactions between services cause dependen-

cies, resulting in the cascading impact on the system. Despite

various fault-tolerance mechanisms introduced, it is still possi-

ble for minor anomalies to magnify their impacts and escalate

into system outages. When a cloud service or microservice

enters an anomalous status, the anomaly can cascadingly prop-

agate through the service-calling structure, causing a degraded

user experience or even a service outage [2].

The cascading impacts hinder system operation and mainte-

nance, deteriorating customer satisfaction. For instance, during

the initial service deployment or service upgrade, all the

services it relies on should be ready. During the failure

mitigation and recovery, the cascading impact will slow the

This work was supported by Key-Area Research and Development Program

of Guangdong Province (No. 2020B010165002), Key Program of Fundamen-

tal Research from Shenzhen Science and Technology Innovation Commission

(No. JCYJ20200109113403826), and the Research Grants Council of the

Hong Kong Special Administrative Region, China (CUHK 14210920).

Engineer Specification

Configuration Files

OpenTracing

Manual Update

Configuration Parser

AID

Data Source

Dependency Analysis

Initial Service

Deployment Service

Upgrade

Architectural

Optimization

Failure

Mitigation

Application Scenarios

Fig. 1. The architecture of DMS.

recovery. Therefore, evaluating and managing the cascading

impacts caused by service dependencies is crucial.

II. KEY INNOVATIONS

This paper classiﬁes the dependency types in cloud systems

and demonstrates the design of the Dependency Management

System (DMS), an end-to-end platform for managing the ser-

vice dependencies in the production cloud system. DMS sup-

ports the full-lifecycle support for service reliability, i.e., initial

service deployment, service upgrade, proactive architectural

optimization, and reactive failure mitigation. DMS integrates

our previous study on the aggregated intensity of service

dependency [2] to characterize the degree of cascading impacts

and provides a reﬁned characterization of dependencies. In

addition, DMS also features automatic conﬁguration parsing

and multi-source dependency fusion for practicality.

III. DEPENDENCY TYPES

The dependency relations in a cloud system are diverse.

In Huawei Cloud, we categorize the dependencies according

to the architectural level, i.e., service-level dependencies and

microservice-level dependencies.

1) Service-level dependency: If the dependency is between

two cloud services, we call it a service-level dependency.

Service-level dependency can be further divided into the

following three subtypes, i.e., deployment dependency, run-

time dependency, and operational dependency.

Deployment dependency indicates dependency during the

deployment of a cloud service. The deployment phase may

rely on some cloud services to create and conﬁgure resources.

For example, the elastic computing service depends on the

API management service to register public APIs. The elastic

computing service also depends on the block storage service

to allocate the required resource.

arXiv:2210.06249v1 [cs.DC] 28 Aug 2022

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

ManagingServiceDependencyforCloudReliability:TheIndustrialPracticeTianyiYang,BaitongLi,JiachengShen,YuxinSuy,YongqiangYangz,andMichaelR.LyuDepartmentofComputerScienceandEngineering,TheChineseUniversityofHongKong,HongKongSAR.Email:ftyyang,btli,jcshen,lyug@cse.cuhk.edu.hkySchoolofSoftwareEngineer...

收起<<

Managing Service Dependency for Cloud Reliability The Industrial Practice Tianyi Yang Baitong Li Jiacheng Shen Yuxin Suy Yongqiang Yangz and Michael R. Lyu.pdf

共2页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Managing Service Dependency for Cloud Reliability The Industrial Practice Tianyi Yang Baitong Li Jiacheng Shen Yuxin Suy Yongqiang Yangz and Michael R. Lyu

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: