[Cloud Residency Co-Creation] Huawei Cloud Global Scheduling Technology and Practice of Distributed Technology

题记：分布式云原生都涉及哪些核心技术？Which scenarios will be involved？How to accelerate enterprise through distributed cloud native？

分布式云原生,是一个新的概念.使用分布式技术,Integrate cloud-native practices、Provide application computing power、Centralized supply、流量治理、The ability to work with data,This creates a ubiquitous cloud-native experience.

1 华为云分布式云原生——Global scheduling technology and practice

分享人：Wang Nannan, Distributed Cloud Architect of Huawei Cloud

分享目录：

• An overall introduction to HUAWEI CLOUD's distributed cloud-native technology
• Global scheduling architecture and technology sharing
• 实践案例

1.1 什么是云？

计算能力作为一种商品,可以在互联网上流通,就像水、电、煤气一样,可以方便地取用.

The status quo of traditional cloud native is fragmented,缺乏全局管理视图：

Distributed cloud native is an application-centric business model,Made a new experience：全局统一,无处不在,随时可用.

• 应用 + 算力：Provide computing resources required by applications anytime, anywhere
• 应用 + 流量：实时、跨域、按需调配应用访问流量
• 应用 + 数据：全业务一体化迁移、容灾、弹性

1.3 How HUAWEI CLOUD does it？

通过 1 个中心、3 个核心能力,4 类区域覆盖,为应用从核心区到业务现场形成广泛覆盖：

• 1 个中心：Provides access to cloud-native infrastructure、应用、用户权限、安全策略的一站式管理.华为云 UCS via a center console、一张分布式网络、Accord cloud native ecology,Four types of area coverage、N 个基础设施
• 3 个核心能力：Application computing power supply、应用流量管理、应用与数据协同
• 4 类区域覆盖：核心区域、热点区域、客户机房、业务现场

1.4 构建 4 个核心技术

• Cloud native services are distributed globally,Unified application ecology
• Distributed security management and control strategies、Establish uniform security and authority standards across the organization
• Multi-cluster management system developed by open source,Simplify Hailiang cluster application management
• 流量、Unified orchestration of data and applications,End-to-end guaranteed distributed services

2 Global scheduling framework and technology sharing

Current public cloud,First you have to choose a region、Select Availability Zone,比如在广州、上海、深圳等,It is recommended to choose nearby.Therefore, the current schedule of the public cloud is still in accordance with region（区域）Plan for manual selection,与用户的实际资源使用之间存在差异.

实际业务需求：

• Calculate multiple regions：Want to calculate away from my data,The user schedules
• Multi-line specifications：Businesses generally test multiple specifications,C3、C6、C63、C7都可以,除了 S The series is no exception
• Computational scale is highly flexible：按需弹性,白天晚上,季度、Holidays fluctuate,Hope to solve large-scale sudden problems

3. 如何解决上述问题？

全域调度：Cloud edge global resource integration,面向垂直场景以满足用户业务需求为目标

全域调度愿景：

1. 供给侧改革：Integrate cloud-edge global resources,整合为全局统一视图,按照租户要求,协同全局资源完成供给
2. 需求侧升级：面向各种垂直场景,and real business needs,抽象业务调度目标,以满足业务需求为最终调度目标.

Global view of the supply side：

core technical problems：多个租户资源如何合理分配,先给谁,不给谁,给多少？

Resume capacity limit、满足用户需求;

1. 如何体现不满足量？
2. How to present nonstop tenant preferences？

Customers pass the needs satisfaction assessment：

Resources are done,How traffic is handled？

General diagram of global scheduling technology：资源、应用、Linking business strategies

4 实际案例

4.1 Internet clients use global scheduling

A head Internet client uses global scheduling,Cope with large-scale traffic bursts.Provides resource recommendation capabilities：

业务特性：

• large burst：The amount of elastic resources for emergencies is normal 5- 9 倍
• 短时长：The duration of burst resource usage < 16 H
  问题：
• How to prepare resources according to the peak will lead to a lot of waste of resources,资源利用率不足 20 %
• If you buy resources according to the daily average,In the event of an emergency, sufficient resources cannot be obtained to support the business.

4.2 HUAWEI CLOUD real-time transcoding service

The real-time transcoding service industry is an Internet business.

1. Multi-regional distribution of business,However, it cannot reasonably utilize cross-regional resources,before the volume of business,Significant reserve resources are required
2. User global access,It's hard to balance cost and experience
3. Various calculation specifications are available,But don't know how the cost can be optimized

1. Global traffic trend judgment,Assists local cluster resiliency decisions

• Forecast business volume,Decide on the global business cluster offloading strategy
• Combined with the capacity of each resource pool, combined elasticity recommendations are given

2. Business collaboration has stateful single line,Rapid volume reduction control

• Based on multiple resource pools Node 调整,The allocation strategy of output services to multi-specification resources,Reduced bootstrap costs

本文参与华为云社区【内容共创】活动第18期.
https://bbs.huaweicloud.com/blogs/364560

任务21：华为云HCDE专场暨分布式技术峰会