Architecture Design Methodology

on Huawei Cloud
 Foreword
⚫ Deploying service systems on the public cloud has become common
across industries. To fully tap into the features of a cloud platform and
ensure successful deployment, selecting a suitable architecture is a vital
step. This lesson describes the basics of cloud computing, Huawei
Cloud infrastructure, and architecture design methodology on Huawei
Cloud, and explains a typical case to help you foster a basic
understanding of cloud service deployment.

2
 Objectives
⚫ Upon completion of this course, you will be able to describe:
 Basic concepts of cloud computing
 Huawei Cloud infrastructure
 Major driving forces for service cloud migration
 Five principles of the cloud service architecture design

3
 Contents
1. Developments and Key Features of Cloud Computing

2. Huawei Cloud Infrastructure

3. Major Driving Forces for Service Cloud Migration

4. Cloud Service Architecture Design Principles

5. Typical Scenario Requirements on the Cloud

4
Cloud Basics and Development Timeline
⚫ Cloud computing refers to providing compute resources through on-demand services over
the Internet. Enterprises do not need to purchase, configure, or manage resources on their
own, and only need to pay for the resources they use.

Release
Release Founding Founding Graduation
of AWS
Release Release of Release of of Docker of OCI of CNCF of K8s
GSX and Lambda
of EC2 OpenStack Cloud
ESX
Non- Release and S3 Version A Foundry
released
virtualization of KVM
by 2013 2014 2015 2015 2018
VMware
2006 2010 2011 Cloud native
2000 2001 2006
Cloud
Virtualization

5
Cloud Computing Development Models

Public cloud

Hybrid cloud
Enterprise firewall

Private cloud

Private cloud: The cloud infrastructure is owned and managed for exclusive use by a single organization.
Public cloud: The cloud service provider owns and operates the cloud infrastructure and provides cloud
services open to the public or enterprise customers.
Hybrid cloud: A combination of public and private clouds, viewed as a single cloud externally.

6
Cloud Computing Service Models
⚫ There are three service models for cloud computing: Infrastructure as a Service (IaaS) that
provides compute and storage services, Platform as a Service (PaaS) that provides
environments for developing and deploying cloud applications, and Software as a Service
(SaaS) that provides applications as services.
Traditional IT IaaS PaaS SaaS

Managed by users
Applications Applications Applications Applications

Managed by users

O&M on the cloud platform

Data Data Data Data

O&M on the cloud platform

Operating environment Operating environment Operating environment Operating environment
Managed by users

Middleware Middleware Middleware Middleware

O&M on the cloud platform

OSs OSs OSs OSs

Virtualization Virtualization Virtualization Virtualization

Servers Servers Servers Servers

Storage Storage Storage Storage

Networking Networking Networking Networking

7
 Cloud Computing Benefits

Reduced Large-scale Quick resource No need for

Easier O&M Operations at different Agile services
investment into resources acquisition capacity guessing geographical locations
fixed assets
• Investments into • Economies of scale, that • Resources can be • Resources can be • More automated • The same • Compute resources
fixed assets are is, the procurement obtained in just a few quickly obtained, and less manual standardized services are quickly obtained
reduced in the early volume from Huawei is minutes. so they are there O&M are used to provide to support agile
stage so that enormous and the price • Therefore, users can when needed. • More application- infrastructure. services.
enterprise resources is lower. easily implement • Due to the large- level and less • API-based access • A solid foundation is
can be used to tap • Huawei Cloud is capable dynamic service scale resources, it infrastructure-level allows resources to laid for companies to
into the core value of providing abundant capabilities. Obtaining can be assumed O&M be obtained in the go cloud native and
of businesses. resources for a large resources for service that resources are same way. promote business
• Fixed assets may number of businesses. innovation is also always available. innovation.
become a burden Therefore, the resource convenient.
for future requirement of one
transformation. For single customer is
example, business generally easy to be
adjustments may be met on Huawei Cloud. It
limited by the reuse can be assumed that
of existing assets. resources are always
available.

Focus on the core value of businesses and avoid repeated work

8
 Contents
1. Developments and Key Features of Cloud Computing

2. Huawei Cloud Infrastructure

3. Major Driving Forces for Service Cloud Migration

4. Cloud Service Architecture Design Principles

5. Typical Scenario Requirements on the Cloud

9
Huawei Cloud's Global Presence for One-stop Services

29
Regions

CloudOcean < 50 ms
78 Global center
Latency
CloudSea < 10 ms
Guizhou, Inner Mongolia,
AZs Anhui, Singapore, and Dublin
Regional center
CloudLake < 5 ms
Beijing, Shanghai, Guangzhou, and Edge access
Hong Kong (China)
2800 Thailand, Indonesia, UAE, South Africa,
Brazil, Argentina, and Mexico, etc.

CDN nodes

10
Huawei Cloud Infrastructure
⚫ A region is a collection of resources divided by a geographic location. Each region comprises one or more AZs and is completely
isolated from other regions. AZs within the same region can communicate with one another through a private network, while those
in different regions cannot.
⚫ An availability zone (AZ) is a geographical area with independent power supply and networks in the same service region. Generally,
an AZ is an independent physical equipment room. One region has multiple AZs. If one AZ becomes faulty, the other AZs in the same
region can still function. AZs within the same region can communicate with one another through a private network.

Long distance
Independent network 1 to n
power supplies There is no
No spillover impacts in standby
CN North- CN East- the event of a disaster data center.
Beijing4 Shanghai3

AZ 1
CN East- AZ 2 AZ 3
Shanghai3a CN East-Shanghai3b CN East-Shanghai3c

Logically
AF- interconnected Private network communication in a region (LAN)
Johannesburg1 network between
CN East-Shanghai3
regions

11
How Do I Select a Region?

01 Compliance is a must.

02 User experience comes first.

03 Functions are region-specific.

04 Cost savings are higher.

12
How Do I Select an AZ?

1. AZs aim to improve availability.

AZ 2. AZs help build active-active and multi-active applications.
3. AZs are equivalent.

Application requirements prevail in the selection.

If there is no special requirement, one AZ is sufficient.

If high availability (HA) is required, select at least two AZs.
One more AZ does not increase costs.
If there are more than three AZs, the availability
improvement is limited and the complexity increases.

13
 Contents
1. Developments and Key Features of Cloud Computing

2. Huawei Cloud Infrastructure

3. Major Driving Forces for Service Cloud Migration

4. Cloud Service Architecture Design Principles

5. Typical Scenario Requirements on the Cloud

14
Major Driving Forces for Service Cloud Migration

01 Technology-driven

02 Service-driven

03 Policy-driven

An open, flexible, easy-to-use, and secure cloud platform provides enterprises

with a new choice in modernizing their IT architectures.

15
Technology-driven
⚫ Cloud migration is essential for enterprises to achieve agile service development, process massive
amounts of data, and build highly resilient IT systems.
History: Traditional IT Phase 1: Resource integration Phase 2: Agile service

Traditional data Data services

management Service data
App App Data development tool
Data
Database Database Intelligent data lake intelligence
Database Big data AI

Traditional IT Virtualization Infrastructure services

Resource pooling
infrastructure
Virtualization silos Compute Storage Networking Container
Unified resource
PM silos VM VM VDI management Automated resource More efficient,
orchestration less expensive
Virtualization
Standard Standard
Server hardware hardware

Traditional development Agile development

App App
Development

deployment
by vendors

DevOps Service
System

O&M

innovation
Microservice Middleware

Container

16
Service-driven
⚫ Cloud vendors provide a large number of PaaS and SaaS services and complete solutions to
help enterprises explore new service requirements and enhance competitiveness.
Huawei Cloud Healthcare Intelligent Twins help
pharmaceutical companies upgrade services

Gene sequencing Drug R&D Clinical research

Single- and multiomics, gene network, and Drug generation optimization, protein structure Medical text structuring, medical knowledge
genome sequencing prediction, dynamics simulation, federated graph, AI-assisted clinical decision-making,
learning, and more biomarker discovery, and risk prediction
Genome engine Drug R&D engine Clinical research engine
Pangu drug molecule model 20+ tasks for Ultra-large healthcare
AutoGenome AutoOmics Pangu medical NLP
knowledge graph with tens of
World's first auto Brand-new multiomics with 100s of millions of online software model with 10s of
billions of vertices and
genome modeling tool AutoML framework parameters upgrade billions of parameters
trillions of edges

Limited Full control of Years Months Diagnostics and 30%

Benefits treatment efficiency
technologies digital sovereignty Lead compound development cycle

Digital sovereignty New business Higher service

control processes efficiency

17
Policy-driven
⚫ Cloud computing is a key industry, and enterprises moving to the cloud are often supported
and promoted by various governments worldwide.
13th Five-Year Science and
Technology Innovation Plan

2016 China
Shaping Europe's Digital Future
2020 Europe

Federal Cloud Computing

Strategy
2019 USA Outline of the 14th Five-Year Plan for
National Economic and Social
Development and Long-Range
Objectives Through the Year 2035

2021 China
More

18
 Contents
1. Developments and Key Features of Cloud Computing

2. Huawei Cloud Infrastructure

3. Major Driving Forces for Service Cloud Migration

4. Cloud Service Architecture Design Principles

5. Typical Scenario Requirements on the Cloud

19
Enterprise Cloud Architecture Design Requirements
The enterprise cloud architecture must meet the 1. Performance
requirements of high reliability, high security, Enterprises are eager for high-performance applications and fast
consistency, and compliance in enterprise IT, and be analysis to quickly acquire information.
capable of flexible, fast, and scalable IT innovation. 2. Cost
Enterprises are keen to optimize the scheduling of O&M
personnel and reduce the manpower input in IT maintenance so
that more resources can be allocated to promote the
Gartner: No one size fits all. combination of service demands and IT technologies, motivating
enterprise innovation.
Nonlinear 3. Security
+ mode Systems of - Enterprises require a complete security deployment with a
Accept innovation strong focus on the security of data centers (physical

Governance
instability infrastructure), access, networking, storage, and data.
Charge

(DevOps,
Systems of 4. Availability
Traditional differentiation When a disaster occurs and one data center is faulty, the other
automated,
mode reusable) data center can run in real time to ensure service continuity.
- Required Systems of +
reliability
5. Maintainability
record
When a fault or risk occurs, services can be recovered or
(ITIL, CMMI, COBIT) switched over as scheduled in a timely manner to ensure service
continuity.

20
Five Principles of Solution Architecture Design

Architecture

Security Reliability Performance Cost- Maintainability

effectiveness
These five principles are used throughout the process of solution architecture design.

21
Cloud Security Customer Requirements

Data confidentiality:

Controllable O&M: • External breach

Continuous services: prevention
• Security policies • Invisible to
• Network attack • Risk identification & unauthorized internal
prevention handling employees
More?
• Intrusion prevention • Auditable and • Invisible to cloud
• Legal compliance traceable operations service providers

01 02 03 04
Key requirements on cloud security

22
Cloud System Security Design Principles

01 Set security objectives based on system compliance standards.

The solution that best matches customer needs is the optimal one. Simply
adding security solutions is not a good idea.

02 Systematically plan the system security.

System security design embodies the bucket effect. That is, the security capability
depends on the "shortest board".

03 Implement security at all layers.

When designing a security solution, analyze the solution layer by layer
based on customer requirements.

04 Assess risks and prepare contingency plans.

Any security risk that can be identified may be a weak point of the system.
Therefore, a proper contingency plan must be formulated to eliminate risks.

23
High-Availability Requirements
⚫ The availability of a system is reflected in two aspects: reliability (service continuity) and
maintainability (quick service restoration).

MTBF
Availability =
MTBF + MTTR

Expressed in 0.99999... or x nines

Reliability Maintainability
Mean time between failures
(MTBF, also known as average life) Mean time to recover
(MTTR)

⚫ Availability can also be regarded as service continuity. Instance availability = (1 – Fault duration/Total service duration) x 100%. For
example, 99.995% availability means that a service can be accessed for 100,000 consecutive minutes (about 69 days), during which
the service is unavailable for no more than five minutes.
⚫ Reliability refers to the probability that products are able to execute the specified functions as required in the specified conditions
over the specified time.
24
 Cloud System High-Reliability Design Principles
HA of the service application layer

Reliability and robustness (with capabilities Set reliability metrics based on

like service retry and isolation, and graceful 01
service requirements.
failure)
02 Ensure that the design aims to
HA of system architecture design eliminate failures.
Service Key
availability 03 Avoid single points of failure.
Single node removal, HA cluster, DR (cross-AZ points
and cross-region DR deployment), data
backup, cloud security, and standard O&M
04 Design a loosely coupled architecture.

HA of cloud services Predict the fault form and prepare a

05
contingency plan.
Reliable infrastructure and automatic recovery
of services after environmental restoration

25
Cloud System Performance Metrics
Latency refers to the time it takes to start performing an
operation. Latency is the most intuitive metric of cloud
Compute computing performance.

Throughput is a metric that evaluates data processing speed.

Networking In terms of data transmission, throughput is expressed in
bytes/second or bits/second. Throughput limit is one of the
common performance bottlenecks.

Storage IOPS refers to the number of input/output operations per

second. It is a measurement of data transmission.

Database Concurrency refers to the capability of running several

programs simultaneously within a period of time.

Latency, throughput, IOPS, and concurrency are key factors related to cloud application performance.

26
Cloud System Performance Design Principles

01 Determine performance metrics based on service requirements.

The solution that best matches customer needs is the optimal one. Simply
adding system parameters is not a good idea.

02 Understand advanced technologies and select appropriate ones.

The use of new technologies significantly improves system performance. However,
new technologies should be used properly as they can also cause new risks.

03 Search for data characteristics and popularity, and design cache.

Reading, writing, and processing an enormous volume of data poses great
challenges to system performance. Proper use of cache policies based on
data features can improve system performance.

04 Achieve system elasticity.

Properly use cloud service features and deploy services in clusters for fast elasticity.

The performance does not need to be extreme. Being slightly higher than the standard requirements is sufficient.

27
Cloud System Cost Optimization Design Principles
Continuously reducing costs: Allocate costs based on the system resource
01 utilization over time.

02 Using cloud hosting services: Customize a plan to pay only for what you use.

03 Balancing supply and demand: Avoid over-provisioning to eliminate unnecessary costs.

Fostering an awareness of expenditure: Accurately track costs and assess the

04
profitability of different departments and products.

Using cost-effective resources: Cost-effective resources and appropriate services and

05
specifications are essential to lowering costs.

06 Optimizing costs at multiple levels: Optimize costs from multiple aspects.

28
Cloud System Maintainability Design Principles

01 Stay well informed of the system status through monitoring.

Comprehensive monitoring of system operating status is the basis for efficient O&M.

02 Reduce manual risks through automation.

Automate operations to avoid any incorrect operations and accumulate O&M experience.

Switch to update or replacement instead of simply repairing.

03
O&M is not just about maintenance. Before locating a specific issue, you can
perform the update or replacement operation to quickly restore services.

04 Focus on applications instead of infrastructure.

Fully tap into the features of a cloud platform so that users can focus on core services.

A good architecture should be easy to implement and maintain.

29
When Can Architecture Design Principles Come
into Play?

Architecture design Architecture acceptance

Follow architecture design principles to solve pending issues.

30
 Contents
1. Developments and Key Features of Cloud Computing

2. Huawei Cloud Infrastructure

3. Major Driving Forces for Service Cloud Migration

4. Cloud Service Architecture Design Principles

5. Typical Scenario Requirements on the Cloud

31
 Vmall Target Service Architecture
Front-end app PC website WAP website App Reservation Flash sales Agency
presentation Browse and display (home Transactions (shopping carts/order
pages/categories/details) placement/payment) Local client +
Page
Web page
Personal center
Other (help/customer services/ads) Page
(orders/returns/replacements/addresses) Page Management
Processing Processing
Page management Session management (same as PC) (same as PC)

Process services Operations

Process systems
Internal purchase Contract phone Flash sales Store management
configuration OMS
Front end
Function services OSM
Order placement Addresses Shopping carts Reservations Flash sales Comments Payment
Open system
Message Delivery
CPS Invoicing Return/Replacement CDS Open gateway
notifications confirmation

Basic services Fulfillment services Common services

Product management system Order management system
User center Order fulfillment system (OFS) RPC
Middle end (PMS) (OMS)
framework
Marketing management Activity management system Warehouse management
Inventory system Message bus
system (MMS) (AMS) system (WMS)

Monitoring
Other systems Third-party systems system

Background Promotion Configuration

Tax & Invoice Settlement ERP Repairing Payments Credit Logistics
channels center

32
Analysis of Vmall Target Service Architecture
⚫ Vmall is a Huawei-operated e-commerce platform that looks to facilitate lives with mobile information. Typical
activities of Vmall include device sales, holiday promotions, and new device launches, which are supported by
Huawei Cloud services.
 The front-end application presentation defines the touch points that allow users to interact with the system.
 Front end
◼ Process services: A complete product sales workflow of Vmall includes five core service processes.
◼ Function services facilitate purchase for better experience.
 Middle end
◼ Basic services are core functions.
◼ Fulfillment services include order fulfillment and warehousing management.
 Backend
◼ The terms "other systems" and "third-party systems" indicate the areas where Vmall integrates with other systems.
 Open system: External participants will perform operations by calling APIs.
 Message communication: The RPC framework and message bus are used for a distributed architecture. The RPC framework is
used for synchronous communication between applications to ensure effective management. Asynchronous messages are
transmitted through the message bus to ensure performance and reliability.

33
Vmall Service Process
Logistics
• Differentiated logistics Sharing
• Optimized • Recommendation User engagement
return/replacement from friends • User bonus points
Payment procedure • Sharing for gifts • Phone exchange
Order placement
• Installment • Better after-sales • Comments discounts
• Diverse presentation
• Improved visual design payments service • Order display
• Guest purchase • COD
Registration and login
• Simplified registration
• Optimized login

Register/Log in Browse & Buy Order & Pay Manage orders

Order
Order center Settlement
User center fulfillment Comments
system
system (OFS)

34
Vmall Service Scenario: Holiday Promotion & New
Product Release
Massive access
requests and high Security & reliability User experience
concurrency
1. Performance: concurrent 1. Robust defense: Massive 1. Reliable: smooth 24/7 access
online purchasing for DDoS attacks can be 2. Smooth: smooth operations
millions of users prevented. without frame freezing at
2. Order placement: 2. Scalper bot prevention: key time points
hundreds of thousands of The system identifies
orders per second scalpers and prevents
3. Payment: tens of them from using
thousands of online automated software-
transactions per second programmed bots to
4. Service volume: 5 to 10 automatically purchase
times (100 times in goods to ensure normal
certain scenarios) normal pricing and availability for
volume legitimate users.

Holiday promotion (Double 11 shopping festival)/New product launch (Porsche) poses huge
challenges to the performance, security, and reliability of cloud infrastructure.
35
Flash Sale Service Requirement Breakdown and
Methods 1. Pressure isolation: flash sale system and main website of the e-commerce platform
Login of hundreds of
millions of users in a 2. Area isolation: damage and non-damage areas
short time
Reliability 3. Cross-region service decoupling: deployment splitting and application routing
Huge volume of orders design
in a short time 4. Service decoupling: vertical service splitting to create several centers

Performance
5. Distributed: distributed support and middleware for all resources (compute and storage)
Millions of concurrent design
requests 6. Linear expansion: distributed data storage for orders and users
Security
High-concurrency design ?
7. DB design: vertical splitting of services that are related and unrelated to users, and
payment in a short time horizontal splitting of large databases

O&M design 8. Performance improvement: read/write isolation, CDN acceleration, and caching
DDoS attacks
9. Attack defense and traffic control: SLB traffic limiting and Nginx attack defense
Cost design
10. Risk control: scalper identification based on XXX rules
Scalper software
11. Activity support: dual-order mode that decouples flash sales and hides the main website

Scenarios Strategy classification Architecture design strategies

36
Cloud E-commerce Application Architecture

security
Cloud
AAD WAF CDN
Offline services
Access
layer

Elastic Load Balance (ELB)

Recommendation
item.xxx.com order.xxx.com car.xxx.com model training
App logs
ModelArts

Shopping carts
authentication
Order mgmt.
Order details
Home page
Front-end

Payment
Search

mgmt.
layer

Login
Replica library

Data Lake ELK

Visualization (DLV)
Service layer

Search center
Order center
User center

Evaluation

Payment
Offering

center

center
center

Cloud monitoring and O&M

Middleware

Log Tank Cloud Trace

Cloud Eye Service (LTS) Service (CTS)
layer

Memcached/Redis DDM node 1 ... DDM node N RabbitMQ/Kafka

More

Application Simple
Big data platform Operations Message
Data layer

DLF
DLF Management Notification
Shard

Shard

Primary Standby Primary Standby (AOM) (SMN)

1

RDS RDS RDS RDS DDS HBase DWS

instance instance instance instance CDM
CDM MRS
MRS DWS
OBS

37
Interactive Tasks

How do we design a good cloud architecture to

deploy the services of the order center?

38
 Now you can understand this architecture:
Region
VPC

Domain
Name

AZ 1
Cache
Service
(DNS) NAT Gateway Web server Application
server Primary
database
Subnet for
Front-end subnet Application subnet Data subnet
external access

Auto Scaling AS
Content Delivery ELB (AS)
Network ELB
(CDN)

AZ 2
VPC Cache
Endpoint Application
NAT Gateway Web server
server Standby
OBS
Subnet for database
external access Front-end subnet Application subnet Data subnet

CN East-Shanghai3

39
 Quiz
1. (Multiple-choice question) After a service system is deployed on
Huawei Cloud, in which of the following aspects can we evaluate
and optimize the service system on the cloud?
A. Reliability
B. Performance
C. Security
D. Maintainability

40
 Acronyms and Abbreviations
⚫ ECS: Elastic Cloud Server
⚫ EVS: Elastic Volume Service

41
Thank You.
Copyright © 2024 Huawei Technologies Co., Ltd. All Rights Reserved.
The information in this document may contain predictive statements including,
without limitation, statements regarding the future financial and operating results,
future product portfolio, new technology, etc. There are a number of factors that
could cause actual results and developments to differ materially from those
expressed or implied in the predictive statements. Therefore, such information is
provided for reference purpose only and constitutes neither an offer nor an
acceptance. Huawei may change the information at any time without notice.