Introduction to architecture - who moved my cake

The author participated in the back-end youth training camp held this year , After listening to one of the introductory lessons on Architecture , Make the next note .

This note is mainly about the architecture information in the back-end development , The content is extensive and introductory , At the same time, in the course, the teacher vividly explained the structure by using the metaphor of the cake shop master selling cakes .

What is architecture

Definition of Architecture

• Abstract description of the overall structure and components of software
• It is used to guide the design of all aspects of software system
• informally , There are many ways to implement a software , Method selection It plays a vital guiding role

stand-alone

The software system needs to provide external services , stand-alone Is to implement all functions in one process , And deployed on a machine .
advantage ： Simple
problem ：

• C10K problem： There is a bottleneck in the service provided
• The service needs to be stopped for operation and maintenance

Monomers 、 Vertical application | vertical partitioning

Monomer architecture ： Distributed deployment
Vertical application architecture ： According to the application of vertical segmentation of monomer
advantage ：

• Levels increase
• O & M does not need to stop service
  problem ：
• Too many responsibilities , Low development efficiency
• The explosion radius is large

SOA、 Microservices | horizontal partitioning

SOA(Service-Oriented Architecture)

1. The different functional units of the application are abstracted as service
2. Define between services Communication standards

stay stand-alone → Monomers → SOA In the process of , The evolution of function is ： process → application → service

SOA With service Centered , namely Service oriented , The communication between services needs to be through Define communication standards To carry out , It can be said to have been carried out horizontal partitioning .

Microservice architecture : SOA Of De centralization Evolution direction , Break down responsibilities and services in more detail , Promote free communication between services , So as to avoid the existence of centralization , Make the coupling lower .

problem :

• Data consistency ： How many cakes have been delivered at the loading dock ?
• High availability ： So many masters , How to cooperate ?
• government ： The oven is broken , How to cope with disasters ?
• decoupling vs Too slight ： The operation and maintenance cost is high , Is it worth it ?

Summary ：

The original intention of architecture evolution : It's like making a cake .

• The demand is growing , After all, we must increase our manpower
• More complicated , In the end, we should divide our work and cooperate

The evolution of architecture : It's like cutting a cake , The cake is getting bigger and bigger , If you can't eat a bite, you have to cut it up

• Cut vertically ( vertical partitioning )
• Cut across ( horizontal partitioning )

 

Analysis of enterprise level back-end architecture

Master LAN passed by the cake shop 3 Years of vigorous development , It has accumulated a good reputation and user base , Next , Need to scale up :

• How about the store : buy / rent
• Master, how to recruit : Master LAN's family came out / From the training class
• Whether to continue to insist on pure handmade ?
• When it's big , The focus of work should be : Refined cake making income / Key directions and future planning of the cake shop

Cloud computing

Cloud computing ： Automated management through software , A service network that provides computing resources , It is the cornerstone of large-scale familiar analysis and storage of modern Internet .

Basics :

• Virtualization technology – Full rent vs Joint tenancy
• Layout plan – The owner vs Rental platform

framework : More and more detailed functions are provided from top to bottom

• Iaas (lnfrastructure as a Service)： Buy a house vs Housing rental platform
• PaaS (Platform as a Service)： Clear the bag vs Full package
• SaaS (Software as a Service)： Zero training vs Hire a trained master
• FaaS (Function as a Service)： handmade vs Batch production of cake machine

The following diagram is in the reverse order of the above architecture , More and more detailed functions are provided from bottom to top , It can be compared with the above architecture partition .

Cloud native

Cloud native technology for organizations （ company ） In public 、 Freedom cloud 、 In the new dynamic environment such as hybrid cloud , It is possible to build and run flexible applications .

among ： Public cloud 、 Freedom cloud 、 Hybrid cloud can be understood as the cloud in cloud computing , The difference lies in the different forms of the cloud .

The following is a further description of each part in the above figure ：

Elastic resources

Elastic computing resource types

• Service resource scheduling
  • Microservices : and 、 Carving
  • Big service : Oven
• Computing resource scheduling
  • On-line : Top selling list
  • offline : Top selling list update
• Message queue
  • On-line : Peak shaving 、 decoupling
  • offline : Big data analysis

Elastic storage resources

• classic
  • object : Promotional video
  • big data : User consumption record
• Relational database
  • Cashier records
• Metadata （ The data is not very large but is used in many scenarios ）
  • Service discovery : The address book of the baker in the cake shop
• NoSQL
  • KV: To a xx The cake

summary : Treat storage resources as services

DevOps

DevOps It is a powerful tool for software delivery in the cloud native era , Throughout the entire software development cycle . Combined with automated processes , Improve software development 、 Delivery efficiency .

Dev Throughout the entire software development cycle ：

Microservice architecture

Communication standards :

• HTTP (RESTful API)
• RPC (Thrift, gRPC)

Microservice Middleware RPC vs HTTP:

• performance
• Service governance
• Agreement interpretability

In the cloud scene , Microservices do not need to implement interaction logic that conforms to communication standards in business logic , It is left to the framework to do .

Service network

Service Grid (Service Mesh):

• The middle layer of communication between microservices
• High performance network agent
• Business code is decoupled from governance

Compared with RPC/HTTP frame :

• Unification of heterogeneous system governance
• Decouple from business processes , The lifecycle is easy to manage

Come here , We use the cake shop as an analogy to the back-end architecture , Here's the picture ：

 

Challenges of enterprise class back-end Architecture

problem

Challenge :

• Infrastructure level
  • Physical resources are limited
    • machine
    • bandwidth
  • Resource utilization is constrained by Deployment Services
• At the user level
  • The network communication cost is large
  • Network jitter leads to higher operation and maintenance costs
  • Heterogeneous environment , The water level of resources in different instances is uneven

Leave online resources and pool

Solutions to improve the utilization of physical resources within the enterprise

Core revenue :

• Reduce the cost of physical resources
• Provide more flexible resources , increase income

Solutions : Leave online resources and pool

• Characteristics of online business
  • IO It's mainly intensive
  • Tidal property ( There are few visits in the early morning 、 There is a large number of visitors in the evening peak )、 The real time ( Delay sensitive )
• Characteristics of offline business
  • Computing intensive is in the majority
  • Non real time

Automatic volume expansion and shrinkage

Core revenue ： Reduce business costs

Solutions ： Automatic volume expansion and shrinkage

• Utilize the tidal nature of online business to automatically expand and shrink capacity

problem ： What index is the expansion and contraction capacity based on ？
example ：CPU utilization 、 Memory utilization

Microservice affinity deployment

Core revenue ：

• Reduce business costs
• Improve service availability

Solutions ： Microservice affinity deployment

• Schedule containers that meet affinity conditions to a host
• Microservice middleware communicates with service grid through shared memory
• The implementation of service grid control surface is flexible 、 Dynamic traffic scheduling

Traffic management

Core revenue :

• Improve microservice invocation fault tolerance
• disaster
• Further improve development efficiency ,Dev0ps Play to the extreme

Solutions : Based on micro middleware & Traffic governance of service grid

• Fuse 、 retry
• unitized
• Complex environment ( function 、 preview ) Traffic scheduling for

CPU Water level load balancing

Core revenue :

• Balance the differences in computing power in heterogeneous environments
• Provide positive input for automatic expansion and contraction

Solutions : CPU Water level load balancing

• IaaS
  • Provide probe resources
• Service Grid
  • Dynamic load balancing

 

The end of the

1. No best architecture , Only the most appropriate architecture
2. How to do architecture design
   • Demand comes first . Find out what problem to solve
   • Industry Research . What solutions are available in the industry
   • Technology selection . Inside / What are the basic components of the community
   • Abnormal situation . Think about it xxx What if I can't
3. Architecture and engineer growth
   • Technology Manager
   • Architects