Shulou(Shulou.com)06/03 Report--

Today, I would like to talk to you about what is a micro-service, many people may not know much about it. In order to make you understand better, the editor has summarized the following content for you. I hope you can get something according to this article.

Monolithic architecture

Concept

Monomer architecture is also called monomer system or monomer application. It is an architectural way to couple all the functions and modules in the system into one application.

Characteristics

The main characteristics of the single architecture are:

1. Packaged into a separate unit (led to a unique jar package or war package)

two。 Run as a process

Monolithic architecture

Advantages

Easy to develop: simple development mode, good IDE support, easy to run and debug.

Easy to test: all functions run in one process, and once the process is started, the system can be tested.

Easy to deploy: only one software package needs to be released to the server.

Easy to scale horizontally: you only need to create a server node, configure the runtime environment, and then publish the package to the new server node to run the program (of course, you also need to adopt a distribution strategy to ensure that requests are effectively distributed to the new node).

Shortcoming

High maintenance cost: when the application has more and more functions and the team becomes larger and larger, the communication cost and management cost increase significantly; when bug appears, there are more and more combination of reasons that may cause bug, resulting in an increase in the cost of analysis, positioning and repair; and in the case of lack of in-depth understanding of global functions, it is easy to introduce new bug when repairing bug.

Long continuous delivery cycle: build and deployment time increases as functionality increases, and any minor modification triggers the deployment pipeline.

The training period of new recruits is long: it takes more and more time for new members to know the background, business and configuration environment.

High cost of technology selection: single-block architecture tends to use a unified technology platform or solution to solve all problems, if you want to introduce new technology or framework later, the cost and risk are high.

Poor scalability: with the increase of functionality, the cost of vertical scaling will become higher and higher, while for horizontal scaling, because all the code runs in the same process, there is no way to expand independently for some of the functions of the application.

Adopting an outdated monolithic architecture makes it difficult for companies to hire potential developers, applications cannot be expanded, and reliability is very low, so let's take a look at what the micro-service architecture looks like.

Micro-service architecture

Concept

Microservice is an architectural style. A large and complex software application consisting of one or more micro-services. Each micro-service in the system can be deployed independently, and each micro-service is loosely coupled. Each microservice focuses on completing only one task and can accomplish that task well.

Architecture core

Core part

Gateway cluster: data aggregation, identity authentication of docking clients, anti-message playback and anti-data tampering, service authentication of function calls, desensitization of response data, traffic and concurrency control, etc.

Business cluster: generally speaking, gateways for mobile and browser access need to be isolated to prevent business coupling.

Local Cache: since the client may need to invoke multiple service aggregations to access the business, the local cache effectively reduces the frequency of service calls and also indicates the access speed. Local caches generally use automatic expiration, which allows a certain amount of data delay in business scenarios.

Service layer: atomic service layer, which implements the basic function of adding, deleting, changing and querying. If you need to rely on other services, you need to call it actively in the Service layer.

Remote Cache: access a layer of distributed cache in front of DB to reduce the number of DB interactions and improve the TPS of the system.

DAL: data access layer. If the amount of data in a single table is too large, you need to use the DAL layer to deal with the data in different databases and tables.

MQ: message queuing is used to decouple dependencies between services, and asynchronous invocations can be performed through MQ.

Database master-slave: a necessary stage in the process of service, used to improve the TPS of the system

Architecture

Common architectures are:

1. Client and server

two。 Component model based architecture (EJB)

3. Hierarchical Architecture (MVC)

4. Service oriented Architecture (SOA)

Characteristics

1. The system is made up of multiple services

two。 Each service can be deployed independently

3. Each service is loosely coupled. The service is highly cohesive on the inside and low-coupled on the outside. High cohesion means that each service focuses on completing only one function.

Advantages

Clear boundaries: for example, an e-commerce platform, we used to deploy on a server, all the code into a war package. Now, we can split it apart: user service, points service, payment service, warehousing service, information service, map service and so on. Each micro service only focuses on a specific business function. In this way, it is relatively easy to develop and maintain a single service, because its boundaries are clear enough, the business is clear enough, the user service is good, and only do the user's thing well. Compared with the previous large and comprehensive single services, the amount of code for each microservice is also relatively small.

High efficiency: single services become more and more with the amount of code, such as millions of lines of code, it may take a long time to compile an application, but now, if there is a problem in one place, for example, if there is a problem with the payment module, you only need to modify the payment module separately, after modifying the payment module, test the payment function separately, and deploy the payment module separately. Without affecting the overall speed of deployment.

Technology stack is unlimited: each service can be implemented using a different technology stack, and since different services communicate through restful API, each service can be implemented using a different technology framework and a different repository.

More expansive: with the development of the business, the number of users becomes more and more, or the number of orders increases sharply. At this time, we can specifically optimize this order service and provide a higher configuration machine for this order service. While other businesses that do not encounter bottlenecks, such as SMS services, we can not move for the time being.

Shortcoming

The cost of operation and maintenance is too high: in the past, you only need to pack a war package and throw it under the tomcat, but now we may need to deploy several or even dozens of micro-services. In that case, how to ensure the normal operation and cooperation of these dozens or even hundreds of micro-services brings great challenges to the operation and maintenance staff.

Distributed system complexity: using the architecture of micro-service, we build a distributed system, which will introduce many problems in the distributed system, such as distributed locks, distributed transactions and so on. At this time, we need to have a proper solution to deal with and deal with the problems of this system, such as transaction, idempotency, network delay, partition, circuit breaker, degradation and so on.

The communication cost is high: because the previous interface calls are all in the same process, I need to pay to call the payment method and need the integral to directly call the method to add points, but now, because the integral module or payment module has been split into separate services, if you want to call it again, it will be called through the request of http. This frequent cross-service communication has a high cost. It is also critical to choose a lightweight and low-cost way of communication that suits your business.

It is difficult to split services: how to split microservices well? This requires us to constantly explore the evolution from a single service to a micro-service architecture, which is a gradual process. In the process of evolution, micro-services are often reconstructed or even re-divided according to business changes, so as to make the architecture more reasonable.

Architectural distinction

MVC architecture

MVC architecture is a single architecture. The technologies we often use: Struts2, SpringMVC, Spring, Mybatis, and so on.

RPC architecture

RPC (RemoteProcedureCall): remote procedure call. It is a protocol that requests services from remote computer programs over the network without knowing the underlying network technology. The technologies we often use: Thrift, Hessian, etc.

Implementation principle: first of all, there needs to be a module to deal with network connection communication, which is responsible for connection establishment, management and message transmission. Secondly, we need to have codec module, because network communication is the bytecode of transmission, and we need to serialize and deserialize the objects we use. The rest is the client side and the server side, the server side exposes the service interface to be open, and the client invokes a proxy implementation of the service interface, which is responsible for collecting data, encoding and transmitting it to the server and waiting for the results to return.

SOA architecture

SOA (ServiceorientedArchitecture): service-oriented architecture

ESB (EnterpariseServceBus): enterprise service bus, service mediation. Mainly provides a service for the interaction between services. ESB includes functions such as load balancing, flow control, encryption processing, service monitoring, exception handling, monitoring emergency, and so on. The technologies we often use: Mule, WSO2

Micro-service architecture

Micro-service is a lightweight service governance solution. The technologies we often use: SpringCloud, dubbo, and so on.

Architectural distinction

Micro-service principle

AKF split principle

The industry has a simple idea for scalable system architecture design, that is, capacity and availability problems can be solved by adding machines.

This concept in the "cloud computing" concept crazy popularity today, has been widely recognized! For a rapidly growing system, capacity and performance problems are of course the first to bear the brunt. However, with the progress of time and the growth of system scale, in addition to the problems of performance and capacity, we also need to face the problem of system complexity caused by the growth of the number of functions and modules, as well as the problem of providing differentiated services brought about by business changes. And many systems, in the architecture design did not fully take into account these problems, resulting in system reconfiguration into the norm, thus affecting business delivery capacity, but also a waste of human and financial resources! In this regard, the book Extensible Art proposes a more systematic extensible model-AKF Extensible Cube (ScalabilityCube). The cube is set along the three axes: X, Y, and Z.

Y axis (function)-focus on the division of functions in the application, based on different business split

The Y-axis extension splits a large overall application into multiple services. Each service implements a set of related functions, such as order management, customer management, etc. The common solution in engineering is service-oriented architecture (SOA). For example, for an e-commerce platform, we can split it into different services to form the following architecture:

Service-oriented architecture SOA

However, by looking at the figure above, it is easy to find that when the number of services increases, the service invocation relationship becomes complex. Add a new function to the system, and the number of services to be invoked becomes uncontrollable, which leads to confusion in service management. Therefore, in general, it is necessary to use the mechanism of service registration to form a service gateway for service governance. The architecture of the system will become as shown in the following figure:

Service gateway governance

X axis (horizontal expansion)-focus on horizontal expansion, that is, "add machines to solve problems"

The X-axis extension is consistent with our previous naive philosophy to solve the problem of capacity and availability by replicating services and data equally. In fact, it is to run multiple instances of micro-service to do the mode of cluster and load balancing. In order to improve the availability and capacity of individual services, each service is divided into X-axis expansion.

X axis (expand horizontally)

Z axis (data partition)-focus on the prioritization of services and data, such as by region

Z-axis expansion usually refers to the division of the system based on the unique needs of the requester or user, and makes the divided subsystems isolated from each other but complete. There are two common schemes for Z-axis expansion in the engineering field:

Unitary architecture: in the field of distributed service design, a Cell is a self-contained closed loop that satisfies all business operations in a partition. As we mentioned above, in the Y-axis extended SOA architecture, the client's selection of server nodes is generally random, but if Z-axis extension is added here, the selection of service nodes will no longer be random, but each unit will be self-contained. As shown below:

PC user

Mobile end user

Data partitioning: for the sake of performance data security, we divide a complete data set into different subsets according to certain dimensions. A Shard is a subset of the overall data set. For example, if users are divided by a tail number, the part of users with the same tail number can be regarded as a partition. Data partitioning generally includes the following ways of data partition:

Data type such as: business type

Data range such as: time period, user ID

Data heat such as: user activity, commodity heat

According to reading and writing, such as: commodity description, commodity inventory

Principle of separation of front and rear ends

What is front and rear separation? Aren't the front and rear ends separated? It starts with the embarrassing jsp. Fine division of labor has always been the principle of making the cake bigger, and it is best for engineers in multiple fields to cooperate without contact with knowledge in other fields, so as to make it more and more efficient and easy to maintain. The template technology of jsp combines html and java code, which makes the front and back ends of traditional MVC development like glue here. The front end makes the page, and the back end turns into a template. If you find a problem, you can find the front end, and the front end does not understand the java code. The purpose of front and rear separation is to break this awkward situation.

Front and rear end separation

The principle of front-end and back-end separation is simply the front-end and back-end code separation. Our recommended pattern is best deployed by physical separation to further promote a more thorough separation. If you continue to directly use server-side template technology, such as: jsp, java, js, html, css are all piled into a page, slightly more complex pages can not be maintained.

Separation principle

This separation has several advantages:

1. The front-end technology is separated, and the respective experts can optimize their respective fields, so that the front-end user experience optimization effect is better.

two。 In the separation mode, the front-end interface is clearer, leaving only the interface model, the back-end interface is simple and clear, easier to maintain.

3. The front-end multi-channel integration scenario is easier to implement, and the back-end services do not need to be changed. Using unified data and model, multiple front ends can be supported: for example: Wechat H6 front-end, PC front-end, Android front-end, IOS front-end.

Stateless service

For stateless services, let's first talk about what state is: if data needs to be shared by multiple services to complete a transaction, that data is called status. A service that relies on this "state" data is called a stateful service, whereas it is called a stateless service.

Stateless service

Then this stateless service principle does not mean that there is no state in the micro-service architecture, but what it really means is to change a stateful business service into a stateless computing service. then the state data is migrated to the corresponding "stateful data service" accordingly.

Scenario description: for example, the data cache and Session cache previously established in local memory should be migrated to distributed cache for storage in the current micro-service architecture, so that the business service becomes a stateless computing node. After migration, we can dynamically scale on demand, and micro-service applications dynamically add and delete nodes at run time, so there is no need to consider how to synchronize cache data.

The communication style of RestFul

In principle, it should be a "stateless communication principle".

Stateless communication

Here we directly recommend a practical preferred Restful communication style because it has many benefits:

1. HTTP, a stateless protocol, has inherent advantages and strong scalability. For example, if you need secure encryption, you can have a ready-made and mature scheme called HTTPS.

2. JSON message serialization, lightweight and simple, human and machine readable, low learning cost, search engine friendly.

3. Language-independent, all the popular languages provide mature RestfulAPI frameworks, which are more ecologically sound than some other RPC frameworks.

Communication

Service communication

WebService, WCF, WebAPI, and ASHX,ASPX.

Active trigger

Data serialization transfer

Cross-platform.

Cross-language

Http penetrates the firewall.

Process communication

Net Remoting:Net platform supervises mail, does not support cross-platform.

GRPC: high-performance, open source and general RPC framework for server and mobile, based on HTTP/2 design, recommended.

WebService

Registry Eureka

The registry mainly provides three core functions:

1. Service registration

When the service provider starts, it registers the information with Eureka Server through Eureka Client, Eureka Server stores the information of the service, and there is a two-tier cache mechanism inside Eureka Server to maintain the entire registry.

two。 Provide registry form

When the service consumer invokes the service, if Eureka Client does not cache the registry, it will get the latest registry from Eureka Server.

3. Synchronization statu

Eureka Client synchronizes the current client state through registration, heartbeat mechanism, and Eureka Server.

Eureka process

Gateway Zuul

API gateway is a more intelligent application server, its existence is like the facade of the entire micro-service architecture system, all external client access needs to be scheduled and filtered through it. In addition to the functions such as request routing, load balancing and verification filtering, it also needs to be combined with the service governance framework, the circuit breaker mechanism when forwarding the request, the aggregation of services and a series of advanced functions. Main core functions:

Service routing forwarding

Authentication verification filtering

Fuse limit protection

Gateway

Authentication & Authorization

Nowadays, there are an endless stream of application development, web applications based on browsers, official accounts based on Wechat, Mini Program, App based on IOS and Android, desktop applications based on Windows and UWP applications, and so on. So many kinds of applications bring challenges to application development. In addition to implementing each application separately, we also need to consider the interaction between each application and the refinement of general modules. Identity authentication and authorization is an indispensable part of every application. Nowadays, the Internet is very demanding for information security, so a set of unified identity authentication and authorization is very important.

IdentityServer4 is such a framework. IdentityServer4 is an authentication and authorization middleware tailored for ASP.NET CORE that implements OpenId Connect and OAuth3.0 protocols.

IdentityServer4

Distributed log

Generally, we need to do log analysis scenarios: you can get the information you want by grep and awk directly in the log file. However, in the large scale, that is, the large number of logs and complex scenarios, this method is inefficient, including how to archive if the log is too large, what to do if the text search is too slow, and how to query in multiple dimensions. Centralized log management is required, with log collection summarization on all servers. The common solution is to establish a centralized log collection system to collect, manage and access the logs on all nodes.

Large-scale systems are usually a distributed deployment architecture, and different service modules are deployed on different servers. When problems occur, most cases need to locate specific servers and service modules according to the key information exposed by the problem. Building a centralized log system can improve the efficiency of locating the problem.

Exceptionless is an open source real-time log collection framework, which can be used in applications based on technology stacks such as ASP.NET,ASP.NET Core,Web Api,Web Forms,WPF,Console,MVC, and provides a Rest interface that can be applied in Javascript,Node.js. It makes log collection easy to use and does not need to know too many technical details and configuration. In the past, we mostly used frameworks such as Log4net,Nlog to collect logs. When the application becomes complex and clustered, the traditional approach may not be very suitable, because collecting logs and analyzing them will become troublesome and a waste of time.

Now the Exceptionless team has provided us with a better framework to do this, which I think is great and meaningful, thanks to them.

ELK is an acronym for three open source software: Elasticsearch, Logstash and Kibana, all of which are open source software. However, a new Beats has been added, which is a lightweight log collection and processing tool (Agent). Beats takes up less resources and is suitable for collecting logs on various servers and transferring them to Logstash. This tool is also officially recommended. At present, due to the addition of the Beats tool to the original ELK Stack members, it has been renamed Elastic Stack and recommended.

ELK

Configure Central ConfigSpring

Config first promotes git-based management, which provides two management dimensions, one is label (that is, branch), and the other is profile. Main core functions:

Service configuration

Function switch

Service configuration

Spring Config

Asynchronous queue MQMQ

As we all know, the mainstream MQ now has rabbitMQ,rocketMQ,kafka. To learn more about rabbitMQ and kafka, check out these two articles:

Detailed explanation of picture and text: Ali beloved [bunny] RabbitMQ's development strategy

Photo description: what are the secrets of Kafka that I have a crush on it?

Core function: cutting peak and filling valley

Flow peaking

Fault tolerant current limiting Hystrix

It is designed to isolate the relationship between the current service and the remote service, system and three-party package when inevitable anomalies occur in the distributed system, so as to prevent cascading failure and lead to the avalanche of the whole system. Main core functions:

Failure and graceful demotion

Real-time monitoring changes related to configuration

Circuit Breaker isolation based on Thread and semaphore

Fault tolerance and current limit

Load balancing, service invocation

Ribbon is a load balancing client, which can well control some behaviors of htt and tcp.

Feign integrates ribbon by default. The main function of ribbon is to provide software load balancing algorithm on the client side. Ribbon belongs to in-process load balancing. it is just a class library integrated into the consumer process, through which the consumer can obtain the address of the service provider. Main core function: load balancing

Load balancing

Continuous integration of Jenkins

When there are many projects, repeating operations is a great waste of time. If you encounter different project groups packaging projects at the same time, the packaging and deployment servers will be queued for use, and testers can only waste their time waiting. In order to solve these problems, we choose to find a suitable continuous integration solution. Automate to complete the repetitive steps. Jenkins also contains many plug-ins, such as code validation and so on. It is the basic technology of CI/CD. Core functions:

Pull substitution code

Package construction

Send resources to the target server

Continuous integration

Distributed cache RedisRedis

Is an ANSI C-based, BSD-licensed, log-based key-value storage component, all of its data structures are stored in memory and can be used as cache, database and message middleware. Core functions:

Main memory database

Various extended functions based on in-memory database

Distributed cache Redis

Distributed transaction

The main ways to implement distributed transactions are:

2PC (two-phase commit protocol, strong consistency, no availability)

3PC

TCC (Try-Confirm-Cancel)

Local message table, RabbitMQ is recommended.

Saga mode

Local message table: MQ distributed transaction-local message table-based on message consistency.

There is a delivery message on the

Get messages downstream

Upstream delivery stability

Downstream acceptance stability

Message queuing Asynchronous

Sub-library and sub-table sharding jdbc

Sharding-JDBC is positioned as a lightweight Java framework, providing additional services in the JDBC layer of Java, in the form of jar packages, without additional deployment and dependency, can be understood as an enhanced version of the JDBC driver, fully compatible with JDBC and various ORM frameworks. Core functions:

Data fragmentation

Separation of reading and writing

Use jdbc to access each database transparently

Sharding-JDBC

SpringCloud

SpringCloud, as we can know from the name, SpringCloud is the product of the famous Spring family, focusing on the development of enterprise-level open source frameworks.

Since its release, Spring Cloud is still developing at a high speed, considering almost all aspects of service governance, and it is very convenient and simple to develop.

Spring Cloud architecture

Service Provider: the provider of the exposure service.

Service Consumer: the service consumer that invokes the remote service.

EureKa Server: service registry and service discovery center.

Spring Cloud architecture

Support protocol

Spring Cloud uses REST API of the HTTP protocol.

Spring Cloud component running

All requests access internal services through a unified API gateway (Zuul).

After receiving the request, the gateway obtains the available services from the registry (Eureka).

After the load is balanced by Ribbon, it is distributed to the specific instances at the backend.

Micro services communicate with each other through Feign.

Spring Cloud component running

Dubbo

Dubbo, born in the Ali family, is the core framework of Alibaba's service-oriented governance, and is widely used in Chinese Internet companies; it only needs to be configured through Spring, and there is no intrusion for applications, and the design is designed to serve their own business.

Although the maintenance version of Dubbo was suspended for internal reasons, the maturity of the framework itself and the perfection of the documents can fully meet the business needs of major Internet companies.

Dubbo architecture

Provider: the provider of the exposed service, which can be started by jar or container.

Consumer: the service consumer that invokes the remote service.

Registry: service registry and discovery center.

Monitor: count the number of services and calls, and call the time monitoring center. (it can be displayed on the console page of Dubbo, and there is currently only a simple version.)

Container: the container in which the service runs.

Dubbo architecture

Support protocol

Dubbo uses the RPC communication protocol and provides serialization methods as follows:

Dubbo:Dubbo default protocol adopts single long connection and NIO asynchronous communication, which is suitable for service calls with small amount of data and large concurrency, and when the number of service consumer machines is much larger than that of service provider machines.

RMI:RMI protocol is implemented by java.rmi.* of JDK standard, blocking short connection and JDK standard serialization.

Hessian:Hessian protocol is used to integrate Hessian services, HTTP communication is used at the bottom of Hessian, Servlet exposure service is adopted, and Jetty is embedded in Dubbo by default.

HTTP: implemented with Http Invoker of Spring.

Webservice: frontend-simple and transports-http implementation based on CXF.

Dubbo component running

Each component needs to be deployed on a separate server, and Gateway is used to accept front-end requests, aggregate services, and call background atomic services in batches. Each Service layer interacts with a separate DB.

Dubbo component run storage

Gateway: front gateway, specific business operations. Gateway is completed automatically through the load balancing mechanism provided by Dubbo.

Service: atomic services, which only provide atomic services related to the business.

Zookeeper: the atomic service is registered with ZK.

After reading the above, do you have any further understanding of what microservices are? If you want to know more knowledge or related content, please follow the industry information channel, thank you for your support.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.