Graduation project: design seckill e-commerce system

1.  Basic business scenarios

• System （
  APP
    And WeChat applet ） The homepage shows the seckill activity , The second kill button is grayed out in non second kill time period , Users can browse the product list even if they don't register and log in 、 Details of the goods ;

• Wechat apps can only watch seckill activities , But you can't participate in the second kill , Only pass
  APP
    The client participates in seckill ;

• Second kill is usually on the shelf at a fixed time , The user can see the product , But I can't click “ Buy now ” The button , Before the second kill, users will quickly refresh the product page , A lot of visits , Enter the order page first when the seckill starts ;

• The second kill is limited and the price is unlimited , Drain the mall as much as possible ; And before the second kill , The mall needs a series of activities ; Users must register and log in before they can participate in seckill ;

• Second kill time , Only the first order submitted by the user is sent to the order subsystem of the website ;

• Each user should ensure that only one product can be killed every second , Only the first submitted order is sent to the order subsystem of the website ;

• Second kill is usually on the shelf at a fixed time , And there can be no “ Oversold ” The problem of ;

• Second kill is a marketing tool , Can not affect the daily application of the mall .

2.  Storage architecture design

2.1  Storage performance estimation

2.1.1  Register login

Daily life is about  100  Million users , Suppose that daily activities account for... Of the total number of users  20%, Then the actual user is about  500  ten thousand , Suppose the second kill attracts  10  Ten thousand new users sign up . Suppose the second kill , Co ownership 80% Users of （ namely ：400 Million users ） Log in to access the system , Need to store 400 Login data of 10000 users .

2.1.2  Commodity information

• The stored product information shares
  10 Categories  *  The most 20 individual  = 200 strip
  data

• Users participating in the second kill will quickly refresh the product page , Suppose each user refreshes  5  Time , Then the number of requests for product information is  400  ten thousand *5=2000  ten thousand . Secsha users' rapid refresh of the product page is mainly focused on the... Before the start of the activity 1 Hours , During this period of time, you can query commodity information
  QPS
   ：
  400W * 5 / (60 * 60) = 5555

2.1.3  seckill

Seckill merchandise is  1000  A power bank 、10  platform  iPhone 12, Second kill valid orders are  1010. Suppose that the average number of users on that day is  1  Orders , Then the order data of the current day has  400  ten thousand   strip , Each user queries the order  2  Time , Then there are  800  ten thousand   Order query request .

2.1.4  Order

Suppose the day of seckill  400W  Users have an average of  1  Orders and focus on the second kill  1  Within hours , To generate the order  

TPS = 400w/60/60 = 1111

; The order data to be stored is  400W  Bar record , The average number of orders queried by users on that day  2  Time ,800W The request .

2.2  Storage architecture design

among ：

• MySQL
    data storage 500 Million user information and 400 Million user login information

• MySQL
    Database maintenance 200 A relational database of product information

• MySQL
    Database storage 400W Data

• Redis
    The cache database stores the request information of commodities 2000 ten thousand , Number of order requests 400 ten thousand

3.  Computing architecture design

3.1  Computing performance evaluation

3.1.1  register

Suppose the second kill attracts  10  Ten thousand new users sign up .

3.1.2  Sign in

It is estimated that there will be 400  Million users log in , And warm up after activities , The time of login and registration is concentrated before the second kill  30 Within minutes , Then register and log in  TPS = 410W/30/60= 2278.

3.1.3  Commodity information

Users participating in the second kill will quickly refresh the product page , Suppose each user refreshes  10  Time , And focus on before the second kill  1  Hours , Then view the product information  QPS  yes

 400  ten thousand *10/60*60 = 11111

.

3.1.4  seckill

Because the second kill is limited , Suppose the goods are in  10S  Internal second kill complete ,400  Thousands of users  10S  Average requests per user  5  Time , Then the second kill request  TPS = 400 W *5/10 = 200W.

3.1.5  Order

• Second kill effective order  1010, Suppose these orders are in  5S  Endogenesis , A second kill order is generated  TPS = 1010/5 = 202.

• According to the previous assumption ,400W  Users have an average of  1  Orders and focus on the second kill  1  Within hours , To generate the order  TPS = 400w/60/60 = 1111;

• The stored order data is  400W  Bar record , The average number of orders queried by users on that day  2  Time , Query order request quantity is  800W; Suppose the query time is concentrated after the second kill  2  Hours , Query order  TPS = 800W/60/60/2 = 1111.

3.2  Load balancing design of computing architecture

3.2.1  Business characteristic analysis

Seckill request  TPS = 400 W *5/10 = 200W., Therefore, it is necessary to adopt multi-level load balancing architecture , Cover  DNS->Nginx-> Multilevel load balancing of gateway ;

3.2.2  Architecture design

Seckill request depends on login status , The login status is generally saved in the distributed cache , Therefore, sending a second kill request to any server can , Choose here “ polling ” perhaps “ Random ” Algorithm .

3.3  Cache architecture design of computing architecture

3.3.1  download APP

Suppose that among the users participating in the second kill activity  10% I haven't downloaded  App( The size is  50MB), And focus on before the second kill begins  1  Hours to download , The required bandwidth is  40  ten thousand *50MB/3600s=5500MB/s, use  CDN  cache .

3.3.2  Business characteristic analysis

The number of requests to view product information is  4000W , Query order request quantity is  800W, Use multi-level load balancing architecture .

3.3.3  Architecture design

Visitors can directly view the product information , Therefore, sending a request to any server can , Choose here “ polling ” perhaps “ Random ” Algorithm .

4.  High availability architecture design

• Second kill activity is not normal , It's a marketing tool , Short duration , However, it will generate dozens of times larger page visit traffic and order traffic than usual , If the second kill activity is deployed with the ordinary application of the mall , It may have an impact on daily business , To improve usability , The seckill system can be deployed independently , Even using a separate domain name , Make it completely isolated .

• Due to the second kill request  TPS  A very large , Queuing architecture can be used , The request is not processed immediately after it is received , Instead, put the request in the queue , The system processes data asynchronously according to its capability .

4.1  Spike queue structure

4.1.1  Business characteristic analysis

Seckill request  TPS = 200W , The request is not processed synchronously after it is received , Instead, put the request in the queue , The system processes data asynchronously according to its capability .

4.1.2  Architecture design

Use queued servers , The essence of technology is ： Request cache +  Synchronous to asynchronous +  Requester polling  .

5.  Scalable architecture design

The technical team takes  Java  Mainly , The microservice architecture has been implemented , You can choose  Spring Cloud  As the basic framework of microservices .

6.  evolution

The first phase is realized according to the above functions （ Commodity type 、 Number 、 The quantity arriving at the queuing module is based on the quantity of goods  5  Double calculation ）;

Phase II operation system , Configure item type 、 Number , Dynamically set the number of requests that arrive at the queuing module through the client / Proportion of commodity quantity .