Design e-commerce seckill system

【 Business background 】

As an architect of an e-commerce startup , Responsible for the design 6.18 Design of big promotion second kill system , Your business model is as follows : 

1.  You choose the best-selling and highly praised products on major e-commerce platforms to sell , No more than... Per category 20 A commodity , At present, we have done 10 Categories ;

2.  This time 6.18 Second kill chose 1000 A power bank ,10 platform iPhone12 As a second kill commodity ; 

3.  The normal daily life is about 100 Million users ;

4.  The boss asked to be foolproof .

【 Technical background 】

1.  The technical team takes Java Mainly , The microservice architecture has been implemented ;

2.  The main channels are self owned App( Include iOS and Android) And WeChat applet , In order to promote the transformation of users into App user , Download only App To participate in the second kill ;

3.  At present, there are only single rooms .

【 Completion requirements 】

1.  Design a complete architecture , Such as storage 、 Load balancing 、 cache 、 High availability 、 Expandable etc ;

2.  about 10 Within the page PPT, each page PPT Explain a design , Including architecture design and design reasons , There is no need to explain the alternatives in detail .

【 Tips 】

1.  Analysis and consideration should be comprehensive , But it doesn't mean that the architecture design should be comprehensive , If it is considered that some design points can not be done after analysis , Explain the reasons uniformly at the end of the operation ;

2.  If there is no thought , Please refer to the module 9 Of  IM  Case study ;

3.  If some information is not complete or detailed enough , You can make certain assumptions , But you can't fly in the sky , It is necessary to make a reasonable inference against the existing business background and technical background .

Basic business scenarios

User scale storage performance estimation

【 register 】

Assume that the daily living user is a registered user 1/10,  It is necessary to save the registration information of tens of millions of users

【 Sign in 】

The daily life is millions , Reading user information when logging in is daily 100 Ten thousand times

【 Browse 】

Suppose there is a second kill 80% Our daily users are online , And refresh the product page every second , Product browsing qps by 80 ten thousand

【 Place an order 】

Suppose you go through the client 、 gateway 、 Business server 、 Cache layer by layer filtering , Finally, the traffic to the storage service is only the number of requests close to the actual number of goods , It only needs 1010qps

【 payment 】

The quantity paid is equal to the total quantity of the goods , be equal to 1010 Bar record

Storage architecture design

Registration information of nearly 10 million users ,100 10000 login request ,2000 Item information

redis Adopt cluster partition architecture , In the second kill scenario, the cluster needs to undertake requests every second 80 Ten thousand times , Each machine needs to bear 4w qps

Load balancing design

Adopt multi-level load balancing , Maximize load sharing and availability

Cache design

High availability architecture design

Scalable architecture design

Split into multiple microservices , It is convenient to continue to expand functions in the future