System design. Short chain system design

目录

What is a short chain

Why do you need short chains

Short-chain principle

301 和 302 的区别

短链系统

方案演进

直接记录 Long-chain and short-chain mapping table,直接查询

存储映射关系,用HashCode优化查询效率

 Redis替代DB存储映射关系

长短1:1Scenario improvements

 Millions of high-performance short-chain systems

---

What is a short chain

Shorter links

Why do you need short chains

Sending text messages should be shorter,美观,节省空间 

The QR code generated by the short chain is less dense,更容易被识别.如图示：

The link is too long to send

Long link is hereQQWechat or Dingding is broken in the middle,很不友好

Short-chain principle

Visit the short chain Follow the long chain to respond to the request

Visit the short chain,Redirect to long chain

一般使用,重定向302

301 和 302 的区别

301永久重定向,That is, the browser only needs to request the long link for the first time,The next time you visit this short link, you will not request it from the short URL server,而是直接从浏览器的缓存里拿.This can improve the browser's access speed,但是也有一个问题,That is, if we want to count the visit data of the active link, there is no way to start.Or is this activity over and I want to remove the access entry,However, due to browser caching, some users may also access it.所以我们一般不使用301.

302 临时重定向,That is, every time the short link is accessed, the short URL server will be requested（除非响应中用 Cache-Control 或 Expired 暗示浏览器缓存）,这样就便于 server 数据监控,所以虽然用 302 会给 server 增加一点压力,But the pros outweigh the cons.

短链系统

基本能力：

1. 生成短链：According to the long chain 获取 短链,并记录
2. Find the long chain：根据短链 映射 长链
3. Whether the query has been generated：According to the long chain Check the short chain
4. 超时过期：短链 Timeout cleanup mechanism

调用流程： 

方案演进

直接记录 Long-chain and short-chain mapping table,直接查询

方案分析：

1. 短链 生成 跟 长链 关系不大,Use a two-dimensional table to record the mapping relationship
2. 长链,短链 是字符串,Creating an index is not cost-effective
3. Concurrency has database query efficiency as a bottleneck
4. Expiration time requires additional implementation

解决痛点：

1. 路径1：存储映射关系,Combine short and long chains 做Hash计算,Then resume indexing.查询的时候,先根据HashCode查询,Then query based on text
2. 路径2：存储映射关系,可以使用redis实现
3. 路径3：Make short and long chains 有逻辑关系,Long chains can be directly reduced using short chains

存储映射关系,用HashCode优化查询效率

Get long chains based on short chains 或者 Get short chains based on long chains：使用类似HashMap的get算法,先根据HashCode查询,Then query according to the short chain text

逻辑：

1. Find short scenes according to the length：
   1. 长链 计算 得到HashCode值
   2. 根据HashCode值查询,Then in the list of queries,再根据 文本查询
2. Find long scenes based on short ones：
   1. 短链 计算 得到HashCode值
   2. 根据HashCode值查询,Then in the list of queries,再根据 文本查询

分析：

1. Database performance bottlenecks still exist
2. Expiration time requires additional logical guarantees

解决：使用rediscan solve these two problems

扩展：当然也可以使用MD5替代HashCode,也可以.

 Redis替代DB存储映射关系

No matter what you useHash算法,may conflict.

For this conflict needs to be resolved,一个CodeCorresponding to multi-text scenarios.

Find the short-chain scenario based on the long-chain：

1. 长链 计算Hash值
2. 根据Hash值获取value,value应该是一个列表,And need to store long and short chains
3. 数据结构：

{     “HashCode”:[        {唯一ID,LongUrl,ShortUrl},        {唯一ID,LongUrl,ShortUrl}     ], }

1. 1:M场景 The timeout mechanism is implemented slightly with logic
   1. 唯一ID作为key设置expire过期时间
   2. 查询到“唯一ID”Need to determine whether it has expired

The advantage is high query performance,Can resist,And comes with an expiration mechanism

The disadvantage is that the maintenance of the structural relationship is slightly cumbersome and complicated

If it can be guaranteed URL和Hash值 1:1关系,The data structure is slightly simpler,Also need to maintain multiplekv关系

It is estimated that most of them will choose this method.

长短1:1Scenario improvements

前提：

1. URL和Hash值 1:1关系
2. Only do it for long chainsHash值,The short chain is relatively short, just inquire directly
3. 使用redis实现

分析：

1. URL根据HashAlgorithms compute directly
2. 冲突解决：Short chains need to be judged heavy,If it is repeated, the long chain will add a fixed string and continue to calculate the short chain judgment weight
3. 判重优化：You can use the Bloom filter to determine whether the short chain already exists
4. hash算法可以是：MD5,SHA加密算法;非加密型MurmurHash 算法

存储kv对：

1. 长 Code 》 短
2. 短 》长

 Millions of high-performance short-chain systems

1. Web是局限

    

 2. OpenResty直接请求Redis

---

END