Interviewer: what are cache penetration, cache avalanche and cache breakdown?

Preface

Original official account ：bigsai

For cache penetration 、 Cache avalanches and cache breakdown often occur in interviews , Let's see what it is today ？

redis Cache penetration

understand

• Focus on through Well , That is to say, through redis Go straight through mysql, It's usually a nonexistent key, Query in the database as null. Every request falls in the database 、 And high concurrency . The database will hang up if you can't carry it .

Solution

• You can find out null Set to key Cache object of .
• Of course , It can also be based on obviously wrong key On the logical level verification .
• meanwhile , You can also analyze user behavior , Whether it's a deliberate request or a reptile 、 The attacker . Restrict user access .
• Others, etc , For example, with a bloon filter ( Super large hashmap) Filter first .

redis Cache avalanche

understand

• An avalanche , It's something Things are pouring in It means , Like an avalanche . ad locum , Namely redis Cache collective Mass collective failure , In the case of high concurrency key A large-scale visit mysql, Make the database crash . Imagine the country Aging population . Later that day people concentrated on 70-80 year , No one's working . The state's labor force creates pressure .
  

Solution

• The usual solution is to key After the expiration date of, add a random number , Give Way key Uniform failure .
• Consider using queues or locks to keep programs within pressure limits , Of course, this scheme may affect the concurrency .
• Hot data can be considered not to be invalid

redis Cache breakdown

understand

Cache breakdown , It means a key Very hot , Constantly carrying big concurrency , Large concurrent centralized access to this point , When this key At the moment of failure , Continuous large concurrency breaks through the cache , Direct request database , It's like brute force .

• Breakdown and penetration are different , Penetration means thinking Bypass redis To make the database crash . And breakdown can be understood as The front is just breakdown , This is usually a large number of concurrent pairs of key Large scale read and write operations . This key Request a large number of databases during cache failure , Too much pressure on the database makes it crash . Just such as In the second kill scene 10000 The dollar mac and 100 The block mac This 100 The order for block will be rushed to the top , Constant requests ( Of course, the specific seckill has its own way of handling. Here is just an example ). So cache breakdown is for a common key A large number of requests caused the database to crash .

Solution

• Mutexes can be used to avoid a large number of requests falling at the same time db.
• The bloon filter , Determine whether a container is in the collection
• You can set the cache to never expire ( Suitable for some cases )
• Do a good job in fusing 、 Downgrade , Prevent system crash .

Original official account ：bigsai