Redis cache breakdown, cache penetration, cache avalanche

Cache penetration

a large number of key Not in cache , The request goes directly to the database , and key It's not in the database ( Finally back to null, A meaningless query ), If hackers create such a large number of request attacks , Cache penetration may occur
Solution ：
1. You can verify at the front end or back end , Filter illegal requests , For example, judge whether the request has been tampered by the summary Algorithm
2. Invalid cache key, Set a shorter effective time （ Not recommended ）
3. The bloon filter , A very clever data structure , The concrete realization is ：
Store all possible request values in the bloom filter , When requested by the user , First, judge whether the value exists through the bloom filter , If no error message is returned directly .

The process is as follows
When stored , Through multiple hash Function calculation , Then set the corresponding position of the bit array to 1
When comparing , The use of multiple hash The algorithm calculates the request parameters , Judge whether each element at the corresponding position in the digit group is 1, If there is any inconformity, it will be judged as fasle
Of course , Different strings may be hashed out in the same position （ You can appropriately increase the size of the digit group or adjust hash function ）

Cache avalanche

It is caused by memory server downtime or simultaneous failure of a large number of hotspot caches at the same time , The request landed in the database
Solution ：
1. For server downtime , It can be distributed （ Master-slave ） To improve usability , Or current limiting , Prevent a large number of requests for access at the same time
2. For hotspot cache invalidation , It can be set to never expire （ Not recommended ）, Or set the expiration time randomly （ Peak staggering failure ）

Cache breakdown

Cache avalanche is aimed at the situation that a large number of caches for different requests are invalidated , Cache breakdown refers to the situation that a large number of requests are the same and the cache fails
Solution ：
Hotspot cache does not expire
Use mutexes , Only one request thread can reload the results into the cache after the cache expires , Other threads wait , Finally, query from the cache .