Redis expiration key deletion strategy [easy to understand]

Hello everyone , I meet you again , I'm your friend, Quan Jun .

List of articles

• Delete... Now
• Lazy deletion
• Delete regularly
• Redis The strategy used

Redis There is a function to set time expiration in , That is, to store in redis The value in the database can set an expiration time . As a cached database , It's very practical . As in our general project token Or some login information , Especially SMS verification codes are time limited , According to the traditional way of database processing , Generally, it's self judgment that is overdue , This will undoubtedly have a serious impact on project performance .

We set key When , You can give me one expire time, It's the expiration date , We can specify this by expiration time key Time to live .

notes ： For the hash table container , The expiration time can only be set for the entire key （ The whole hash table ）, You can't set an expiration time for a single element in a key .

If a key is expired , Is it deleted from the memory immediately after the expiration time ？ If not , When will it be deleted after expiration ？

There are actually three different deletion strategies ： （1）： Delete... Now . When setting the expiration time of the key , Create a callback event , When the expiration time reaches , Key deletion is performed automatically by the time processor . （2）： Lazy deletion . The key is out of date , No matter . Each time from dict In the dictionary, press key When the value is taken , Check this first key Has it expired , If it's expired, delete it , And back to nil, If it doesn't expire , Return the key value . （3）： Delete regularly . Every once in a while , Yes expires Dictionary check , Delete the expiration key inside . You can see , The second is passive deletion , The first and third are active deletion , And the first one is more real-time . The following is a specific analysis of these three deletion strategies .

Delete... Now

Immediate deletion can ensure the maximum freshness of data in memory , Because it guarantees that the expired key value will be deleted immediately after expiration , The memory occupied by it will also be released . however Delete right now cpu Is the most unfriendly . Because deletion will occupy cpu Time for , If it happens to be cpu When you're busy , For example, when doing intersection or sorting calculations , Will give you cpu Cause additional stress .

And now redis How the event handler handles time events – Disordered list , Find a key The time complexity of is O(n), So it is not suitable for dealing with a large number of time events .

Lazy deletion

Inert deletion means , After a key value expires , This key value will not be deleted immediately , But wait until the next time it is used , Will be checked to expire , Only then can it be deleted . So the disadvantage of lazy deletion is obvious : Waste of memory .dict Dictionary and expires Dictionaries should store the information of this key value .

for instance , For some data updated by time point , such as log journal , After expiration, you may not get access for a long time , In this way, during this period of time, we will waste so much memory to save log. This is very memory size dependent for performance redis Come on , It's more lethal .

Delete regularly

From the above analysis , Deleting immediately will take up a lot of... In a short time cpu, Lazy deletion will waste memory for a period of time , So scheduled deletion is a compromise . Scheduled deletion is ： Delete every other period of time , And by limiting the execution time and frequency of deletion operations , To reduce the number of delete operations cpu Influence . On the other hand, scheduled deletion also effectively reduces the memory waste caused by lazy deletion .

Redis The strategy used

redis The expired key deletion policy used is ： Lazy deletion plus regular deletion , Combination of the two .

Publisher ： Full stack programmer stack length , Reprint please indicate the source ：https://javaforall.cn/127895.html Link to the original text ：https://javaforall.cn