Optimization ideas from ordinary query commodities to highly concurrent query commodities

• version 0（ The initial release ）

         problem 1： More and more goods ,redis Carrying more and more

• version 1

         problem 1 Solutions ： Set timeout

         problem 2： We usually initialize at startup redis Product inventory of , If set like this , A large number of cache failures may occur at the same time .

• version 2

         problem 2 Solutions ： take redis The effective time of changes from fixed time to variable time .

         problem 3： For popular products, no one has browsed them for half a year , Can this pair reduce us redis What about your storage

• version 3

         problem 3 Solutions ： Every time we go from redis The products we get are all popular products , So we postpone reading （ In the red frame ）, In this way, we can simply realize a cold and hot separation . 

        problem 4： If users or hackers keep accessing non-existent products （ Cache penetration problem ）, In addition to the use of Bronx filters , What can we do in the back-end code .

• version 4

        problem 4 Solution ： Since there are empty products found , Then let's return to the front end with an empty product ,redis Set a shorter effective time in , With Bloom filter , It can effectively prevent cache penetration .

        problem 5： If you encounter a sudden explosion （ Hotspot reconstruction ）, In an instant, tens of thousands of visits hit our database , It directly leads to the downtime of our database （ Cache breakdown ）, What do I do .

• version 5 

        Will be taken from redis Get the code of the product, extract it and package it , Or a little messy . The following is the version 5 The optimization of the .

        problem 5 Solution ： Realize with lock , Here I also use double check （DCL） To guarantee .

        problem 6： For production environment , When the concurrency is particularly high, it is still very easy to have double write inconsistencies , What do I do .

• version 6 

         problem 6 Solution ： Database query and redis Set this step in locking , To try to ensure that our double writing is consistent .

        problem 7： Two locks have been added , Whether locks can be optimized .

• version 7

 

         problem 7 Solutions ： We can use read-write lock for this situation of reading more and writing less , Ensure that threads can execute in parallel when reading , Increase throughput .

        summary , When using the cache, you basically need to lock , If the concurrency is so large that this framework can't bear it now, it can be used more guava perhaps ehcache Wait for the local caching framework .