IO intensive and CPU intensive

One 、 Computationally intensive

• Very much CPU Computing resources , Let each CPU The core is involved in the calculation ,CPU Take full advantage of performance , Avoid excessive thread context switching , The ideal solution is ： Number of threads = CPU Check the number +1

• Can also be CPU Check the number *2, Want to see JDK edition , And CPU To configure ( The server CPU Have hyper-threading ).1.8 To increase parallel computing , Want to Number of threads = CPU Number of kernel threads *2

Two 、IO intensive

The Internet 、 disk IO （ And DB、 cache ）, once IO, The thread waits , Do not execute until the end . Set the number of threads , Do something else while waiting , Increase of efficiency .

Thread context switching comes at a cost .IO Dense formula ： Number of threads = CPU The core number /(1- Block coefficient ) The blocking coefficient is general 0.8~0.9, Dual core CPU Ideal number of threads 20, The dynamic thread pool looks at the pressure measurement .

3、 ... and 、1000 Concurrent threads ,10 Taiwan machine , Each machine 4 nucleus , Design thread pool size

Ideas ：CPU、IO Intensive situations . Reasonable parameter configuration through pressure measurement . Thread pool dynamic adjustment .

CPU Concentrated ： Reduce context switching , Number of core threads 5, Queue leader 100, The maximum number of threads is consistent with the number of core threads .

IO Concentrated ： Allocate a small number of core threads , Better use CPU, Number of core threads 8, Queue leader 100, Maximum thread pool 10

It's all theoretical values . Number of threads from the core 5 Start the pressure test , Comparing the results , Determine the appropriate settings . limit ： The number of core threads and the maximum number of threads are 4, queue 96, just 100 request