《Java Concurrency in Practice》ch11 Performance and Scalability

1. 没有免费的午餐

While the goal may be to improve performance overall, using multiple threads always introduces some performance costs compared to the single-threaded approach. These include the overhead associated with coordinating between threads (locking, signaling, and memory synchronization), increased context switching, thread creation and teardown, and scheduling overhead.

当发现内核CPU占用率较高（>10%），通常表示调度活动发生的很频繁，这很可能是由I/O或者竞争锁导致的阻塞引起的(When a thread blocks because it is waiting for a contended lock, the JVM usually suspends the thread and allows it to be switched out)。

可以使用vmstat/mpstat命令查看CPU使用率，如下：

/home/a/j/nomad2:vmstat 1
 kthr      memory            page            disk          faults      cpu
 r b w   swap  free  re  mf pi po fr de sr s0 s1 -- --   in   sy   cs us sy id
 0 0 0 8011960 6610640 15 13 0  0  0  0  0  0  0  0  0  299  118   84  0  0 100
 0 0 0 6795008 5783928 0 10  0  0  0  0  0  0  4  0  0  410  338  214  0  0 100
 0 0 0 6794944 5783864 0  2  0  0  0  0  0  0  0  0  0  389  284  192  0  0 100
 0 0 0 6794944 5783864 0  0  0  0  0  0  0  0  0  0  0  362  249  162  0  0 100
 0 0 0 6794944 5783864 0  0  0  0  0  0  0  0  0  0  0  380  246  172  0  0 100

2. 什么是可伸缩性

Scalability describes the ability to improve throughput or capacity when additional computing resources (such as additional CPUs, memory, storage, or I/O bandwidth) are added.

3. 避免不成熟的优化

Avoid premature optimization. 

在对性能的调优时，一定要有明确的性能需求（这样才能知道什么时候需要调优，什么时候应该停止），此外还要一个测试程序以及真实的配置和负载等环境。以测试为基础，不要猜测。
It is therefore imperative that any performance tuning exercise be accompanied by concrete performance requirements (so you know both when to tune and when to stop tuning) and with a measurement program in place using a realistic configuration and load profile. Measure, don't guess.

4. Amdahl定律

在N个处理器的机器中，最高的加速比，F为串行执行部分的比率

5.  监测CPU的利用率

如果CPU没有得到充分利用，通常有以下原因：

1) Insufficent load.
2) I/O-bound.
3) Externally bound. 
4) Lock contention.