《Systems Performance: Enterprise and the Cloud》读书笔记系列（五） —— 第二章（四）

1. There are two common perspectives for performance analysis, each with different audiences, metrics and approaches. They are  workload analysis and resource analysis. They can be thought of as either top-down or bottom-up analysis of the operating system software stack.

2. Resource analysis begins with analysis of the system resources: CPUs, memory, disks, network interfaces, busses and interconnects.  Activities include:  

a) Performance issue investigation: to see if a particular type of resource is responsible;

b) Capacity planning: for information to help size new systems, and to see when existing system resources may be exhausted.

3. Resource analysis focuses on utilization, to identify when resources are at or approaching their limit. Metrics best suited  for resource analysis are included:

a) IOPS;

b) Throughput;

c) Utilization;

d) Saturation.

These measure what the resource is being asked to do, and how utilized or saturated it is for a given load. Other types  of metrics, including latency, are also use to see how well the resources is responding to the given work load.

4. Workload analysis examines the performances of the applications: the workload applied and how the application is responding. The targets for workload analysis are: 

a) Requests: the workload applied;

b) Latency: the response time of the application;

c) Completion: looking for errors.

5. Metrics best suited for workload analysis include:

a) Throughput(transactions per second).

b) Latency.

These measures the rate of the requests and the resulting performance.