UNIT-1Elements of Monitoring and Tuning
UNIT 1 Elements of Monitoring and Tuning
译文：内容监控和调优
调优主要是对procfs（/proc/sys）和sysfs（/sys）目录里参数进行调整
(1)
Compensating for physical characteristic of hardware 补偿硬件物理特性
(2)
Performance tuning has sometimes been called a “black art” due to the sometimes arcane knowledge required to effectively tune a system.
性能调优被称为“黑色艺术”
(3)
Performance tuning requires an understanding of both the hardware and the components of a system and the software components of a system as well as an understanding of the interactions between the systems.
调优需要了解系统的硬件和软件并且需要理解它们之间的互相联系
(4)
Performance management is the process of making sure that adequate computing resources are available to accomplish business needs of all users.
性能管理是一个确保每一个用户有足够的计算机资源来进行商业需要
(5)
Performance tuning and management will not always involve working specifilly with hardware and software .the microprocessor that is at the heart of your computer system does not care whether it is running at five percent utilization or ninety-five percent utilization.
性能调优不仅仅是软件和硬件，微处理器是系统的心脏
(6)
Recognize and avoid traps 认识和避免陷阱
(7)
Get baseline performance data 得到基准性能数据
(8)
International system of units(SI) prefixes are decimal
Kilo(K) =10^3=1000
Mega(M)=10^6=1000000
Giga(G) =10^9=1000000000
IEC standard binary prefixes reduce confusion
Kibi (ki) =2^10=1024
Mebi(Mi)=2^20=1048576
Gibi(Gi) =2^30
(9)
Bc计算器
3/5
0
scale=2
3/5
.60
[root@shm-chong ~]#
(10)
Monitoring
Methods
.Ad hoc:interactive commands
.Scalable:SNMP,RHN,Scripts
Appropriate paramenters should be:
.relevant
.simple
.reproducible
.measurable
Graphical output can help in spotting trends
.Gnuplot,MRTG
监控的方法有两种：使用交互式的命令监控和使用SNMP,RHN,Scripts
监控图形化输出有两个软件：MRTG,Gnuplot
注意：一定要学会这两个软件的使用，并且理解SNMP协议的原理和由来
(11)
The foucus of monitoring is on:
.tracking performance metrics
.altering when performance is out-of-specification
The focus of profiling is on:
.establishing baseline performance criteria
.finding hot spots in application runtime
.identifying runtime characteristics as a basis for tuning
监控的目的：
跟踪新能指标，超出系能指标报警（altering）
分析的目的：
建立基准的系能标准
查找程序运行时的热点
(12)
Concepts apply to entire system and participating subsystems
未来需要了解系统和子系统的监控
(13)
Sample application model
(14)
Sample hardware model
(15)
Before tuning the kernel
Is syslogd a bottleneck?
Disable reverse name lookups 关闭反向解析
Set loglevel to warn for most production daemons
Disable /etc/cron.daily/makewhatis.cron
Consider disabling /etc/cron.daily/mlocate.cron
Never disable /etc/cron.daily/prelink
内核调优之前：
Syslog的瓶颈：
关闭反向解析，把日志级别调整到warn
关闭生成whatis数据库的计划任务
关闭locate的计划任务
不要关闭prelink
(16)
Procfs
Persist in /etc/sysctl.conf or /etc/rc.local
Sysctl operates against a subset of /proc/sys
Sysfs
Persist in /etc/rc.local
(17)
Kernel-doc
[root@shm-chong ~]# yum -y install kernel-doc
Installed:
kernel-doc.noarch 0:2.6.18-274.18.1.el5
Complete!
[root@shm-chong ~]# cd /usr/share/doc/kernel-doc-2.6.18/Documentation/