下面的内容那个来自Jeff Dean在LADIS大会上做的报告.
Designs,Lessons and Advice from Building Large Distributed Systems

Numbers Everyone Should Know

devices latency
L1 cache reference 0.5 ns
Branch mispredict 5 ns
L2 cache reference 7 ns
Mutex lock/unlock 25 ns
Main memory reference 100 ns
Compress 1K bytes with Zippy 3,000 ns
Send 2K bytes over 1 Gbps network 20,000 ns
Read 1 MB sequentially from memory 250,000 ns
Round trip within same datacenter 500,000 ns
Disk seek 10,000,000 ns
Read 1 MB sequentially from disk 20,000,000 ns
Send packet CA->Netherlands->CA 150,000,000 ns

关于相关设备的吞吐量(throughput)可以参考List of Device bit rates

新增部分磁盘设备的读写效率信息 added 2010-06-04

Product ~IOPS Capacity
7200 RPM Disk 100 2TB
10K RPM Disk 140 600GB
15K RPM Disk 160 600GB
Fusion-IO Device (SLC) 100000* 160 GB
Fusion-IO Device (MLC) 70 000* 320 GB
Intel SSD (SLC) 3500 writes / 35000 reads* 64 GB
Intel SSD (MLC) 8500 writes / 35000 reads* 160GB

来源 understanding the role of io bottleneck from Percona

对Oracle性能优化比较有用的部分常量.

事件名称 每次等待最大所需
容忍延时
此延时下每秒可处理
事件次数
单块磁盘读(物理读) 10 ms 0.010,000 s 100
从Oracle Buffer Cache中
的单块访问(逻辑读),500MHz CPU
20 μs 0.000,020 s 50,000
从Oracle Buffer Cache中
的单块访问(逻辑读),1G Hz CPU
10 μs 0.000,010 s 102,400
从Oracle Buffer Cache中
的单块访问(逻辑读),2G Hz CPU
5 μs 0.000 005 s 204,800
通过WAN传输SQL*Net消息 200 ms 0.200,000 s 5
通过LAN传输SQL*Net消息 15 ms 0.015,000 s 67
通过IPC传输SQL*Net消息 200 ms 0.001,000 s 1,000

此表来源 : useful Constants for the Oracle Performance Analyst

新看到部分IOPS以及延时的计算公式, 放在这儿, 可能对大家还有点用户.

RPM
Rotations
Per
Minute
Rotations
Per
Second
Rotations
Per
Mili-second
Full
Rotation
Rotational
Latency
(Half
Rotation)
Average
Seek
Time
IO
Time
IOPS
(x) (x/60) (x/60,000) (1/
[x/60000] )
(1/
[x/60000] ) / 2







Y Z (Y+Z) (1/[Y+Z])*1000
15,000 15,000/60 15,000/
60,000
4ms 2ms 4ms 6ms 167
10,000 10,000/60 10,000/
60,000
6ms 3ms 5.15ms 8.15ms 122



10ms 5ms 9ms 14ms 71
7,200 7200/60 7,200/
60,000
8.4ms 4.2ms 9.9ms 14.1ms 71

摘自James Koopman 在 DBAsupport.com上的文章 Measuring Disk I/O—A Vendor View