一直以來有一個模糊的概念：
循序讀寫和隨機讀寫的效能，和 raid stripe size 有關。

雖然一般來說不太理這件事，既然常常兩種應用都有，就按照 raid card 的預設值設個中立的 64KB 或 128KB 就好。

但最近想把這件事情搞清楚，發現 google 大神越教越迷糊。

oracle 建議要設大一些： http://www.dba-oracle.com/t_setting...stripe_size.htm

toms hardware 測試也是大一些好: http://www.tomshardware.com/reviews...RTS,1735-4.html

網路仁兄 3ware raid card 的測試也是大一些好：
http://blog.tinlans.org/2010/04/29/...B8%AC%E8%A9%A6/

areca raid card 的說明符合我長久以來的印象：
A larger stripe size produces better read performance, especially if your computer does mostly sequential reads. However, if you are
sure that your computer does random reads more often, select a smaller stripe size. (但此建議跟 oracle 不太相符了)。

3ware raid card 的建議和 areca 相反(和 oracle 一樣),
但是上面 3ware card 的實際測試似乎不太支持這樣的說法：
The default stripe size of 64KB will give the best performance with applications that have many sequential reads and writes. A larger stripe size will give better performance with applications that have a lot of random reads and writes. In general, the smaller the stripe size, the better the sequential I/O and the worse the ransom I/O. The larger the stripe size, the worse the sequential I/O and the better the random I/O.

關於為何 stripe size 大反而 random I/O 比較好
(和一般人直覺相反），看到常見的解釋如下，
with a small stripe size, all disks in an array will contribute to the IO of the same (larger) file, thereby multiplying the sequential bandwidth most of the time. With a large stripe size, disks will be able to service small files separately and independently, thereby increasing the random IOPS of the array.

所以，究竟什麼情境下該把 stripe size 設小呢？
為何各方說明，理論，與實測不太一致..