Raw Devices and File Systems 觀後發問

[watermelonyu]

再看完TWP_Oracle_HP_files.pdf這篇文章後

我有一個疑問

我的環境是rp8400,EMC Cx600, online jfs 3.5

目前這個環境只有我在用,所以我可以定沒有其他I/O

我的測試是用iozone工具,只單測試filesystem

發現使用mincache=direct and convosync=direct 反而會比較慢,單是測試filesystem就比較慢了,難道oracle 用了就會快這麼多,光enable 上面兩個,oracle 的transaction throughput 可以多到10000,

我的測試數據如下......

只用delayong and nodatainlog filesystem可以跑到350MB/sec

但是加上mincache=direct and convosync=direct就馬上降到270MB/sec

是在使用mincache=direct and convosync=direct有要注意的地方,還是用這幾個參數要搭配oracle才有用

因為我們目前所有的oracle 的環境都沒有用上這幾個function,如果只是使用mincache=direct and convosync=direct就可以快這麼多,那我一定要用看看

[watermelonyu]

Hi Joey,



I find the doc on HP knowledge DB.

It need test to find which configuration is best for your application. Thanks a lot for your sharing.

[watermelonyu]

Hi,Sir

could you share this doc with me

thanks

[watermelonyu]

Hi Joey,

I am so sorry make you confused. The doc what I meant is "TWP_Oracle_HP_files.pdf".



For JFS detail options, please refer

"VERITAS File System 3.5 (HP OnlineJFS/JFS 3.5) Administrator's Guide". You can download it from http://docs.hp.com

Tks.

[watermelonyu]

終於找到答案了

因為我用的IO Zone 並不適用random access,所以測試上數據會比較差

Oracle 的使用比較偏向random access的話,使用

mincache=direct, convosync=direct,效能才會提升

如果i/o偏向sequential access,則用delaylog, nodatainlog比較好



In general, for VxFS filesystems use these mount options:

delaylog, nodatainlog

For VxFS filesystems with primarily random access, use:

mincache=direct, convosync=direct

“What???” The short version: When access is primarily random, any read-ahead I/O performed

by the buffer cache routines is “wasted”: logical read requests will invoke routines that will look

through buffer cache and not get hits. Then, performance degradation results because a physical



read to disk will be performed for nearly every logical read request. When mincache=direct is

used, it causes the routines to bypass buffer cache: I/O goes directly from disk to the process’s

own buffer space, eliminating the “middle” steps of searching the buffer cache and moving data

from the disk to the buffer cache, and from there into the process memory. If mincache=direct is

used when read patterns are very sequential, you will get hammered in the performance arena,

because very sequential reading will take big advantage of read ahead in the buffer cache,

making logical I/O wait less often for physical reads. You want much more logical than physical

reading for performance (when patterns are sequential).

[watermelonyu]

You need have Online JFS to enable mincache=direct, convosync=direct

two options.

Thanks for your performance sharing.