Linux

Hi Guys,

I just wondering if any of the AIX gurus with some exposure to Linux Redhat can tell me whether in Linux ( Red Hat or OpenSuse ) we have a similar option/capability like the CIO (Concurrent I/O) which is currently available on AIX. The reason I'm asking is because in the past we suffered a lot of locking/enqueue issues at the filesystem level and basically after implementing CIO all those issues became a thing of the past. Any comments Please will be highly regarded.

thanks....

No, there is nothing that exactly matches concurrent I/O. AFAIK. You can back into it by turning off file caching, for example:

You can drop_caches for a one-time flush:

Writing to this will cause the kernel to drop clean caches, dentries and
inodes from memory, causing that memory to become free.

Code:

To free pagecache:
	echo 1 > /proc/sys/vm/drop_caches
To free dentries and inodes:
	echo 2 > /proc/sys/vm/drop_caches
To free pagecache, dentries and inodes:
	echo 3 > /proc/sys/vm/drop_caches

This is a non-destructive operation and dirty data objects are not freeable, the
user should run `sync' first.

You can mount filesystems

Code:

 mount -o sync ....

to bypass file caching
The dirsync flag also bypasses file metadata update caching. Doing these can increase your exposure to possible data loss for long term operation.

You can use the O_DIRECT flag on open.

If this question refers to a db applications I would bet the system designers are aware of all of this for a linux install. Are you experiencing performance issues?

Hi Jim,

Thanks A lot. Really appreciate your comments. In regards to your question about having performance issues no because at this stage is merely a discusion but I have had issues in the past and basically I believe for Database servers the big vendors of Oracle and DB2 suggest to bypass the filesystem cache as the databases they have their own cache/buffercache and also that memory that you release by bypassing the filesystem cache can be allocate it to make the DB caches larger.

thanks again.