IBM DS6000 Series ユーザーズマニュアル

DS6000 Series: Concepts and Architecture

In AIX all storage is managed by the 

 (LVM). It virtualizes 

physical disks to be able to dynamically create, delete, resize, and move logical volumes for 
application use. To AIX our DS6000 logical volumes appear as physical SCSI disks. There 
are some considerations to take into account when configuring LVM.

Striping is a technique for spreading the data in a logical volume across several physical disks 
in such a way that all disks are used in parallel to access data on one logical volume. The 
primary objective of striping is to increase the performance of a logical volume beyond that of 
a single physical disk.

In the case of a DS6000, LVM striping can be used to distribute data across more than one 
array (rank).

Refer to Chapter 11, “Performance considerations” on page 219 for a more detailed 
discussion of methods to optimize performance.

LVM has the capability to mirror logical volumes across several physical disks. This improves 
availability, because in case a disk fails, there will be another disk with the same data. When 
creating mirrored copies of logical volumes, make sure that the copies are indeed distributed 
across separate disks.

With the introduction of SAN technology, LVM mirroring can even provide protection against a 
site failure. Using long wave Fibre Channel connections, a mirror can be stretched up to a 10 
km distance.

Another application for LVM mirroring is online (non-disruptive) data migration. See 
Chapter 15, “Data migration in the open systems environment” on page 289.

AIX provides several modes to access data in a file system. It may be important for 
performance to choose the right access method.

Synchronous I/O occurs while you wait. An application’s processing cannot continue until the 
I/O operation is complete. This is a very secure and traditional way to handle data. It ensures 
consistency at all times, but can be a major performance inhibitor. It also doesn’t allow the 
operating system to take full advantage of functions of modern storage devices, such as 
queueing, command reordering, and so on.

Asynchronous I/O operations run in the background and do not block user applications. This 
improves performance, because I/O and application processing run simultaneously. Many 
applications, such as databases and file servers, take advantage of the ability to overlap 
processing and I/O. They have to take measures to ensure data consistency, though. You can 
configure, remove, and change asynchronous I/O for each device using the 

 command 

or SMIT.