Download IBM XIV Storage System Architecture and

Transcript
Now, in contrast, see Figure 6-2, which shows a LUN virtualized into 1 MB partitions and
distributed evenly across 180 drives in a 15-module XIV Storage System configuration.
ƒ XIV full rack
• Every LUN is carved out of 180 spindles
• Every LUN enjoys the performance of 180
spindles
Figure 6-2 XIV Storage System full rack with 180 drives
In a traditional storage system, it is common for an application’s performance requirements to
be greater than the performance capabilities of a single RAID array. To engage multiple
traditional RAID arrays into providing the performance requirements of an application, one or
more of the following techniques must be used:
򐂰 Break up the application into multiple LUNs and distribute them among several
RAID arrays
򐂰 Use operating system logical volume striping techniques to distribute the operating system
across the performance capabilities of multiple RAID arrays
򐂰 Storage subsystem striping techniques that create LUNs that are striped across multiple
RAID arrays
It is common for these traditional storage system performance techniques to be combined to
meet the performance requirements of an application. These techniques are a necessity with
traditional storage. They can become challenging if there is ever the need to change these
configurations because more storage capacity or more performance capacity is required.
These activities require specialized skills and much work.
With XIV Storage System, there is no need to resort to these multiple techniques. Each XIV
Storage System LUN has only one disk performance limitation and that is the full
performance potential of the entire XIV Storage System, which includes all disks and modules
in the XIV Storage System configuration. The consequence of this massive parallel approach
and the way data is effectively laid out within the XIV Storage System grid is that LUNs can be
easily resized without physical disk layout concerns. It is unlikely that an XIV Storage System
LUN will ever be reconfigured for performance reasons.
Chapter 6. Performance
289