Compaq 108164-003 - ProLiant - 800 White Paper page 30

ECG025.0997
W P
HITE APER
(cont.)
. .
. .
. .
Disk-Related Measurement Terms
. .
. .
. .
Within this document we discussed disk-related performance characteristics and how
. .
these measurement terms can affect the performance of an entire disk subsystem.
. .
. .
Understanding how a hard disk works and the measurement terms used in the industry
. .
provides insight on the possible affect(s) disks can have on the entire disk subsystem.
. .
. . For example, if your average access time on the hard disk is poor, it can become a
. .
bottleneck. Thus causing poor performance throughout the disk subsystem because other
. .
. .
components are waiting on the slowest device, which in this example is the hard disk.
. .
. .
. .
Understanding the Transfer Rates within the Disk Subsystem
. .
. .
. .
We also discussed in detail each part of the disk subsystem and how it all works together
. .
using different transfer rates. Knowing the integral parts of the disk subsystem and how
. .
. .
they transfer data from one component to another helps you quickly identify potential
. .
bottlenecks. The slowest component in the disk subsystem generally determines the
. .
. .
overall throughput of the system.
. .
. .
. .
. . The Importance of File System Cache
. .
. .
Improving performance on a disk subsystem comes with understanding the impact of file
. .
. . system caching. File system cache is the single fastest component within the disk
. .
subsystem; therefore, we know that this device is the least likely of all the disk subsystem
. .
. . components to be a performance bottleneck. If this device did become a bottleneck,
. .
however, simply add more memory to the server. This is a great way to improve disk
. .
. . subsystem performance.
. .
. .
. .
Benefits of Scaling
. .
. .
. .
Throughout this paper we learn through many examples that scaling or adding more
. .
hardware to your disk subsystem typically provides better system performance. Listed
. .
. .
below are the scaling benefits we discovered during our testing.
. .
. .
. .
Eliminates bottlenecks
. .
. .
. .
Increases I/O concurrency
. .
. .
. .
Increases cumulative transfer rate
. .
. .
. .
Decreases idle time on the controller
. .
. .
. .
Increases cumulative disk capacity
. .
. .
. .
Increases disk controller cache
. .
. .
. .
Your goal is to produce the right balance for an effective high-performing disk subsystem.
. .
It is for you to choose which level or mixture of scaling is right for your environment. The
. .
. .
test results and data contained in this paper should help you with that choice.
. .
. .
. .
. .
. .
. .
30