Filter
Top Products
Chapter 6 Managing the File System
Working With Stripe Groups
StorNext User’s Guide 113
Working With Stripe Groups
A stripe group is a logical disk volume in a file system that consists of one
or more LUNs used to store metadata information, journaling
information, and user data. Stripe groups are used to create logical
volumes that can span multiple controllers on an array and even multiple
arrays for even greater performance. Stripe groups are composed of
LUNs of the same size and disk class based on your cost and performance
requirements. (Fibre channel provides the highest performance and duty
cycle.)
A stripe group contains definitions about read and write permissions,
real time I/O constraints, a stripe breadth definition, multi-pathing
methodology, and an affinity association. A file system can contain
multiple stripe groups.
Stripe groups are bound together to create a StorNext file system. When
data is written into the file system, two critical things happen.
First, data is separated from metadata. Metadata operations are typically
small and random, and they require a lot of head movement on disks.
Data however, tends to be written in large sequential patterns with less
head movement. By separating data and metadata, thrashing is
minimized and performance is maximized.
The second critical thing that can happen when data is written to the file
system is file steering. Stripe groups can be different sizes and categories
of disk, so you could have a file system with one stripe group of SATA
disk intended for proxy files or temporary storage (i.e., less critical data
on lower duty cycle, slower disk). You might have another larger fibre
channel stripe group for storing raw content (i.e., high value data, on
higher duty cycle, higher performance disk).
Getting data to specific stripe groups is accomplished using affinities, a
mapping that ties a directory in the file system to a specific stripe group.
When you write files to a directory, StorNext uses affinities to
Note: For more information about using the snfsdefrag
command, refer to the snfsdefrag(1) man pages.