Table of Contents
Advertisement
vCenter Configuration Manager Hardware and Software Requirements Guide
SQL Server IO Configuration
When it comes to Disk IO subsystems, many IT organizations tend to "paint with a broad brush" and do
not analyze the individual technical drivers behind a given system. Unfortunately this tends to result in
many SQL Server installations that are configured with RAID (Redundant Array of Inexpensive Disks) 5
arrays, which, from a SQL Server standpoint, is not preferred. RAID 5 is generally a compromise between
write performance and data redundancy. The more redundant you make a system, the more work you
have to do in order to write data. SQL Server is extremely disk-write intensive, which means performance
suffers when SQL is configured with RAID 5. Knowing how each of the RAID levels work can help the
SQL DBA to configure the disk IO subsystem in the most efficient manner.
RAID configuration levels are as follows:
RAID 0: RAID level 0 is also known as "Striping Without Parity". In this configuration, each block of
n
data is written to each disk in the array in a "round robin" fashion. This means that each disk in the
array only holds a portion of the total data written. Depending on the array configuration, this
drastically improves read performance, as data can be read in small parallel chunks. This method also
provides improved write performance, as data can be written in parallel (However, some time is spent
breaking the data into the "stripe" that will be written). Unfortunately there is no fault-tolerance in this
model, so anytime a drive fails in the array, the entire array fails. A minimum of 2 drives is required for
RAID 0, and the resulting size of the array is calculated by adding the sizes of the drives together.
RAID 1: RAID level 1 is also known as "Disk Mirroring" or "Disk Duplexing" (Mirroring on a single
n
channel, Duplexing when multiple channels are used). In this configuration, each bit of data that is
written to a single disk is duplicated on the second disk in the array (RAID 1 is limited to 2 physical
disks). This means that the array is capable of increasing the read performance (In a duplexed
environment, the performance is theoretically doubled), while at the same time provide fault tolerance
in the event that a drive fails. Write performance is not affected by RAID 1. Only 2 drives can participate
in a RAID 1 array, and the size of the array is the same as a single disk.
RAID 5: RAID level 5 is also known as "Disk Striping with Parity". Like RAID 1, data is written to each
n
disk in the array in a "round robin" fashion, but there is also an additional block of data written as
"parity". This parity information can be used to rebuild the array in the event of a disk failure. RAID 5
is by far the most popular RAID configuration in data centers, and represents an effective compromise
between read performance and fault tolerance. Since time must be spent calculating the parity stripe,
write performance isn't as good as RAID 0. A minimum of 3 disks is required for RAID 5. The size of
the array is calculated by taking the added size of the total disks and then subtracting the size of 1 disk
(e.g. 80GB + 80GB + 80GB = total array size of 160GB).
RAID 0+1: RAID 0+1 is also known as a "Mirror of Stripes". In this configuration, 2 RAID 0 arrays are
n
also mirrored with RAID 1. This gives you the fast read/write performance of RAID 0 as well as the
fault tolerant features of RAID 1. Think of this configuration as "Performance First, then Fault
Tolerance".
RAID 10: RAID 10 is also known as a "Stripe of Mirrors". In this configuration, multiple RAID 1 arrays
n
are also striped. Think of this configuration as "Fault Tolerance First, then Performance".
When examining the various RAID levels for use with SQL Server, some easy guidelines can be followed:
76
VMware, Inc.
Advertisement
Table of Contents
