IBM TotalStorage DS6000 Series: Performance Monitoring and Tuning

Chapter 2. Hardware configuration planning 37

RAID 5 is well suited for random access to data in small blocks. Most data transfers, reads, or

writes, involve only one disk and hence operations can be performed in parallel and provide a

higher throughput. In addition to this efficiency for random transaction operations, the RAID 5

implementation of the DS6000 is able to work in a

strip write style when processing

sequential operations for maximum sequential throughput.

2.8.3 RAID 10 array

The other possible RAID implementation in a DS6000 is RAID 10 (also known as RAID 1+0).

A RAID 10 array consists of a set of disks for

user data and their mirrors. There is no parity

disk to rebuild a failed disk. In case one disk becomes unusable, then its

mirror will be used

to access the data and also to build the

spare.

The DS6000 architecture for maximum availability is based on

two spare drives per loop (and

per capacity, per rpm). If the first array configured in a loop is defined as a RAID 10 Array, it

will be defined with one data disk plus one mirror plus two spares (made from 1 Array Site).

Or three data disk plus three mirrors plus two spares (made from 2 Array Sites). This is a

1+1+2S or 3+3+2S array configuration. This will happen for the first array of each capacity

installed in the loop if configured as RAID 10.

Once the

two spare per capacity rule is fulfilled, then further RAID 10 arrays in the loop will be

configured by the DS6000 as two/four data disks plus two/four mirrors—this is a 2+2/4+4

array configuration. Figure 2-16 to Figure 2-18 on page 38 illustrate the three arrangements

of disks that can be found in the DS6000 when there are RAID 10 Ranks.

Figure 2-16 RAID 10 Array implementation 1

IBM

TotalStorage

RAID 10 configuration example

 When using same capacity and rpm.

data spare

mirror

RAID 10

3+3+2S

RAID10

4+4

Two spares are configured on

first two array per loop.

Spares are not configured

on the rest of array.

DA

DA

1750-511

1750-EX1-1

1750-EX1-3

1750-EX1-2

Loop 0

Loop 1

38 IBM TotalStorage DS6000 Series: Performance Monitoring and Tuning

Figure 2-17 RAID 10 Array implementation 2

Figure 2-18 RAID 10 Array implementation 3

RAID 10 is also known as RAID 1+0, because it is a combination of RAID 1 (mirroring) and

RAID 0 (striping). The striping optimizes the performance by striping volumes across several

disk drives (three or four DDMs in two Array Sites). RAID 1 is the protection against a disk

IBM

TotalStorage

RAID 10 configuration example

 When using different capacity and rpm.

RAID 10

3+3+2S

Two spares are configured on

first two array per loop.

Two additional spares are

configured for faster DDMs per loop.

DA

DA

1750-511(146GB / 10krpm)

1750-EX1-1(146GB / 10krpm)

1750-EX1-3(73GB / 15krpm)

1750-EX1-2(73GB / 15krpm)

Loop 0

Loop 1

data spare

mirror

RAID 10

3+3+2S

RAID 10

4+4

RAID 10

4+4

IBM

TotalStorage

RAID 10 configuration example

 When using different capacity.

RAID10

3+3+2S

RAID10

4+4

DA

DA

1750-511(300GB / 10krpm)

1750-EX1-1(300GB / 10krpm)

1750-EX1-3(146GB / 10krpm)

1750-EX1-2(146GB / 10krpm)

Loop 0

Loop 1

–In case, initially configure 146 GB array and then add 300 GB DDMs , two

hot spares are configured for each capacity arrays per loop.

Two larger capacity hot spares are

configured on each loop.

data spare

mirror