SCF Reference Manual for the Storage Subsystem (G06.28+, H06.05+, J06.03+)

8 Disk Load Balancing
Disk load balancing seeks to distribute the disk work load across all fabrics and processors in the
system. This chapter describes load balancing between the X and Y fabrics and pairs of processors
in a NonStop S-series processor enclosure. A processor enclosure and its I/O enclosures are called
a topology branch.
The principles that apply to the internal disks in these examples also apply to 45xx disks in a
modular disk subsystem.
Using reports from products like Guardian Performance Analyzer and Measure, you can assign
each disk in a topology branch to run its primary processes in a specific processor. For more
information, see the GPA Manual and Measure User’s Guide. This chapter describes:
• “Disk Load Balancing Between Fabrics” (page 118)
• “Disk Load Balancing Between Processors” (page 119)
“Guidelines for All G-Series RVUs” (page 119)◦
◦ “Guidelines for G06.11 and Later RVUs” (page 119)
◦ “Topology Branch Example for G06.11 and Later RVUs” (page 120)
◦ “A Closer Look at Eight Disks” (page 122)
• “Disk Load Balancing on G06.10 and Earlier RVUs” (page 123)
“Understanding SAC Ownership” (page 123)◦
◦ “Guidelines for G06.10 and Earlier RVUs” (page 123)
◦ “Topology Branch Example for G06.10 and Earlier RVUs” (page 123)
◦ “A Closer Look at Eight Disks” (page 124)
◦ “PRIMARY Command Function in G06.00 Through G06.10” (page 125)
◦ “Establishing a Disk Load Balance in RVUs Prior to G06.11” (page 130)
• “Maintaining a Disk Load Balance” (page 133)
Disk Load Balancing Between Fabrics
Because a storage adapter in a NonStop S-series server is permanently assigned to a fabric
(adapters in slots 50, 51, and 53 use the X fabric; adapters in slots 52, 54, and 55 use the
Y fabric), you can examine a detailed STATUS ADAPTER report to see whether the mirrored disks
on an storage adapter are balanced between the X and Y fabrics.
Examine this display:
-> STATUS ADAPTER $ZZSTO.#*.GRP-3*, DETAIL
The disks are balanced if the primary (-P or -B) disk or the mirror (-M or -MB) disk has an asterisk.
For example, if $OPER on the PMF CRU in slot 50 is balanced between the X and Y fabrics, the
primary (-P or -B) disk OR the mirror (-M or -MB) disk has an asterisk in a detailed INFO ADAPTER
display:
Slot 50:
2 $OPER-MB STARTED 0,298 1,269
…
1 $OPER-P *STARTED 0,298 1,269
118 Disk Load Balancing