vparresources2.5 (2010 09)

ManualsBrandsHP ManualsSoftwareHP-UX Manpages

vparresources2(5) vparresources2(5)

(For OA Based Partition Management Systems)

Deconﬁgured CPU-cores are not activated in a vPar. Any CPU-cores scheduled for deconﬁguration are

not activated at the next vPar boot. CPU-cores having its container or parent resource (socket, blade etc.)

deconﬁgured is also not considered for activation in a vPar.

A vPar is booted if there are enough healthy CPU-cores available to boot the vPar. Since deconﬁgured

CPU-cores are not used, user speciﬁed requirements for vPar CPU-core assignment may not be met in

some cases. The vPar database information is appropriately modiﬁed to reﬂect the actual CPU-core

assignments given to the vPar. If the user does not want the vPar database information to be modiﬁed,

they should disable the parspec change policy attribute for the nPartition. In that case, if CPU-core

requirements speciﬁed by the user cannot be met because of unavailability of CPU-cores, boot of the

vPars is not allowed and the vPar conﬁguration in the vPar database is left unmodiﬁed.

Indicted or parent-indicted CPU-cores are still considered for activation at boot time. Only if they actu-

ally get deconﬁgured or scheduled for deconﬁguration, they are excluded from activation.

Memory

SLM and ILM

There are two major types of memory designation, Socket Local Memory, (SLM), and InterLeaved

Memory, (ILM). With SLM, entire memory address ranges are from a single socket. For best perfor-

mance these ranges should be on the same socket as the CPU cores. ILM is an address range of memory

whose adjacent addresses reside on one or more sockets in the underlying nPartition. In this topic, we

describe the mechanics of managing each memory type.

Memory Granularity

The nPartition memory is sliced into multiple memory granules by ﬁrmware for ease of memory assign-

ment to vPars. The size of these memory granules can be optionally speciﬁed by the user. Memory

granularity refers to the maximum amount of memory that will be in each of these memory granules.

Note that memory granules should not be confused with actual memory DIMMs.

The granularity for SLM and ILM can be different and can be speciﬁed separately by the user. However,

following applies to both types of memory:

• Memory granularity is speciﬁed at nPartition creation time. Any modiﬁcation thereafter requires a

reboot of the nPartition.

• The minimum values (ILM and SLM granularity) are 256 MB.

• The default value is OS dependent and may be adjusted based on the total memory available in the

nPartition.

• Memory is assigned to virtual partitions in multiples of granule sizes.

Hewlett Packard recommends that users do not change the default memory granularity value chosen by

the system, unless there is a speciﬁc need to do that.

In a situation, where user wants to manually specify the memory granularity value, following rules apply:

• Any chosen granularity must be an integral power of 2, not just a multiple of 256. For example, 512 is

a legal value, but 768 is not.

• Although a granularity must be an integral power of 2, memory can be assigned in any multiple of

that value. For example, if the SLM granularity is 256 MB, it is legal to assign 768 MB of SLM to a

vPar.

• Itanium-based platforms have a platform-dependent limit to the number of SLM granules per blade or

ILM granules that may be conﬁgured. You can determine speciﬁc limits for your installation by using

the

parstatus command.

• User speciﬁed memory granularity value can be overridden by ﬁrmware because of certain restrictions

imposed by the distribution of memory among individual blades. In this case, memory granularity

chosen by ﬁrmware is used to round up the memory sizes assigned to individual vPars in the nParti-

tion. The vPar database is also updated to reﬂect these memory size adjustments, unless user has

explicitly disabled parspec change policy attribute for the nPartition.

• User also should take into consideration the maximum number of ILM and SLM granules per nParti-

tion supported for the platform. The number of memory granules possible in the nPartition with the

user speciﬁed memory granularity value and the total memory in the blades assigned to the nPartition

should be less than the maximum number of ILM (or SLM) granules per nPartition supported for the

platform. The user speciﬁed memory granularity gets automatically adjusted if the value speciﬁed is

not in conformance with the above rule, unless user has explicitly disabled parspec change policy

4 Hewlett-Packard Company − 4 − HP-UX 11i Version 3: September 2010