GPFS FPO weighting algorithms

Due to processing by IBM, this request was reassigned to have the following updated attributes:
Brand - Servers and Systems Software
Product family - IBM Spectrum Scale
Product - Spectrum Scale (formerly known as GPFS) - Private RFEs
Component - Product functionality

For recording keeping, the previous attributes were:
Brand - Servers and Systems Software
Product family - IBM Spectrum Scale
Product - Spectrum Scale (formerly known as GPFS) - Private RFEs
Component - V4 Product functionality

A configurations where disks with different IOPS profiles are mixed in the same storage pool is inherently problematic, and should be avoided if at all possible. An attempt to deal with the problem through FG assignment introduces significant extra complexity, but does not provide a substantial enough benefit to justify the effort.

1. for different disk capacity(e.g. iops)

in current GPFS, there are two ways to leverage the disk capacity(e.g. high IOPS).
make these kind of disks as one storage pool and put data requiring high speed of read/write into the pool

if we put them with lower IOPS disk in one failure group, if GPFS provides the rule for user to specify more IO over this disk, this will make the allocated disk space unbalanced among these disks.

2. for uneven disk number in different failure groups, so far, one workaround is to make them even in disk number among all failure groups. for those cases in which the nsd disks can't be added or removed, we will consider how to handle this well.

no matter for 1 or 2, from another view, making disk balanced in space usage is the current rule. providing customer one way to customize the "rules" will involve some changes that might impact GPFS in shared storage mode and we need further discussion inside team.

so far, we are in planning of TL2 items and we will have some discussion in the following weeks and will update the progress here. thanks

Yong ZY Zheng

Creating a new RFE based on Community RFE #60130 in product GPFS.