IBM sub-capacity licenses are dependent on “HW threads” instead of physical cores when calculating full-capacity license consumption. I am not seeing the hyper-threaded cores on any of my BigFix Inventory reports.
Can BigFix report the HW threads for the different SPARC processors?
Try this pulled from https://bigfix.me and was written by: @jgstew :
if (exists properties whose (it as string contains "cpupackage")) then ((thread of it*count of it) of cpupackage) else ERROR "cpupackage property not available"
or try:
Count of processorsAs @jgo points out, there are multiple inspectors that could give this info that pull from different sources, and not all sources are available on all platforms.
sums of threads of cpupackages
But that seems to only count the threads of a single CPU package. The right answer might actually be this if there are multiple CPUs:
(count of cpupackage * it) of sums of threads of cpupackages
It is hard for me to test some of these inspectors because I don’t have a lot of test systems lying around with multiple physical CPUs.
It should also be possible to read this info using the DMI/SMBIOS inspectors. On windows, the WMI inspectors should also work. On Mac, the IO Kit Registry should have this same info.
Another option you could do to confirm the results of the inspectors is to use a command that should return the correct results you expect and run that with bigfix, output the results to a file, and read that back with relevance. This is not an ideal solution, particularly for all computers, but it is a useful validation method you could use on select systems to directly compare the different options available directly with Relevance inspectors to what you see as the correct count based upon running a command. You can use this to then be more confident of the direct relevance results elsewhere.
Based upon the testing I just did, it doesn’t seem like this info is available in SMBIOS, at least not on my test system, which surprised me:
unique values of (it as string) of ((name of it | ""), ( (name of it | "") & " : " & (it as string | "")) of values of it) of structures of smbios
This relevance should return a mostly exhaustive list of what is available in SMBIOS.
At least for my VM, the number of “CPU socket” returned was way higher than the number of CPUs / Threads.
It seems to me that BigFix Inventory should already be doing this without any special scripts or commands. The Audit Snapshot report for IBM sub-capacity licensing provides a total consumption report across all platforms for each title. It cannot be accurate if it is not determining the number of HW threads per core as described in the license counting rule document found at ftp://public.dhe.ibm.com/software/passportadvantage/SubCapacity/Scenarios_Sun.pdfs.
If it is already determining the HW threads, I just need that data to appear on a report such as the Software Installations Report.
I believe your information may be outdated (the presentation you reference is from 2009). I found this link:
PVU requirement for SPARC T4 from Oracle (formerly Sun) and is dated Sep 2011.
Also, neither the current Eligible Processor technology Document nor the current PVU Table per Core for RISC processors make any mention of special counting rules for hardware processor threads.
According to the Eligible Processor Technology Document the following Oracle/Sun processors are supported by ILMT v9.0.0.0 or later:
UltraSPARC III, IV, T1, T2, T3
SPARC T4, T5.
… supported by ILMT v.9.2.6 or later:
SPARC M7 and S7
… supported by ILMT v9.2.11 or later
SPARC M5 & M6
I recommend you contact your IBM Sales rep for the officially official response, however since the foregoing is just my opinion.
I was also wondering if threads actually need to be counted, and only cores.
The IBM website “Virtualization Capacity License Counting Rules” [https://www-01.ibm.com/software/passportadvantage/Counting_Software_licenses_using_specific_virtualization_technologies.html] includes links to ten guides for “Specific Eligible Virtualization Environments”. This website is supposed to be the authority but I have contacted my account rep for confirmation.
The eligible processor document and the PVU tables do not contradict the guide specifically for Sun and Fujitsu. The Sun and Fujitsu guide further defines what are to be considered as CPUs and cores for the SPARC processors. It does appear to be unnecessarily complex.
I see what you mean. If sub-capacity for Sun requires thread counting, I don’t see how it is implemented in ILMT. And since ILMT is supposed to be authoritative for those version of the OS, it appears to be a contradiction.
I agree, ILMT is supposed to be authoritative for IBM sub-capacity.
I received confirmation from my IBM account rep that the 2011 pdf document on calculating sub-capacity licensing for SPARC processors is current.