HMC

Hello,

I know HMC 10.3.1060.0  introduces a nice dashboard to see power consumption and carbon footprint based on the overall complete server.But do we have any option in it so that these metrics can also be seen based on LPAR level? The reason is that there are multiple clients hosted on one physical server so the requirement is to find out usage and carbon footprint for each client.

------------------------------
Rohit Chauhan
Senior Technical Specialist
Norway
------------------------------

Hi Rohit, 

It sure is a nice dashboard, but a good question. Unlike the Performance Dashboard, far as I know you cannot see the power and carbon footprint metrics for individual LPARs on a selected System on the Environmental Dashboard nor in any exported data. 

When exporting metrics data 'By tier' from the Environmental Dashboard, it generates a bunch of data in either JSON or CSV format that includes the properties and performance metrics for each LPAR on the selected system. But no EnergyPowerUtil.powerReading like it does if you export the data 'By source' for the selected system.    

I wonder if it is not possible to apportion power usage and carbon footprint metrics to individual LPARs.  Unless you can indirectly apportion a system's power usage to the resident LPARs based on an LPAR's CPU utilisation... ?  

It also seems misleading where when you export data 'By tier', the resultant file name starts and ends with 'ByTier_EnergyMetrics_ManagedSystem......._LPAR.csv'.  Unless I am missing something I don't see any energy metrics in the data.

EDIT - After I first posted the above, I asked the following question on our enterprise Microsoft Co-pilot:

Q - Is there a way to estimate LPAR power usage based on system-level data?

A - Yes, you can estimate the power usage of individual LPARs based on system-level data using a few different approaches. Here are some methods:

1. CPU Utilization Metrics: By using CPU utilization metrics such as Processor Utilization Resource Register (PURR) and Scalable Processor Utilization Resource Register (SPURR), you can estimate the power consumption of each LPAR. These metrics help track how much CPU time each LPAR is using, which can be correlated with power usage (1).
2. Modeling and Estimation Techniques: Research has been conducted on modeling energy consumption at the LPAR level. For example, a study on IBM POWER9 systems demonstrated a method to estimate LPAR energy consumption with high accuracy by using utilization metrics and comparing them with instrumented energy measurements (2).
3. Weight-Based Allocation: Each LPAR has an assigned weight that determines its share of the system’s resources. By understanding the weight and the overall system power consumption, you can estimate the power usage for each LPAR based on its relative weight (3).

These methods can provide a good estimate of LPAR power usage, though for precise measurements, dedicated power monitoring tools or software might be necessary.

Links for the above

(1)  Rethinking CPU utilization with PURR/SPURR on an IBM Power LPAR

(2) Modeling and Estimation of LPAR Energy Consumption for IBM POWER9 Systems for MASCOTS 2023 - IBM Research

(3) https://pivotor.com/library/content/Chapman-LPARConfigs_Share202403.pdf

Perhaps IBM also have a tool available that use the above methods.  

------------------------------

Regards,

Jozsef Torok
IT Engineer - IBM i and Power Systems
Spark New Zealand Trading Ltd
------------------------------