Power efficiency discussion

[chuckk281]

Francisco is working with a customer on power supply efficiency:

**************

Hi experts, we have a customer making deep studies about power efficiency (they buy thousands of servers every year so this is a critical matter to them) and they have found a discrepancy in the ilo power meter:

The hardware is a C7000 with n+n redundancy; They increment the load in the processing needs and they make a metering in both the iLO and the PDU. What they have found is when load begins to rise, ilo has a zone where it says the power is decreasing, meanwhile the PDU shows that it is indeed rising. You can see the table below. From a certain point onwards, ilo and PDU measures are again aligned.

What could be happening? Is this a normal behavior (i.e. is ilo really not accurate in it’s readings?)

 

On the other hand they are asking for deep technical documentation about how do our powers supplies work in C7000:

-Their redundancy possibilities an % of maximum load for each configuration

-Thresholds when a  Power supply is switched on and off

-Switching times in case of failure, peaks that result, etc

Do we have such documentation updated for the new platinum power supplies?

***********

Reply from Monty:

*************

HP c-Class BladeSystem the ProLiant server iLO only measures the server power input.  The iLO power report is accurate and can be used to limit the power consumed by a server with power capping.

The OA provides readings and history of the entire enclosure power output and power input.

With N+N redundancy – the enclosure is connected to two different PDU – and you would need to combine the readings of both those PDU to view total enclosure power input. 

What is missing from the iLO report of power consumption is the power in every other server bay, power in all the interconnect bays and the power consumed by all the enclosure fans.

Here are answers to your other questions:

-Their redundancy possibilities an % of maximum load for each configuration     

For N+N redundancy – the OA allocates up to 50% of all the supplies capacity.  All the active power supplies share the total enclosure load

-Thresholds when a  Power supply is switched on and off

The OA has a proprietary algorithm for the enclosure Dynamic Power Savings feature.  This algorithm is designed to maximize the entire enclosure power efficiency through a wide range of enclosure loads.  When Dynamic Power Savings is enabled, the OA disables and enables pairs of power supplies in order to operate the remaining supplies at their highest efficiency.  For more details – see the c7000 Enclosure technology brief http://h20000.www2.hp.com/bc/docs/support/SupportManual/c00816246/c00816246.pdf section on Dynamic Power Saver mode on page 28

-Switching times in case of failure, peaks that result, etc

When Dynamic Power Savings is enabled, the power supplies that are in standby are immediately enabled if a different power supply fails.  In addition – the OA proprietary algorithm ensures that only half the power supplies that are enabled would be required if the AC input on one side failed – to prevent loss of power to the enclosure.

Do we have such documentation updated for the new platinum power supplies?

                The c7000 enclosure tech brief describes the different power supplies available for the c7000 enclosure – with a link to the current enclosure QuickSpecs.

***************

Comments or questions?