- Integrated Systems
- About Us
- Integrated Systems
- About Us
03-28-2013 08:35 AM
Issue with Onboard Administrator power limits for C7000 DC power supplies
Question from David and his customer:
From my customers perspective he believes the OA is adding ~17% padding on top of the 100% max utilization power for the power and on the system it is not consistent by slot for the same hardware. This is up to a 20% difference between slots equipped with the same hardware. He believes this seems excessive for power loss based on slot locations and the hardware should not have more than a 5% difference.
He mentions this because he desires to use N+N in a DC C7000. The configuration does not have enough power in N+N and they require this since they only have a A and B source.
1. Is there any way to limit OA power to be able to use N+N configurations with 16 blade configurations?
2. Is there any plans on increasing output power in DC power supplies?
His note to me…..
We have a user that has exceeded the OA calculated power allocation for a DC based c7000 shelf in a redundant 3+3 mode (6750 Watts) for two blades (the OA is calculating a need for an addition 635 Watts) of the sixteen blades (only fourteen will poweron).
They have the following installed.
- Two OA Modules
- Ten Cooling Fans
- Six DC Power Supplies (redundant 3+3 mode)
- Two ProCurve 6120XG
- Two ProCurve 6120G/XG
- Sixteen BL460c Gen8 Blades (Two E5-2670, Sixteen 8GB DIMMs, Two 600Gb Disks)
Per the HP Power Advisor, the above Blade should not consume more than 330 Watts at 100% Utilization (Max Circuit Size of 390 Watts) with AC Gold Power Supplies (Cannot install DC Supplies in Tool). The OA budgeting is all over the place for identical blades (there cannot be that large of difference between slots).
I need help ASAP in making this configuration work for a 3+3 power redundant mode. Is there a way to push the OA budget down to allow all sixteen blades to poweron?
Help from Monty:
The Onboard Administrator does not add any watts to the server blade power allocation. The power allocation to a server blade is based on information programmed into the server base board FRU eeprom at the factory. This value is typically much higher than actual maximum power consumption based on the configured CPUs, memory and options. We developed a power calibration for c-Class server blades at launch in 2006 which is calculated by the server BIOS at the beginning of POST and communicated to the OA through the iLO when POST is complete. The OA then updates the power allocated to each server based on that power calibration. This explains the power allocation difference between each server blade – the largest power consumption variables are the differences in the power consumption of each CPU and DIMM.
The power allocated to each interconnect bay is only based on the FRU eeprom value for that interconnect module.
The power allocated to the fan subsystem is based on the maximum power consumption of the fan subsystem for enclosure cooling at max ambient temperature.
The OA Power Limit and Enclosure Dynamic Power Capping do not increase the amount of power available from the power supplies – they only decrease the enclosure Power Capacity so that an enclosure can share a circuit with other devices and not trip the circuit breaker.
I am surprised that Power Advisor only indicated 330W for the configured server blade. I suspect this was not the same configuration as the servers listed below.
Now to answer your numbered questions:
- Yes – Power Limit or Enclosure Dynamic Power Capping can be used to limit the power consumption of an entire enclosure. But no – this does not increase the enclosure Power Capacity – therefore no more servers will be powered on.
- I’m not aware of any plans for new c7000 DC Power Supplies.
The printout below indicates the enclosure requires 7385W of power allocation – and you only have 6750W of power in N+N Redundant mode with three 2250W DC supplies.
Are the redundant DC inputs to this enclosure from independent battery banks?
Any other suggestions?