Power Management - Dell PowerEdge M1000e Technical Manual

Dell
5.1.3

Power Management

Power is no longer just about power delivery, it is also about power management. The M1000e
System offers many advanced power management features. Most of these features operate
transparently to the user, while others require only a one time selection of desired operating modes.
Shared power takes advantage of the large number of resources in the modular server, distributing
power across the system without the excess margin required in dedicated rack mount servers and
switches. The M1000e has an advanced power budgeting feature, controlled by the CMC and
negotiated in conjunction with the iDRAC on every server module. Prior to any server module
powering up, through any of its power up mechanisms such as AC recovery, WOL or a simple power
button press, the server module iDRAC performs a sophisticated power budget inventory for the
server module, based upon its configuration of CPUs, memory, I/O and local storage. Once this
number is generated, the iDRAC communicates the power budget inventory to the CMC, which
confirms the availability of power from the system level, based upon a total chassis power inventory,
including power supplies, iKVM, I/O Modules, fans and server modules. Since the CMC controls when
every modular system element powers on, it can set power policies on a system level.
In coordination with the CMC, iDRAC hardware constantly monitors actual power consumption at
each server module. This power measurement is used locally by the server module to insure that its
instantaneous power consumption never exceeds the budgeted amount. While the system
administrator may never notice these features in action, what they enable is a more aggressive
utilization of the shared system power resources. Thus the system is never ―flying blind‖ in regards
to power consumption, and there is no danger of exceeding power capacity availability, which could
result in a spontaneous activation of power supply over current protection without these features.
The system administrator can also set priorities for each server module. The priority works in
conjunction with the CMC power budgeting and iDRAC power monitoring to insure that the lowest
priority blades are the first to enter any power optimization mode, should conditions warrant the
activation of this feature.
Power capping is set at the chassis level for our blade servers and not at the blade server level, so
components like processer, memory can throttle down when necessary on lower priority blade
servers. An allocation is taken out for the infrastructure (fans, IO modules) and then the remainder is
applied to the blades, and then throttling is applied if required to get under the cap. If all the blades
are setup with the same priority, then it will start throttling down processor, memory, and so on. A
variety of BIOS settings will throttle the processor or not depending on load:
If power consumption demands exceed available power, the enclosure ―throttles‖ back the power
supplied to blades as prioritized in the CMC. The blades will not shut down; rather they will slow
down if necessary; Dell designed the system this way on purpose, in response to customer feedback
that they did not want the blades to shut themselves down under any condition. I/O modules, on the
other hand, will shut down prior to permanent damage, as they are less tolerant to power variation
than the blade server hardware.
The M1000e is compliant with the PMBus Specification 1.1, using this power management standard
for status, measurement and control. The M1000e power supplies continuously monitor AC input
current, voltage and power, enabling exposure of data to Dell™ OpenManage™ IT Assistant or to other
enterprise-level management tools. Real time power consumption is viewable per system.
PowerEdge M1000e Technical Guide 29