S2600GZ and S2600GL

Intel® Server Board S2600GZ/GL TPS Technology Support
Table 20. Intel
®
Intelligent Power Node Manager
IT Challenge
Requirement
Over-allocation of power

Ability to monitor actual power consumption
 Control capability that can maintain a power budget to enable
dynamic power allocation to each server
Under-population of rack space
Control capability that can maintain a power budget to enable increased rack
population.
High energy costs
Control capability that can maintain a power budget to ensure that a set
energy cost can be achieved
Capacity planning

Ability to monitor actual power consumption to enable power usage
modeling over time and a given planning period
 Ability to understand cooling demand from a temperature and airflow
perspective
Detection and correction of hot spots
 Control capability that reduces platform power consumption to
protect a server in a hot-spot
 Ability to monitor server inlet temperatures to enable greater rack
utilization in areas with adequate cooling.
The requirements listed above are those that are addressed by the C600 chipset Management Engine (ME)
and Intel
®
Intelligent Power Node Manager (NM) technology. The ME/NM combination is a power and thermal
control capability on the platform, which exposes external interfaces that allow IT (through external
management software) to query the ME about platform power capability and consumption, thermal
characteristics, and specify policy directives (for example, set a platform power budget).
Node Manager (NM) is a platform resident technology that enforces power capping and thermal-triggered
power capping policies for the platform. These policies are applied by exploiting subsystem knobs (such as
processor P and T states) that can be used to control power consumption. NM enables data center power
management by exposing an external interface to management software through which platform policies can
be specified. It also implements specific data center power management usage models such as power limiting,
and thermal monitoring.
The NM feature is implemented by a complementary architecture utilizing the ME, BMC, BIOS, and an ACPI-
compliant OS. The ME provides the NM policy engine and power control/limiting functions (referred to as Node
Manager or NM) while the BMC provides the external LAN link by which external management software can
interact with the feature. The BIOS provides system power information utilized by the NM algorithms and also
exports ACPI Source Language (ASL) code used by OS-Directed Power Management (OSPM) for negotiating
processor P and T state changes for power limiting. PMBus*-compliant power supplies provide the capability to
monitoring input power consumption, which is necessary to support NM.
Below are the some of the applications of Intel
®
Intelligent Power Node Manager technology.
 Platform Power Monitoring and Limiting: The ME/NM monitors platform power consumption and
hold average power over duration. It can be queried to return actual power at any given instance. The
power limiting capability is to allow external management software to address key IT issues by setting a
power budget for each server. For example, if there is a physical limit on the power available in a room,
then IT can decide to allocate power to different servers based on their usage – servers running critical
systems can be allowed more power than servers that are running less critical workload.
 Inlet Air Temperature Monitoring: The ME/NM monitors server inlet air temperatures periodically. If
there is an alert threshold in effect, then ME/NM issues an alert when the inlet (room) temperature
exceeds the specified value. The threshold value can be set by policy.
 Memory Subsystem Power Limiting: The ME/NM monitors memory power consumption. Memory
power consumption is estimated using average bandwidth utilization information
 Processor Power monitoring and limiting: The ME/NM monitors processor or socket power
consumption and holds average power over duration. It can be queried to return actual power at any
given instant. The monitoring process of the ME will be used to limit the processor power consumption
through processor P-states and dynamic core allocation
 Core allocation at boot time: Restrict the number of cores for OS/VMM use by limiting how many
cores are active at boot time. After the cores are turned off, the CPU will limit how many working cores
Revision 2.4
55