Cisco Cisco UCS C22 M3 Rack Server Libro blanco
© 2014 Cisco and/or its affiliates. All rights reserved. This document is Cisco Public.
Page 1 of 15
White Paper
Cisco UCS C-Series M4 Server Power
Management
Management
August 25, 2014
Sriranjan Bose, Technical Leader, Cisco Data Center Group
Chitkala Sethuraman, Software Engineer, Cisco Data Center Group
Vikram Bodireddy, Software Engineer, Cisco Data Center Group
This white paper provides an overview of the features and best practices for the new
advanced power management capabilities available with Cisco UCS
advanced power management capabilities available with Cisco UCS
®
C-Series M4
rack servers.
Introduction to Power Capping
Power capping provides the ability to limit the power consumption of a system, be it a blade server or a rack-mount
server, to some threshold that is less than or equal to the system’s maximum rated power. The feature typically
server, to some threshold that is less than or equal to the system’s maximum rated power. The feature typically
offers incremental benefits with the ability to scale a greater number of systems, as well as providing even better
control over power consumption.
There are two major types of power capping. Static power capping involves distributing a fixed budget of power
across multiple servers by calculating the individual system consumption using its rated or nameplate power rating.
For example, if the maximum power rating of a server is given as 340 watts (W) and a rack in a data center is
equipped with 10 servers, but the power available to the rack is only 3100W AC, the available power is sufficient to
supply an average of 300W per rack server. In this case, each server can be capped at a maximum of 300W to
avoid exceeding the capacity of the power supply. An obvious drawback of this approach is the inconsistency of
power ratings across rack servers. This rating depends on the configuration of the server, affected by the types of
DIMMs, CPUs, and adapters the system is equipped with. Another drawback is that the power consumption of the
systems in the rack may not be peaking at the same time. So the server power consumption will be capped even
when the overall power budget has not reached full capacity.
5