Specifications

Executive Summary
The evolution of cloud computing has
resulted in highly efcient and carefully
optimized data centers with increased
server density and capacity that makes
considerations on energy consumption
and utilization extremely critical along
with several other factors that were
not as signicant in smaller data centers
of the past. To support this evolution,
Intel works with end users to create
an open data center roadmap of usage
models that address key IT pain points
for more secure, efcient, and simple
cloud architectures built on a foundation
of transparency. This paper describes
an Energy Management reference
architecture based on Dell, Intel, and
EDCM* (Beijing ZZNode Technologies Co.,
Ltd.) solutions with usage models aimed at
data center power efciency and optimal
utilization of provisioned power and
cooling capacity.
The goal of energy management usage
models is to optimize productivity per
watt in order to reduce total cost of
ownership (TCO). Requirements include
the capability to monitor and cap power
in real-time at server, rack, zone, and data
center levels. This means the ability to
monitor and manage aggregated power
consumption within a rack, zone, or data
center based on available power and
cooling resources
In this reference architecture we used
Dell PowerEdge C-Series Servers with
Intel® Intelligent Power Node Manager
1
(Intel Node Manager) and EDCM Energy
Management Software
2
which uses
Intel® Data Center Manager
3
(Intel DCM)
to provide data center energy efciency
through real time power monitoring of the
servers, power capping and policy based
energy management.
We describe the following energy
management use cases in detail along
with experimental results and data.
1. Real-time Server Energy Usage
Monitoring, Reporting and Analysis
to get continuous and actual
energy usage visibility via agentless
monitoring of the servers along with
other devices and systems in the
enterprise network, data center and
facilities. The actionable reporting
and analysis with real-time power
monitoring enables reduction in
energy cost and carbon emissions.
2. Power Guard Rail and Optimization
of Rack Density by imposing power
guard to prevent server power
consumption from straying beyond
preset limit. The deterministic power
limit and guaranteed server power
consumption ceiling helps maximize
server count per rack and therefore
return of investment of capital
expenditure per available rack power
when rack is under power budget
with negligible or no per server
performance impact.
3. Disaster Recovery/Business
Continuity by applying significantly
lower power caps to lower power
consumption and heat generation
when unforeseen circumstances
like power outage and cooling
system failure occurs. In these
scenarios it may be appropriate to
set aggressively lower power caps
to though performance would be
affected. The use case illustrates
how it works at a data center location
or a group of servers.
4. Power Optimized Workloads to
achieve power efficiency. Workload
profiles are built and a maximum
performance loss target set.
Experiments determine how much
capping can be applied before
the performance target is hit.
The approach is to match actual
performance against service level
requirements. For workloads that
were not processor intensive, we
were able to optimize server power
consumption by approximately
20 percent without an impact on
performance. For workloads that
were processor intensive, for the
same 20 percent power saving,
we saw an 18 percent decrease in
performance. For a 10 percent power
reduction, performance decreased by
14 percent.
5. Data Center Energy Reduction
through Power Aware Support for
Multiple Service Classes showcases
the ability to enforce multiple SLAs
across different populations of users
with different priority workloads.
Workloads that ran over a period of
eight hours realized 25 percent less
energy consumption.
The paradigm of cloud computing brings
opportunity for data center efciency.
Energy management usage models
addressed here can substantially help to
meet power management requirements.
EDCM Energy Management Solution
can manage a wide range of devices
and systems in the data center to
reduce energy cost; however this paper
focuses on its usage models on servers,
specically Dell PowerEdge C-Series
servers with Intel power management
technologies.
Introduction
Cloud computing is the new model for IT
services that has emerged to break the
trend of decline in exibility combined
with increase in costs. It is an approach
to computing that uses the efcient
pooling of an on-demand, self-managed
infrastructure, consumed as a service.
This approach extrapolates applications
and information from the complexity
of underlying infrastructure, so IT can
support and enable business value. In
concert with Dell, Intel, and other industry
leaders, EDCM helps reduce energy costs
in cloud data centers with its innovative
agentless energy management solutions.
4
Intel® Cloud Builders Guide: Data Center Energy Management with Dell, Intel, and ZZNode