Cisco Cisco UCS C460 M2 High-Performance Rack Server White Paper
Page
IT & DATA MANAGEMENT RESEARCH,
INDUSTRY ANALYSIS & CONSULTING
Data Center Management: The Key Ingredient for Reducing Server Power while Increasing Data Center Capacity
©2010 Enterprise Management Associates, Inc. All Rights Reserved. | www.enterprisemanagement.com
Distributed, Policy-based Management
As indicated in the beginning of this paper, a key aspect of today’s data center includes moving to a
service-oriented management model via a centralized policy management engine, coupled with intel-
ligent self-managing hardware elements. The management engine pools data center resources, interfaces
with lower-level managers (such as provisioners and element managers), and ties it all together with
policy-based management capabilities. Data center assets are combined into pools, and policies define
management “units of work,” bundling all components required to deliver an IT service into an easily
managed profile that works for both virtual and non-virtual architectures.
Cisco took the lead in this area by creating UCS “Service Profiles” as a core part of UCS Manager. A
Cisco took the lead in this area by creating UCS “Service Profiles” as a core part of UCS Manager. A
UCS Service Profile is a policy-based grouping of related hardware, including compute, storage and
networking profiles, plus VMs, all in a self-contained package that travels with the business service as
UCS Manager moves loads around the data center to maximize resource utilization and performance. In
a VM environment, Cisco has a unique ability to deliver network QoS and security policies that adhere
to, and move with, every virtual machine using their VIC (Virtual Infrastructure Card). This is inherent
in their VIC and delivered by the hardware, removing this workload from the processor, further increas-
ing performance and therefore performance/watt.
This is a key UCS advantage, as it allows complete abstraction of a service (work) from the hardware on
This is a key UCS advantage, as it allows complete abstraction of a service (work) from the hardware on
which it runs, enabling fully automated, policy-based provisioning and de-provisioning. Service Profiles
significantly reduce the excess capacity burden described above by enabling pooled burst and high avail-
ability (HA) capacity to be rapidly provisioned based on business need. It can also assist with dynamically
balancing data center workload, ensuring maximum performance from each hardware component.
In this EMA head-to-head comparison, Cisco UCS wins the architecture and management comparison
In this EMA head-to-head comparison, Cisco UCS wins the architecture and management comparison
hands-down.
Cisco and HP: A Quantitative Power Consumption Analysis
It is always useful to have deep quantitative data when comparing two vendors, particularly from a
power perspective. As indicated earlier in the paper, power savings at the server level translate not only
into direct OPEX savings, but also produce even greater benefits in terms of reduction of power plant
demand and impact on the environment.
As indicated earlier, CPU power requirements are equal between vendors, since all blade server vendors
As indicated earlier, CPU power requirements are equal between vendors, since all blade server vendors
are using the same Intel chips. This is helpful, since it removes one of the largest power consumers as a
variable when comparing vendors. The differences in power consumption therefore boil down to items
like relative airflow (and the fans required for cooling), network hardware, and so on.
Cisco commissioned a quantitative comparison of UCS and HP blade servers in March 2010, con-
Cisco commissioned a quantitative comparison of UCS and HP blade servers in March 2010, con-
ducted by Principled Technologies (PT),
5
based in Durham, North Carolina. PT is a highly respected
testing laboratory used by many industry leaders, including Cisco, Dell, IBM, Intel and many others.
PT conducted an exhaustive power comparison between Cisco UCS B200 M1 blade servers in a Cisco
PT conducted an exhaustive power comparison between Cisco UCS B200 M1 blade servers in a Cisco
UCS 5108 Blade Server Chassis and HP ProLiant BL460c G6 servers in an HP BladeSystem c7000
Enclosure, and the entire test is available on the PT website.
6
The PT test measured overall power
5
http://www.principledtech.com
6
http://www.principledtechnologies.com/clients/reports/Cisco/UCSPower0310.pdf