Architecting for Green Computing

Posted by Abel Avram on Dec 27, 2008

Sections

Operations & Infrastructure

Topics

Architecture ,

Green Computing ,

Deployment / Datacenter ,

Operations

Tags

Microsoft

Green computing is becoming more and more important every day. System architects need to take into account energy consumption and to find ways to reduce it through: system virtualization, server consolidation, smart unit positioning in data centers, and others.

The latest issue of The Architecture Journal was dedicated to Green Computing. Lewis Curtis, Principal Architect on the Microsoft Platform Architecture Team, divided the topic into the following 5 areas:

• Physical. The Genome research project, described in “Wireless Sensor Network for Data Center Monitoring”, uses heat distribution data from a wireless sensor network to optimize data center design and server provisioning in order to avoid overcooling the entire data center.
• Operating Platform. Hardware resources are often allocated based on a worst-case scenario that may happen with a low frequency. As a result, complete farms may be as much as 90 percent underutilized.
• Sustainable Intelligence. An Energy Usage Profile (EUP) is an essential tool for measuring energy consumption in various domains such as hardware, operating systems, users, and applications.
• Application Development. Green computing discussions today tend to focus on the platform, hardware, and data centers. However, application inefficiencies, such as suboptimal algorithms and inefficient usage of shared resources causing contentions, are originators of higher CPU usage and, therefore, energy consumption.
• The Cloud. Insofar as this these models [cloud computing] consolidate organizations, consumption has the potential to be remarkably reduced as the Internet scale data centers in which services are hosted can make efficient use of shared resources (servers, storage, cooling mechanisms, and so forth).

In an article entitled “Green Maturity Model for Virtualization”, Kevin Francis and Peter Richardson explain how to use virtualization to reduce energy consumption. They see 4 types of computing: Local, Logical, Data Center and Cloud Computing, the last offering the most advanced form of virtualization and therefore representing the greenest computing:

Cloud computing provides the next big thing in computing — some interesting architectural constructs, some great potential from a monetary aspect, and a very real option to provide a more environmentally friendly computing platform.

The lowest-hanging fruit in the transition to virtualization are in test, development, and other infrequently used computers. Moving these machines into a single virtual environment reduces the physical footprint, heat produced and power consumed by the individual servers.

In their article, “Profiling Energy Usage for Efficient Consumption”, Rajesh Chheda, Dan Shookowsky, Steve Stefanovich, and Joe Toscano analyze how much energy each hardware component consumes and compute how much can be saved by virtualization, optimization and shutting down unnecessary services. A single Dell PowerEdge M600 blade server consumes about 380W when idle and 450W during load. The total cost of energy for a year is about $300, the equivalent of 7,700 pounds of CO2. Considering that a company may have dozens, hundreds or even thousands of such servers, the energy intake becomes increasingly costly and the CO2 emissions are accordingly high. Several steps to be taken to reduce the energy consumption are:

• Understanding your hardware and its energy usage.
• Understanding your operating system and its energy usage.
• Understanding your application and its energy usage.
• Understanding the external impacts of your application.

Another approach to green computing is to analyze how energy is utilized in data centers. By getting a thermal image of a data center, one can understand which units generate more heat and which are cooler. Then units can be arranged to optimize heat generation and dispersion with a direct impact on energy consumption. That is what Microsoft’s Data Center Genome project is all about. According to an EPA report cited by “Project Genome” article, “U.S. data centers alone consumed 61 billion kWh in 2006 — enough energy to power 5.8 million average households.”, and the amount is to double by 2011, so it’s understandable why data center’s energy consumption needs to be understand and reduced.

Green computing is a necessity these days and it will become increasingly important in the future when energy consumption levels will raise increasing the CO2 emissions related to global warming according to many scientists. Computer architects will play an important role in providing a greener computing and therefore a better life on planet Earth.

• Older >