Data Center Power Consumption Can Be a FAWNy Business

I'm a sucker for a good pun. (Of course, some would say that there's no such thing as a "good" pun.) So when I came across a paper entitled FAWNdamentally Power-efficient Clusters, I had to read it. And you know what? Not only did I enjoy reading the paper, I actually learned a thing or two about a very important topic.

As a power-efficient alternative for data-intensive computing, the authors -- Vijay Vasudevan, Jason Franklin, David Andersen, Amar Phanishayee, Lawrence Tan, Michael Kaminsky, and Iulian Moraru -- propose a cluster architecture called a FAWN, short for "Fast Array of Wimpy Nodes." A FAWN consists of a large number of slower but efficient nodes that each draw only a few watts of power, coupled with low-power storage. Prototype FAWN nodes are built from 500-MHz embedded devices with CompactFlash storage that are typically used as wireless routers, Internet gateways, or thin clients. According to the authors, their preliminary evaluation results demonstrate that a FAWN can be up to six times more efficient than traditional systems with Flash storage in terms of queries per joule for seek-bound applications and between two to eight times more efficient for I/O throughput-bound applications.

Why is this important? Because, the authors point out, power is becoming an increasingly large financial and scaling burden for computing and society. The costs of running large data centers are becoming dominated by power and cooling to the degree that companies such as Microsoft and Google have built new data centers close to large and cost-efficient hydroelectric power sources. Studies have projected that by 2012, three-year data center energy costs will be double that of server equipment expenditures

They go on to say that power consumption and related cooling costs have become a primary design constraint at all levels, limiting the achievable density of data centers and large systems, and pushing processor manufacturers towards alternative architectures like multicore architectures.

There's nothing FAWNy about excessive power consumption, so take a few minutes and read this paper. It really is an important issue. FAWNdamentally Power-efficient Clusters.


Parallel Pattern 5: Stencil
All memory addresses used for reads are expressed as offsets
Distributing Work Across Cores Using .NET
A roll-your-own ThreadPool implementation
Looking For The Lost Packets: Part 2
Looking For The Lost Packets: Part 1

Real World Parallelism Webinar Series
  • February 18, 2010
    Lock Contention, Using Intel Parallel Studio to Improve Performance
    Speaker: Vasanth Tovinkere, Software Engineer, Intel Corporation (Bio)

    Vasanth Tovinkere is a software engineer in the Developer Products Division (DPD) at Intel. His current role involves defining novel approaches to understanding and visualizing parallel performance and consulting with strategic customers to help them prepare and deliver code for the multicore world. Vasanth has been involved in the development of automatic semantic event detectors for digital sports technologies in Intel Labs. He also has been awarded three patents and has two patents pending.

    Abstract:
    Discover how easy it is to use the power of Microsoft Visual Studio and Intel Parallel Studio to find performance issues due to lock contention in threaded applications. This ensures that shipped applications can take better advantage of multicore processors. In this webcast, we provide live demonstrations that show how to identify lock contentions issues with Visual Studio and Intel Parallel Studio, an add-in to Visual Studio that helps developers create fast, reliable code on multicore processors.t.