Warning: preg_match(): Compilation failed: group name must start with a non-digit at offset 8 in /var/www/html/includes/MagicWord.php on line 592
Load balancing - Green software

Load balancing

From Green software

Revision as of 14:41, 16 September 2010 by Green (Talk | contribs)
Jump to: navigation, search

Load balancing is a technique to distribute workload evenly across two or more computers, network links, CPUs, hard drives, or other resources. In computer networking load balancing is mostly applied to avoid bottlenecks or increase throughput. Load balancing between computing nodes however has a more interesting feature: Two slower processing units can get the same work done as a single highly clocked processor but without producing as much heat and thus improving the energy efficiency.


Heat production

The higher the clock frequency, the more energy is transformed into heat that dissipates and goes unused. The amount of power that is needed to cool a CPU is called the 'thermal design power (TDP)'. When comparing TDP values,

Two CPUs that can be compared are the 'Intel Core 2 Duo E6750' and 'Pentium 4 506'. Both processors run at 2.66 GHz. The Pentium 4 has a TDP value of 84 W, and the Core 2 Duo has signifcantly lower TDP of 65 W.

Image:Http://www.few.vu.nl/~sge900/MultiCoreEE.tif from [1] shows how energy efficiency of multi-core processors relate to single-core processors.


The decrease in operating temperature of computing nodes also increases the expected lifetime of the equipment significantly<source needed>.


Load balancing can be very hard to accomplish. To use a multi-core processor efficiently, programs should be programmed in a way that they use multiple threads. Making a program use more threads is called parallel programming. Parallel programming involves taking care of communication between the different threads and splitting up the work and data appropriately. Most IT professionals have never been schooled in parallel programming.



See Also

Multithreaded programming


This best practice was recognized as such by IT professionals, described in Energy efficient software.
This best practice is mentioned in:
Petter Larsson. 2008. Energy-Efficient Software Guidelines. White Paper for the Intel Software Solutions Group.

Personal tools