Featured post

Top 5 books to refer for a VHDL beginner

VHDL (VHSIC-HDL, Very High-Speed Integrated Circuit Hardware Description Language) is a hardware description language used in electronic des...

Monday, 3 March 2014

Do U Know? Mobile devices said to consume more energy on storage tasks

28058-clipart-illustration-of-a-battery-mascot-cartoon-character-flexing-his-arm-musclesDo u know that “Flash storage takes power to write - a 20 volt jolt to each cell - but needs almost none to maintain. The real power hog is the inefficient storage software stack that eats 200 times the power required for the hardware.”

Given the always-on mobile infrastructure - background updates, instant messages, email, updates, file sync, logging and more - lots of background storage I/O is happening all the time. And it's eating your device's power budget.

Researchers from Microsoft and the University of California at San Diego benchmarked how Android and Windows RT mobile devices used energy for storing data. They focused on activities that occur with the screen off, since displays are a major power consumer when lit. "Measurements across a set of storage-intensive micro benchmarks show that storage software may consume as much as 200x more energy than storage hardware on an Android phone and a Windows RT tablet," the research team wrote in a paper. "The two biggest energy consumers are encryption and managed language environments."

Results
On Windows RT they found that the OS/CPU/DRAM overhead was between 5 and 200 times the power used by the flash storage itself, depending on how DRAM power use was allocated. File system APIs, the language environment and encryption drove the CPU power consumption during I/O. Full disk encryption - protecting user data - incurred 42 percent of CPU utilization.

On an Android phone, the encryption penalty is even worse: 2.6–5.9x more energy per KB over non-encrypted I/O.

For applications, the team found that on Windows RT, the energy overhead in a managed environment is 12.6–18.3 percent while overhead on Android is between 24.3–102.1 percent. It appears that Android's algorithms are not optimized for application I/O power efficiency.