Data Compression and Energy Consumption on a WirelessNetworked Handheld Computing Device

1
Data Compression and Energy Consumption on a
Wireless-Networked Handheld Computing Device

• Daniel OMalley
• Supervisor Prof. Zhiyuan Li

2
Handheld Computing Devices

• Challenges
• Few Computational Resources
• Small Form Factor
• Battery Life
• Mobility Requirements
• Input, Output Mechanisms

3
Handheld Computing Devices

• Benefits
• Convenience
• Similar functionality to a desktop computer
• Always have important data with you
• Mobility
• Laptops can be too big
• Can be kept in a pocket

4
Wireless Networking With Handheld Computers

• Bluetooth
• Small range
• Lower power usage, slower speeds
• 802.11
• Wider range
• Higher power usage, faster speeds
• How do we reduce the power usage?

5
A Hardware Approach

• Use different energy modes in hardware
• Reduces average energy consumption
• Reduces transfer speeds

6
A Software Approach Data Compression

• Compress the data on the server
• Transfer compressed data to the handheld
• Decompress data on the handheld
• Previously shown to be effective on Linux
• Does the same hold for Windows CE?

7
Compression Algorithms

• LZW (ncompress)
• Fast, low compression ratios
• LZ77 (gzip, zlib)
• Somewhat fast, somewhat high compression ratios
• Burrows-Wheeler (bzip2)
• Slow, high compression ratios

8
Equipment Used

• Compaq iPAQ 3630
• Backpaq with 11Mbit WiFi card (802.11b)
• HP 3458a Digital Multimeter
• Linux Workstation
• Acts as server to handheld
• Windows Workstation
• Communicates with Multimeter

9
Transfer, Compression Schemes

• Download compressed the entire file. Decompress
  with gzip, ncompress, bzip2, or zlib
• Interleave the download and decompression using
  threads.

10
Three Zlib Methods

• zlib 1
• Download compressed file, and write it to the
  file system. Decompress the file on the file
  system.
• zlib 2
• Download compressed file, and keep it in memory.
  Decompress the file that is in memory.
• zlib with interleaving
• Create a thread to download the file, and a
  thread to decompress the file.
• The decompression thread runs when the download
  thread is blocking.

11
Average Energy Consumption Ratio
12
Disparity between gzip and zlib 1

• The same algorithm and scheme is used.
• Gzip, bzip2, ncompress on Windows CE
• Use CElib
• CElib introduces overhead into many function
  calls.
• CElib eases porting of applications from Unix to
  Windows CE.
• Zlib on Windows CE
• Does not use CElib.
• Native library

13
Average Energy Consumption Ratio
14
Summary of Results

• LZ77 is superior to LZW and Burrows-Wheeler for
  these purposes.
• Significant power gains can be realized when a
  native version of LZ77 is used.
• Interleaving is ineffective under Windows CE

15
Work Involved

• Port bzip2, gzip, and ncompress to Windows CE.
• Write software to control the download and
  decompression.
• Write software to communicate with the
  multi-meter.
• Actually perform the experiment.

16
Future Directions

• Block-by-block adaptive compression scheme
• Can compression be used effectively to upload
  files?
• What algorithms are best suited for this purpose?
• Alternative techniques

17
Happy Birthday

• Questions