1. Wireless Infrastructure: Networks and Issues (2) H. Scott Matthews February 26, 2003

2. Recap of Last Lecture  Wireless generally a ‘radio’ technology  Dependent on antennas (cell sites)  Cell sizes getting generally smaller  Spectrum allocation has become an increasingly complex problem as there have been more demands for it (FCC)  Number wired/wireless users ~ equal

3. FCC’s involvement  In telecom, the government tends to regulate the devices not the network  E.g. licenses spectrum for use  Certifies devices (e.g. phones) compliant  Industry/professional groups (e.g. IEEE) generally set equipment/network standards

4. Management Metrics  Different type of problem since networks are generally private  Subscribers  Number, growth, net additions  Voice quality  Time to login to system, call access time  Percent completed calls or call failure rate  Coverage area  Percent of US, Percent of population,..  Financial  Margin (profit) per minute, subscriber  Others?

5. Wireless ‘Data’ Networks  IEEE 802.11b (used on campus)  11 Mbps, using 2.4 GHz spectrum (unlicensed!)  14 channels, 2.4 to 2.4835 GHz (80 MHz)  Different channels legal around world, only chan. 1, 6, and 11 have no overlap  Designing a big network means reusing channels and considering overlaps  Usually uses PC cards, access points + wired  Industry group (WiFi alliance) certifies products  802.11a: 54Mbps @ 5 Ghz, 12 channels no overlap - 500 MHz of frequency  802.11g backwards compatible with 802.11b, but boosts speed to 54 Mbps

6. CMU Campus Wi-Fi Network  CMU campus: ubiquitous wired, wireless networks  Every room on campus ‘wired’, every space ‘wireless’  10,000 users; 350 wireless antennas (about 30 users each)  How much electricity used?  Functional, but not equivalent, comparison  Show energy “to network 10,000 users wired/wireless”  Only ‘network’ - not ‘attached devices’ - in boundary

7. Campus Network Model 120 Wiring Closets Office/room equipment 350 Wireless Antennas Main computer center

8. Two Data Sources  Campus has building-level electricity meters installed  Several buildings have more than one meter when areas have higher than average use  Used for “Main computer center electricity”  Not so useful for electricity of room/equipment  Portable power meters to measure electricity use of pieces of equipment  Measure one of each, scale up via inventory

9. Summary of Estimates  Network electricity 6% of total campus - 1.7 kWh/ft 2  Wireless endpoints use 10x less electricity than wired  Caveats: speeds, installation and maintenance requirements different  Wireless speed bump coming (10x) but electricity use expected go up only 50%  Relevance: more voice wireless than wired in the world

10. Overall Voice Network Elec  Do similar analysis, estimate PSTN and wireless voice network electricity use  PSTN = Public Switched Telephone Net  Consider number and kW of cell sites  Total energy use of sector, etc.  Get estimate of 30 TWh/yr  < 1% of US electricity consumption

11. Other Issues  Ad hoc = latin for ‘for this (time)’  Ad hoc networks are temporary, maybe one use systems  Difference in use and design of networks  Don’t have to be ‘operating’ all the time  ‘Beaming’ with palm pilots is an example  New ‘Bluetooth’ devices will be too  Useful for sensor networks (coming soon!)  Issues with designing/managing ad hoc?

12. See publicinternetproject.org For details, more research

13. Open/Public Wireless Nets  Example of more formalized/larger ad hoc networks (not fully “ad hoc”)  Campus wireless is not an example because you need to be registered to use  Communities building small-medium wireless networks with their own broadband connection and wireless points (hotspots)  Could have ‘network name’ commonality but no password/authentication/registration  There are people who drive around looking for ‘open’ wireless networks just for fun  Note: these guys need more work / less free time

14. Implications of Open Nets  Coordination (e.g. same network name)  Security!  Preventing questionable traffic  Hacking/cracking/spamming  Leeching (free rider problem)