Cool Spots Yuvraj Agarwal CSE UCSD Trevor Pering

Cool. Spots Yuvraj Agarwal, CSE, UCSD Trevor Pering, Intel Research Rajesh Gupta, CSE, UCSD Roy Want, Intel Research

Motivation: Wireless Power Is a Problem! Depending on the usage model, the power consumption of emerging mobile devices can be easily dominated by the wireless interfaces! Power breakdown for a fully connected mobile device in idle mode, with LCD screen and backlight turned off. Cool. Spots

Opportunity: Devices With Multiple Radios Many devices already have multiple wireless interfaces… • • • PDA’s HP h 6300 (GSM/GPRS, BT, 802. 11) Mobile Phones - Motorola CN 620 (BT, 802. 11, GSM) Laptops (Wi-Fi, BT, GSM, …) These radios typically function as isolated systems, but what if their operation was coordinated to provide a unified network connection? Cool. Spots

Properties of Common Radio Standards Higher throughput radios have a lower energy/bit value … have a higher idle power consumption …and they have different range characteristics! Cool. Spots

Low-power Access Within a Wi. Fi Hot-spot Mobile Device (e. g. , cell-phone) Wi-Fi Hot. Spot Cool. Spots

Your entire house would be covered by a Wi. Fi Hot. Spot… Your TV would be a Bluetooth-enabled Cool. Spot! Cool. Spots

Inter/Intra Technology Power Management Cool. Spots Bluetooth Wi-Fi Wi. Fi Active BT Sniff Wi. Fi Active BT Active 5. 8 m. W 81 m. W Wi. Fi Active PSM 264 m. W Wi. Fi Active 990 m. W Cool. Spots implement inter-technology power management on top of intra-technology techniques to realize better power & performance than any single radio technology. Cool. Spots

Cool. Spots Network Architecture 5 4 Backbone Network Access point changes routing table on “switch” message from mobile device Cool. Spot Access Point BT Wi. Fi 1 Low-power Bluetooth link (always maintained, when possible) BT Wi. Fi Mobile Device IP address on Backbone Subnet Cool. Spots Switching is transparent: applications always use the IP address of the local subnet. Infrastructure Computers 3 Wi. Fi link is dynamically activated based on switching determination 2 Mobile device monitors channel and implements switching policy

Switching Overview Three main components contribute to the behavior of a multi-radio system: where, what, and when Position: Where you are • Need to address the difference in range between Bluetooth and Wi. Fi Benchmarks: What you are doing • Application traffic patterns greatly affect underlying policies Policies: When to switch interfaces • A non-intrusive way to tell which interface to use Cool. Spots

Where: Position Bluetooth and Wi. Fi have very different operating ranges! (approx. 10 m vs. 100 m) • Optimal switching point will depend on exact operating conditions, not just range • Experiments and (effective) policies will measure and take into account a variety of operating conditions Position 1 Bluetooth channel capacity depends on range, so the further away you are, the sooner you need to switch… Position 2 In some situations, Bluetooth will not be functional and Wi. Fi will be the only alternative Position 3 Cool. Spots Base Station

What: Benchmarks Baseline: target underlying strengths of wireless technologies • Idle: connected, but no data transfer • Transfer: bulk TCP data transfer Video: range of streaming bit-rates varying video quality • 128 k, 250 k, 384 k datarates • Streaming data, instant start WWW: realistic combination of idle and data transfer conditions • Idle: “think time” • Small transfer: basic web-pages • Bulk transfer: documents or media Cool. Spots

When: Policies The switching policy determines how the system will react under different operating conditions bluetooth-fixed (using sniff mode) cap-static-X time > Y kbps > X bandwidth-X Use Bluetooth Channel Cool. Spots kbps < Z cap-dynamic time > Y kbps < X wifi-fixed (using PSM) Z = kbps Use Wi. Fi Channel wifi CAM (normalization baseline)

Experimental Setup • Characterize power for Wi. Fi and BT – – – Multiple Policies Different locations Suite of benchmark applications Benchmark suite Test Machine (TM) • Stargate research platform – – ETH 400 Mhz processor, 64 MB RAM, Linux Allows detailed power measurement Wi. Fi is Net. Gear MA 701 CF card Bluetooth is a CSR Blue. Core 3 module • Use the geometric mean to combine benchmarks into an aggregate result • m. W BT Base Station (BS) RM Wi. Fi • Tested using “today’s” wireless: – – Data Acquisition (DA) ETH = Wired Ethernet BT = Bluetooth RM = Route Management Moved devices around on a cart to vary channel characteristics Cool. Spots Mobile Device (MD) SP Distance adjustment m. W = Power Measurements Wi. Fi = Wi. Fi Wireless SP = Switching Policy

Switching Example: MPEG 4 streaming - Simple bandwidth policy Wi-Fi - Switch from Wi. Fi to BT when application has buffered enough data Bluetooth Switch : Wi-Fi -> BT Demonstrates how switching is transparent to unmodified applications! Cool. Spots

Results Overview (Intermediate Location) • blue-fixed does well in terms of energy but at the cost of increased latency • cap-dynamic does well in terms of both energy and increased latency Cool. Spots

Impact of Range/Distance Missing data indicates failure of at least one application, and therefore an ineffective policy! Cool. Spots

Results across various benchmarks wifi-fixed consumes lowest energy for data transfer, any bluetooth policy for idle Overall, cap-dynamic does well taking into account energy and latency Video benchmarks really highlight problems with wifi-fixed and bandwidth-x Cool. Spots

Cap-Dynamic Switching Policy • Switch up based on measured channel capacity (ping time > Y) • Remember last seen Bluetooth bandwidth (Z=kbps) • Switch down based on remembered bandwidth (kbps < Z) cap-dynamic policy time > Y Z = kbps < Z Cool. Spots

Switching Policies – Analysis • “Wifi-Fixed” Policy (Wi. Fi in Power Save Mode) – – Works best for as-fast-as-you-can data transfer Higher power consumption, especially idle power • “Blue-Fixed” Policy • “Bandwidth” Policy • “Capacity-Static” Policy • “Capacity-Dynamic” Policy – Very low idle power consumption – Increases total application latency, fails at longer ranges – Static coded bandwidth thresholds, fails to adapt at longer ranges – Switches too soon (bandwidth-0) or switches too late (bandwidth-50) – Estimates channel capacity and uses that to switch up – Fails at longer ranges due to incorrect switch-down point – Dynamic policy, remembers the last seem switch-up bandwidth – Performs well across all benchmarks and location configurations! Cool. Spots

Conclusions • A dynamic system can leverage the different underlying radio characteristics to reduce communication energy while still maintaining good performance • Advanced policies can adapt well to changing operating conditions – Application behavior – Radio link quality • Evaluation of Cool. Spots policies shows around a 50% reduction in energy consumption over the present power management scheme in Wi. Fi (PSM) across a range of situations Cool. Spots

Thank you! Questions? Cool. Spots