• Save
Green Radio08
Upcoming SlideShare
Loading in...5
×
 

Like this? Share it with your network

Share

Green Radio08

on

  • 1,028 views

A review of ongoing efforts on dynamic power management with multiple radios

A review of ongoing efforts on dynamic power management with multiple radios

Statistics

Views

Total Views
1,028
Views on SlideShare
1,025
Embed Views
3

Actions

Likes
0
Downloads
0
Comments
0

1 Embed 3

http://www.slideshare.net 3

Accessibility

Categories

Upload Details

Uploaded via as Microsoft PowerPoint

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

Green Radio08 Presentation Transcript

  • 1. Aggressively Duty-Cycled Platforms: Embedded, Mobile & Multi-Core Some observations from the field Rajesh Gupta UC San Diego mesl . ucsd . edu Nokia Meeting, August 2008 “ Future lies in system architectures built for aggressive duty-cycling”
  • 2. New Age Computing & Communications Devices
    • New devices with markedly different usage of energy and power (than desktops and laptops)
      • 6-10X variation in power from sleep to various active modes;
      • Even larger variation in radio power, TX/RX ratio
    Power Supply Battery DC-DC Converter Communication Radio Modem RF Transceiver Processing Programmable  Ps & DSPs (apps, protocols etc.) Memory ASICs Peripherals Disk Display
  • 3. Much Wider Dynamic Range of Power/Energy Consumption
      • 6-10X variation in power from sleep to various active modes (Even larger variation in radios, TX/RX ratio)
    Stargate – Mobile Research Platform (Device is in “idle” state) packet Transmit Processing Transmit Amplifier d packet Receive Processing 50 nJ/bit 100 pJ/bit/m
  • 4. Multiple Radios Are Common
    • Serving as isolated interfaces to (isolated) networks
      • Multiband radios: over wide frequency range and bw
      • Multimode radios: different protocols in a single baseband
    • Upto 10+ radios on a set
      • Cellular:
        • Quad-Band GSM/EDGE/GPRS – 850/900/1800/1900 MHz; Tri-Band HSDPA & WCDMA – 850/1900/2100 MHz, others
      • Non-Cellular WAN:
        • WiMAX: ~2.5GHz and ~3.5 GHz; also 5.8 GHz unlicensed
      • Increasing movement to license 700 MHz – 3.5 GHz:
        • E.g., WCS at 2.3 GHz, 700 MHz ‘issues’
      • Short-Range (LAN/PAN):
        • Unlicensed ISM & UNII– 2400/5100/5800 MHz
        • 928 MHz ISM band has not been used for many years
        • Mostly for short-range (WLAN, WPAN, cordless, etc.)
        • Other (Mesh, Metro WiFi, etc.) apps still speculative
      • UWB: New Short-Range unlicensed spans 3-10 GHz
  • 5. Achieving High Energy Efficiency: Lessons Learnt
    • Reduce distance
      • Physical, logical
    • Minimize wasted work
      • Shutdown, slowdown, procrastinate
    • Specialized processing
      • In a generalized execution environment
    Mult-Core DPM GreenLight Coherent Coprocessing
  • 6. Algorithmically, there are basically two ways to save power
    • Shutdown through choice of right system & device states
      • Multiple sleep states
      • Also known as Dynamic Power Management (DPM)
    • Slowdown through choice of right system & device states
      • Multiple active states
      • Also known as Dynamic Voltage/Frequency Scaling (DVS)
    • DPM + DVS
      • Choice between amount of slowdown and shutdown
    Competitive and Adversarial Approaches using Probabilistic Model Checking Machine Learning Techniques Convex Optimization for Thermally Efficient Multi-Cores
  • 7. Architecturally: ‘Collaborate’
    • Exploit the wide dynamic range of power consumptions (against capabilities)
    • Duty cycle the more power consuming resource using the other
  • 8. Collaborating Radios Can
    • Improve Performance
      • Aggregate connectivity
    • Improve Reliability
      • Radios as backup interfaces
    • Improve Security
      • Multiple/Side-Channel Authentication
    • Improve Efficiency (Spectral, Energy)
      • Dynamically match radios to traffic, range
      • Use radios to page another, duty cycle other radios
    • Collaborating radios have a great potential for system-wide improvement
      • Energy, mobility management, capacity enhancement, channel failure recovery, networking, security, ….
      • We focus on energy.
  • 9. Our Work
    • CoolSpots
      • Use BT radios as paging radios to WiFi
    • SwitchR
      • Multiple radios without base-station modifications to improve energy/bit, idle power, throughput
    • Cell2Notify
      • Use Cellular radios for event notification to WiFi
    • SoftSpeak
      • Mitigate impact of VOIP traffic over WiFi infrastructure
    • Somniloquy
      • Use network interface smarts to duty-cycle computers
  • 10. Collaborating Radios can improve efficiency, reliability,…
    • 50% energy reduction with CoolSpots
    • VOIP with Cell2Notify can reduce power 1.7-6.4x over WiFi and better than Cellular radios!
    Switch : Wi-Fi -> BT Bluetooth Wi-Fi
  • 11. GreenLight: Putting Machines To Sleep Transparently Somniloquy enables servers to enter and exit sleep while maintaining their network and application level presence. Peripheral Laptop Low power domain Network interface Secondary processor Network interface Management software Main processor, RAM, etc
  • 12. Takeaways
    • Energy efficiency is a system level concern
      • That is dealt with a coordinated strategies across processing, communications and networking
    • Algorithmically we look for the right combination of slowdown and shutdown strategies
      • Driven by increasingly real, accurate and timely sensor data that push the available slack to thermal limits
    • When all low power design tricks have been exhausted, duty cycling remains the source of continuing improvements in efficiency
      • By continually reaching to the higher levels of decision making: CAPTURE INTENT
  • 13. Recent Publications
    • SwitchR: Reducing System Power Consumption in a Multi-Client Multi-Radio Environment       - Yuvraj Agarwal, Trevor Pering, Roy Want, Rajesh Gupta, IEEE Symp. On Wearable Computers, 9/8
    • "A Gateway Node with Duty-Cycled Radio and Processing Subsystems for Wireless Sensor Networks", Zhong Yi Jin, Curt Schurgers, R. Gupta, ACM Trans. Design Automation in Electronic Systems, June 2008.
    • "Improved Distributed Simulation of Sensor Networks based on Sensor Node Sleep Time", Z. Jin, R. Gupta , 4th IEEE/ACM Intl. Conference on Distributed Computing in Sensor Systems (DCOSS) , June 2008
    • "Improving the Data Delivery Latency in Sensor Networks with Controlled Mobility", R. Sugihara, R. Gupta , 4th IEEE/ACM Intl. Conference on Distributed Computing in Sensor Systems (DCOSS) , June 2008 (Best paper for Systems track)
    • "A different approach to sensor networking for SHM: Remote powering and interrogation with unmanned aerial vehicles", Todd, M., et al , 6th Intl. workshop on Structural Health Monitoring , 2007.
    • "Temperature-Aware Processor Frequency Assignment for MPSOCs Using Convex Optimization," S. Murali, D. Atienza, G. De Micheli, R. Gupta, IEEE/ACM/IFIP CODES+ISSS / ESWeek, September 2007
    • “ An Embedded Platform with Duty-Cycled Radio and Processing Subsystem for Wireless Sensor Networks”,   Z. Jin, C. Schurgers and R. Gupta, Embedded Computer Systems: Architectures, Modeling and Simulation Conference (SAMOS), July 2007
    • "Wireless Wakeups Revisited: Energy Management for VoIP overWi-Fi Smartphones", Y. Agarwal, R. Chandra, A. Wolman, P. Bahl, K. Chin and R. Gupta, MobySys'07, Puerto Rico, June 2007
    • "CoolSpots: Reducing Power Consumption Of Wireless Mobile Devices Using Multiple Radio Interfaces", T. Pering, Y. Agarwal, R. Gupta and R. Want, Fourth International Conference on Mobile Systems, Application and Services (MobiSys), Uppsala, Sweden, June 18-22, 2006
  • 14. Credits: Projects and Teams
    • Completed Efforts
      • Power Aware Distributed Systems (PADS)
        • Mani Srivastava, UCLA
        • Cristiano Pereira
      • Formal Methods in Power Management
        • Sandy Irani, UC Irvine
        • Sandeep Shukla, Virginia Tech
        • Ravindra Jejurikar, Dinesh Ramanathan
    • Ongoing
      • System level Power Management
        • Zhen Ma, Yuvraj Agrawal, Zhong Yi Jin
      • Location awareness
        • SPATIAL PROGRAMMING
        • Ryo Sugihara, R. K. Shyamasundar, IRL & TIFR, India
        • DYNAMIC RESOURCE DISCOVERY
        • Jeffrey Namkung , Chalermek Intanagonwiwat, Amin Vahdat
    • Launching: GreenLight
      • Energy Efficient Parallelization for High Performance Computing
  • 15. Collaborating Radios Can
    • Improve Performance
      • Aggregate connectivity
    • Improve Reliability
      • Radios as backup interfaces
    • Improve Security
      • Multiple/Side-Channel Authentication
    • Improve Efficiency (Spectral, Energy)
      • Dynamically match radios to traffic, range
      • Use radios to page another, duty cycle other radios
    • Collaborating radios have a great potential for system-wide improvement
      • Energy, mobility management, capacity enhancement, channel failure recovery, networking, security, ….
      • We focus on energy.
  • 16. Typical power distribution Power breakdown for a fully connected mobile device in idle mode, with LCD screen and backlight turned off.
    • Cellular voice radio (GSM) highly optimized for low idle power
      • Cingular 2125: GSM radio consumes 38 times less power than Wi-Fi !
  • 17. Common Radio Standards Higher throughput radios have a lower energy/bit value … have a higher idle power consumption And they have different ranges.
  • 18. Consider: BT and WiFi
    • Objective: Always-on low-power operation with high peak bandwidth and overall energy efficiency
    • Two possibilities:
      • Use BT to page WiFi as needed
      • Build a switching hierarchy for energy efficient operation
        • Effectively expand the power states available at the system level
        • Switching policies are key to a good implementation.
    WiFi Active WiFi Active WiFi PSM WiFi Active BT Active WiFi Active BT Sniff Bluetooth Wi-Fi 264 mW 990 mW 81 mW 5.8 mW
  • 19. 1. BT as a paging radio
    • Scenario : An application on C1 wants to communicate with C3
    • C1 turns its 802.11 radio ON
    • C1 starts communication, sends data to AP through 802.11
    • AP matches C3’s destination IP with its BT address
    • AP sends WAKE-UP page to C3 via it’s BT interface, C3 turns on it’s 802.11 radio on receiving the WAKE-UP page
    • When C1 finishes sending data it switches OFF its 802.11 radio
    • If all connections to and from C3 are closed, AP sends SLEEP page
    • On receiving SLEEP page C3 turns OFF its 802.11 radio
  • 20. Simple paging (with range compensation)
    • Implemented iPAQs (3870), familiar linux and CISCO PCM-350, built-in BT
    • Measured power and latency on FTP and SSH sessions
    Power Savings for 802.11 card only vs PSP : 41% (SS1) to 95% (SS2) Throughput - Same as Awake Mode (CAM) , maximum throughput Latency - Setup latency, amortized across session
  • 21. 2. CoolSpots: Radio Hierarchy
  • 22. CoolSpots Network Architecture Infrastructure Computers CoolSpot Access Point BT WiFi BT WiFi Mobile Device Backbone Network IP address on Backbone Subnet Low-power Bluetooth link (always maintained, when possible) 1 Mobile device monitors channel and implements switching policy 2 WiFi link is dynamically activated based on switching determination 3 Access point changes routing table on “switch” message from mobile device 4 Switching is transparent: applications always use the IP address of the local subnet. 5
  • 23. Technical Challenge: Design of Switching Policies
    • Three main components contribute to the behavior of a multi-radio system
    • Position: Where you are
      • Need to address the difference in range between Bluetooth and WiFi
    • 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
  • 24. When: Policies bluetooth-fixed (using sniff mode) wifi CAM (normalization baseline) wifi-fixed (using PSM) bandwidth-X cap-static-X cap-dynamic kbps > X kbps < X kbps < X time > Y time > Y kbps < Z Z = kbps Use WiFi Channel Use Bluetooth Channel
  • 25. Experimental Setup
    • Characterize power for WiFi & BT
        • Multiple Policies
        • Different locations
        • Suite of benchmark applications
    • Stargate research platform
        • 400Mhz processor, 64MB RAM, Linux
        • Allows detailed power measurement
    • Tested using “today’s” wireless:
        • WiFi is NetGear MA701 CF card
        • Bluetooth is a CSR BlueCore3 module
      • Use the geometric mean to combine benchmarks into an aggregate result
      • Moved devices around on a cart to vary channel characteristics
    Test Machine (TM) Base Station (BS) RM Mobile Device (MD) SP Data Acquisition (DA) ETH BT WiFi mW Distance adjustment ETH = Wired Ethernet mW = Power Measurements BT = Bluetooth WiFi = WiFi Wireless RM = Route Management SP = Switching Policy Benchmark suite
  • 26. Switching Example: MPEG4 streaming
    • Simple bandwidth policy
    • Switch from WiFi to BT when application has buffered enough data
    Switching is transparent to unmodified applications! Switch : Wi-Fi -> BT Bluetooth Wi-Fi
  • 27. Results (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
  • 28. CoolSpots Results across various benchmarks w ifi-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
  • 29. 3. VoIP in Enterprise: Cell2Notify Internet IP Phone Soft Phone LAN Access Point SIP Proxy Smart Phone Wi - Fi interface GSM interface Base Station ATA GSM Network PSTN Enterprise Network Register GSM number Incoming VoIP call Disable Wi-Fi Match VoIP to GSM number Call GSM number Enable Wi-Fi Complete call setup over Wi-Fi ATA = Analog Telephony Adapter
    • Problem: Wi-Fi has to be ON to receive incoming calls.
      • Wi-Fi power consumption is high even when idle, reduces battery lifetime: Cingular 2125 : GSM (6.25days), Wi-Fi (9Hrs) !
  • 30. Power Consumption of a Smartphone Cingular 2125
    • Used to estimate energy savings for the Smartphone
      • Using real usage patterns from 3 different enterprise users
      • Lifetime based on the integrated 1150mAH @ 3.7V Li-ion battery
    1113.811 Wi-Fi (send/recv) 441.82 Wi-Fi (Connected) 1042.44 Wi-Fi (searching) 27.38 GSM Idle 15.688 All Radios Off (Flight mode) Power (mW) Scenario
  • 31. Battery Lifetime : Smartphone
    • Substantial increase in battery lifetime depending on usage!
      • John: 230% improvement, James : 540%
      • Beth improves lifetime by 70% despite very heavy usage
    70% 230% 540%
  • 32. Alternative: VoIP over Cellular Data Network?
    • VoIP over cellular data network (1xEvDO,GPRS/EDGE)
      • Expensive: requires subscription to data plan
      • Poor performance: Cellular data networks not optimized for VoIP
      • Greater power consumption than Wi-Fi for VoIP traffic !
  • 33. Summary
    • Multiple radios open up many possibilities for system-level performance and reliability increases
    • CoolSpots shows ~50% reduction in energy consumption over current power management in WiFi across applications, ranges
    • Cell2Notify: specific application (VoIP over Wi-Fi) shows battery life can be extended 1.7-6.4x, with maximum of 2 additional rings (7s)
    • Many improvements possible that take into account
      • Application behavior, Radio link quality, Network queues instead of ping latency, other scenarios (multi-user environments, p2p configurations)
      • Network infrastructure instead of standalone CoolSpots APs