PAWP A Power-Aware Web Proxy for WLAN Clients Marcel Ro u, Michael - - PowerPoint PPT Presentation

December 3, 2004 Marcel C. Rosu, IBM T.J. Watson

PAWP – A Power-Aware Web Proxy for WLAN Clients Marcel Roşu, Michael Olsen, and Chandra Narayanaswami

IBM T.J. Watson Research Center

Lu Luo

Carnegie Mellon University

December 3, 2004 WMCSA 2004

Power-Aware Web Proxy Usage

Access Point

Wireless Client

Internet Internet

Firewall

Wireless Client Wireless Client

Access Point

Wireless Client

WLANs Origin Server Origin Server Origin Server Caching Proxy (optional)

PAWP Proxy

December 3, 2004 WMCSA 2004

Power Optimization for Wireless NIC

• Active power consumption in WLAN interface:

5-10% in notebooks, 50-90% in PDAs

• Existing power-reduction approaches for WLAN clients:

802.11 Power Saving Mode – limited power saving during active

transmissions

MAC level – extending sleep time Transport level – energy efficient protocols The unpredictability of incoming traffic causes waste of power

• Our approach – Power Aware Web Proxy (PAWP), using:

A web proxy to shape HTTP traffic going into client’s WNIC

Based on:

Application domain knowledge MAC level configuration Network conditions

December 3, 2004 WMCSA 2004

Outline

802.11 Power Management Interactions with Incoming WLAN Traffic PAWP Architecture Traffic Shaping Rules Experiments Testbed Methodology Results Conclusions

December 3, 2004 WMCSA 2004

802.11 Power Management

Power Modes Active Power Save Power States Awake (both Modes, always when listening to beacon from station)

• PRISM3 PCMCIA card: 848mW

Doze (Power Save Mode)

• PRISM3 PCMCIA card: 25mW

Transition between modes always initiated by station Frame exchange with access point Active -> Power Save after idle configurable period Power Save -> Active after sending/receiving frame

December 3, 2004 WMCSA 2004

Analysis of Incoming WLAN Traffic

Tbeacon Tlisten

Power

Time [beacons] Pdoze Pawake Pdoze Pawake

Power N DATA

Tbeacon 1 2 3 4 5 6 7 8 9 Tlisten Time [beacons]

N DATA

1 2 3 4 5 6 Packet arrival at AP Packet arrival at Proxy

b) Proxy (Ttimeout=25ms)

Packet arrival at AP

a) Direct (no proxy) (Ttimeout=100ms)

Data transferTimeout

Sources of wasted energy in WLAN interfaces Proposing: Power Aware W/S Proxy that modulates WLAN data transmission into intervals of high and no traffic

December 3, 2004 WMCSA 2004

Comparison on Power Consumption

eBay.com

0.5 s/div 20 mV/div 20 mV/div 0.5 s/div 20 mV/div 0.05 s/div

DIRECT PROXY

www.eBay.com DIRECT PROXY ACPIspec.pdf

December 3, 2004 WMCSA 2004

Outline

802.11 Power Management Interactions with Incoming WLAN Traffic PAWP Architecture Traffic Shaping Rules Experiments Testbed Methodology Results Conclusions

December 3, 2004 WMCSA 2004

Power-Aware Web Proxy Architecture

Client Side Threads Fetching Threads

WLAN Clients Web Servers

Blackboard

Oracle

Rules

Information per Client Information per Request

December 3, 2004 WMCSA 2004

PAWP: Compensating Content Delays

Delaying Content Release Compensating for Delays

Prefetch Embedded

Objects

Pipeline Requests Pipeline Responses Prioritize Tasks Major Challenge

Handling HTTP Cookies

Delay Data Release Release Data

N N N N Y Y Y Y

CurrentTime – TimeOfLastRequest < Ttimeout TotalBytesReceived >= MinBytes ObjectsReceived >= MinObjects CurrentTime – TimeOfLastSend < MaxDelay

December 3, 2004 WMCSA 2004

Outline

802.11 Power Management Interactions with Incoming WLAN Traffic PAWP Architecture Traffic Shaping Rules Experiments Testbed Methodology Results Conclusions

December 3, 2004 WMCSA 2004

Experimental Testbed

Oscilloscope (VellemanPC S64i) Data collection PC Digital Multimeter (HP3458A)

R=0.53Ω Vdd=3.3V +

• VR

Wireless Client (IBM ThinkPad) Intersil PRISM3 PC Card with Extender

Power measurement environment for wireless client network interface card HTTP protocol trace collection using IBM PageDetailer

• Downloading time distribution
• Information on web objects
• HTTP headers

December 3, 2004 WMCSA 2004

Complete, Across-The-Board Experiments

(sec) 30 60 90 120 150 180 210 Direct Proxy1 Proxy2 Proxy3 Proxy4 Direct Proxy1 … …

Based on the new experimental testbed

Experiments on each proxy configuration can be done in < 30 sec Quick, automatic switching between configurations Measurements in each set are close in time – avoided deviation

1st set 2nd set

December 3, 2004 WMCSA 2004

Experimental Results (1)

Website Size [kB] / Num of Objects Connection Type Download Energy [J] Download Time [s] Throughp ut [kB/s] cnn 281kB/84 Direct Proxy 2.47 2.25 (-9%) 8.13 7.33 (-10%) 34.6 nytimes 253kB/76 Direct Proxy 2.36 1.89 (-22%) 8.17 5.78 (-29%) 30.1 washingtonpost 535kB/73 Direct Proxy 6.14 2.83 (-54%) 9.08 8.58 (-6%) 56.0 Internet Explorer bbc 61kB/31 Direct Proxy 2.10 1.05 (-50%) 3.56 3.37 (-5%) 17.1 cnn 252kB/84 Direct Proxy 3.30 1.37 (-59%) 4.63 3.88 (-16%) 54.3 nytimes 190kB/45 Direct Proxy 3.29 1.11 (-66%) 6.85 3.20 (-53%) 23.3 washingtonpost 504kB/67 Direct Proxy 4.99 2.20 (-56%) 7.34 7.01 (-5%) 44.4 Mozilla

December 3, 2004 WMCSA 2004

Experimental Results (2)

10 20 30 40 50 60 Throughput [kBps] 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0% 100.0% Eproxy/Edirect [%]

Relative energy consumption with Proxy vs. Direct case throughput Cost and Benefits of Proxy Features

NY Times (www.nytimes.com) 240kB/77

Download Energy [s] Download Time [s]

Direct (no proxy) 2.70 8.75 Proxy: all features disabled 2.46 8.95 Proxy: scheduling, prefetching 2.38 8.05 Proxy: scheduling, prefetching, request & response pipelining 2.15 7.54 Proxy: all features on 1.94 6.99

December 3, 2004 WMCSA 2004

Conclusions

PAWP challenges No client modifications Visible to clients Invisible to servers Don’t over-shape traffic

• Avoid increasing download times

Lessons learned Page design matters (cookies) HTTP usage is increasing PAxP extends savings beyond Web browsing

December 3, 2004 WMCSA 2004

Questions?