Infostations A New Design for Wireless Computing Christopher

  • Slides: 32
Download presentation
Infostations: A New Design for Wireless Computing Christopher Crammond Charuhas Pandit Yaling Yang

Infostations: A New Design for Wireless Computing Christopher Crammond Charuhas Pandit Yaling Yang

Computing Trends 1980 s 1990 s 200? Desktop Laptop Network Wireless

Computing Trends 1980 s 1990 s 200? Desktop Laptop Network Wireless

The Downfall of 3 G • 3 G clearly addresses the issue of increased

The Downfall of 3 G • 3 G clearly addresses the issue of increased data rate. • What about. . . – System Asymmetry – Cost per Bit

3 G: System Asymmetry • Voice - equal amounts of traffic on both uplink

3 G: System Asymmetry • Voice - equal amounts of traffic on both uplink and downlink • Data - produces more traffic for downlink • 3 G wastes bandwidth

3 G: Cost Analysis • Voice requires 10 Kbit/s at v cents/min • Implying

3 G: Cost Analysis • Voice requires 10 Kbit/s at v cents/min • Implying a data rate of 1 MByte/s corresponds to 13 v cents/MByte • if v = 10, require $1. 30 per Mbyte • 3 G is expensive

Exit 3 G, Enter Infostations • Infostation: an isolated pocket of high bit -rate

Exit 3 G, Enter Infostations • Infostation: an isolated pocket of high bit -rate coverage • Smaller areas of coverage • Lower Transmit Power • Lower cost per bit

Infostation: The Premise • Designed to account for system asymmetries • Acknowledges data is

Infostation: The Premise • Designed to account for system asymmetries • Acknowledges data is bursty. • Tolerate longer delays. • Disjoint pockets of coverage allow for reducing the reuse factor N = 1. • Use unlicensed frequency band.

Achievable Data Rates • Optimistically: 500 Mbit/s in 100 MHz Band – Preliminary analysis

Achievable Data Rates • Optimistically: 500 Mbit/s in 100 MHz Band – Preliminary analysis under ideal conditions • Realistically: 4 Mbit/s in 1 MHz Band – Infostations at 100 meter intervals – Using current technology – Iacono August 1998

Mathematical Justification • Optimization Problem: – With finite energy, when to transmit? Water-filling Principle

Mathematical Justification • Optimization Problem: – With finite energy, when to transmit? Water-filling Principle • Solution: – For optimum throughput, transmit the most power to the user with the best channel – Implies use of smaller coverage i. e. , (r/R)<1

Calculations: Infostations • No need for ubiquitous coverage • Better SIR conditions • Smaller

Calculations: Infostations • No need for ubiquitous coverage • Better SIR conditions • Smaller cluster size • More bandwidth available gives higher data rate

SIR in 2 D Infostation system

SIR in 2 D Infostation system

Which system is better? 2 M bps 6 M bps

Which system is better? 2 M bps 6 M bps

Answer: Depends! • User Density • Importance of throughput vs. delay • Type of

Answer: Depends! • User Density • Importance of throughput vs. delay • Type of application offered

Queuing Analysis • Modeled as M/M/1 queue with reneging • If coverage (r/R) increases,

Queuing Analysis • Modeled as M/M/1 queue with reneging • If coverage (r/R) increases, – Worst case SIR decreases, so data rate available decreases, so m decreases. – Dwell time in area increases, so reneging rate (n) decreases – Arrival rate (l) increases

Relating l, n, m to r/R • Arrival rate l = pru. E[v] u

Relating l, n, m to r/R • Arrival rate l = pru. E[v] u = mobile user density v = mobile speed • Reneging rate n = 2 E[v]/pr • Service rate m related to data rate, c, related to modulation scheme • Performance of modulation scheme related to r/R.

Mailbox accumulation – performance parameter • Between infostations, mailbox accumulates messages • X =

Mailbox accumulation – performance parameter • Between infostations, mailbox accumulates messages • X = accumulated mailbox size between successive infostation visits (random variable) • E[X] = lu tc (m/c), where lu is the messages per user per second and tc is the average time between visits to Infostations. • tc= (p. R 2)/(2 r E[V])

2 D Numerical Example u=(0. . . 0. 0004), R=500, E[V]=2. 5, m=2 e

2 D Numerical Example u=(0. . . 0. 0004), R=500, E[V]=2. 5, m=2 e 6, lu=0. 01

What Does it all Mean? • Lowest E[X] for r/R =. 25 (reduced coverage)

What Does it all Mean? • Lowest E[X] for r/R =. 25 (reduced coverage) and 64 QAM (high modulation) • Thus justifying the use of infostations • For 1 D, delay is also better (WPMC ‘ 99)

Infostation Protocol Layering • Intelligent transmission protocol for Infostations. Application Transportation Wireless TCP for

Infostation Protocol Layering • Intelligent transmission protocol for Infostations. Application Transportation Wireless TCP for Infostations Network Mobile IP Data Link Control Media Access Control WINMAC Physical

WINMAC • Support both fairness and preemptive service • Multiple access & channel resource

WINMAC • Support both fairness and preemptive service • Multiple access & channel resource allocation • Transmission rate switching • Retransmission scheme

Multiple Access Protocol • Slot Allocation Algorithm: When a mobile with higher priority arrive,

Multiple Access Protocol • Slot Allocation Algorithm: When a mobile with higher priority arrive, assign it most of the channel resources. Otherwise provide fair service. • TDMA/TDD configuration: A frame consist of a number of time slot and each slot can carry one packet • Reservation and Transmission: • Look for beacons • Choose a rate based on the highest rate and the largest # of available slots • Send reservation and registration request • Transmit data

Transmission Rate • Rate Switching algorithm (suppose 3 rates): – Listen to beacons of

Transmission Rate • Rate Switching algorithm (suppose 3 rates): – Listen to beacons of 3 data rates – Monitor Packet Error Rate of current channel condition – Suppose transmission in rate 2 and find out that 2 out of 3 recent beacons of rate 3 is heard and PER<0. 1 then switch to rate 3. Otherwise, if less or equal to one beacon out of 3 is heard of rate 2 and PER>0. 5, then switch to rate 1.

Retransmission Scheme • Combine SR (selective repeat) & GBN (Go back N) ARQ. •

Retransmission Scheme • Combine SR (selective repeat) & GBN (Go back N) ARQ. • SR: Only require sender to retransmit the lost packet. • GBN: Require the sender to retransmit all the packet after the error or lost packet. • Using SR within the processing ability of hardware or software and use GBN beyond the ability.

Architecture UMTS SS 7 signaling Network Infostation Controller Radio Port Proxy Cache Radio Port

Architecture UMTS SS 7 signaling Network Infostation Controller Radio Port Proxy Cache Radio Port Internet ATM or Frame Relay Radio Port

Infostation and Mobile IP • Current Proposals for Mobile IP fit well into the

Infostation and Mobile IP • Current Proposals for Mobile IP fit well into the Infostation concept. Correspondent Host Home Agent Mobile Host Foreign Agent

Infostations and Mobile IP I • When the mobile recognizes that it is no

Infostations and Mobile IP I • When the mobile recognizes that it is no longer connected to its home network, it is assigned a care-of-address from the network it is visiting. In an infostation system, the IC will assign the care-ofaddress to the mobile. • The mobile then informs the home agent about its current location to allow packets to be forwarded to it.

Infostations and Mobile IP II • The mobile can also inform the other party

Infostations and Mobile IP II • The mobile can also inform the other party of its current care-of address, which the other party can use to directly communicate with the mobile. • When a mobile enters the RP of another Infostation it would register there, obtain a new care-of-address, and continue the session using the new Infostation.

TCP Solution • Split Connection: A TCP connection between a fixed host and mobile

TCP Solution • Split Connection: A TCP connection between a fixed host and mobile host is split into two connections at the BS. • End-to-End approach-TCP smart: The snoop agent in the Infostation Controller (IC) monitors all of its TCP connections. The IC will cache all segments received from the fixed Host and not yet acknowledged by the MH.

Application Design for a Network of Infostations • Example: A map service for vehicles

Application Design for a Network of Infostations • Example: A map service for vehicles on a highway. High bandwidth info-station coverage

Open Infostation Research • Transmitting in unlicensed bands • Randomized movement of mobile. •

Open Infostation Research • Transmitting in unlicensed bands • Randomized movement of mobile. • Data transfers over multiple sessions. • All the hassles of increased bit-rate.

Infostation Downside • Difficult to track down explicit numbers. • Exclusively at WINLAB. •

Infostation Downside • Difficult to track down explicit numbers. • Exclusively at WINLAB. • Lack of commercial interest. • Dependent of failure of 3 G.

Bibliography • Borras, Joan and Roy D. Yates. "Infostation Overlays in Cellular Systems. "

Bibliography • Borras, Joan and Roy D. Yates. "Infostation Overlays in Cellular Systems. " IEEE WCNC (1999): 495 -9 • Elaoud, Moncef and Parameswaran Ramanathan. "TCP-SMART: A Technique for Improving TCP Performance in a Spotty Wide Band Environment. " IEEE (2000): 17 -83 -7 • Goodman, David. "The Wireless Internet: Promises and Challenges. " computer. org 2/01 • Frenkiel, Richard, B. R. Badrinath, Joan Borras, and Roy D. Yates. "The Infostations Challenge: Balancing Cost and Ubiquity in Delivering Wireless Data. " IEEE Personal Communications (2000): 6671. • Iacono, Ana Lucia and Christopher Rose. "Minimizing File Delivery Delay in an Infostation System. " WINLAB, 1998. • Irvine, J. , D. Pesch, D. Robertson, and D. Girma. "Efficient UMTS Data Service Provision using Infostations. " IEEE (1998): 2119 -23. • Yates, Roy D. , Narayan B. Mandayam. "Challenges in Low-Cost Wireless Data Transmission. " IEEE Signal Processing (2000): 93 -102. • Wu, Gang, Churng-Wen Chu, Kevin Wine, James Evans, and Richard Frenkiel. "WINMAC: A Novel Transmission Protocol for Infostations. " IEEE VTC'99. 1999. • Ye, Tao, H. Arno Jacobsen, and Randy Katz. "Mobile Awareness in a Wide Area Wireless Network of Info-stations. " Berkeley.