Block-sw itched Netw orks: A New Paradigm for W ireless Transport - - PowerPoint PPT Presentation

▶
block sw itched netw orks a new paradigm for w ireless
SMART_READER_LITE
LIVE PREVIEW

Block-sw itched Netw orks: A New Paradigm for W ireless Transport - - PowerPoint PPT Presentation

Mar 26, 2023 473 likes •834 views

Block-sw itched Netw orks: A New Paradigm for W ireless Transport Ming Li, Devesh Agraw al, Deepak Ganesan, and Arun Venkataram ani University of Massachusetts Am herst U NIVERSITY OF U NIVERSITY OF M M ASSACHUSETTS ASSACHUSETTS , A , A

slide-1
SLIDE 1

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Block-sw itched Netw orks: A New Paradigm for W ireless Transport

Ming Li, Devesh Agraw al, Deepak Ganesan, and Arun Venkataram ani University of Massachusetts Am herst

slide-2
SLIDE 2

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

W hat You Buy vs. W hat You Get

TCP perform s poorly over w ireless links

Advertised capacity: 1 1 Mbps

4 0 x 2 x 1 .6 x

slide-3
SLIDE 3

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

E2E Feedback

1 . E2 E Transport

E2 E rate control is error-prone E2E retransmissions are wasteful E2E route disruptions cause unavailability

Rate Control Rate Control

Dest Source

Congestion? Link loss?

Loss Rate RTT

slide-4
SLIDE 4

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

1 . E2 E Transport

E2E rate control is error-prone E2 E retransm issions are w asteful E2E route disruptions cause unavailability

Source Dest

Redundant Transmissions

E2 E Retransm issions

P

slide-5
SLIDE 5

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

1 . E2 E Transport

E2E rate control is error-prone E2E retransmissions are wasteful E2 E route disruptions cause unavailability

Route disruptions due to mobility

slide-6
SLIDE 6

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

2 . Packet as Unit of Control

Channel access Link layer ARQ

Listen Backoff RTS/ CTS

slide-7
SLIDE 7

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

2 . Packet as Unit of Control

Channel access Link layer ARQ

Timeout Backoff Timeout Backoff

slide-8
SLIDE 8

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

3 . Com plex Cross-Layer I nteraction

Transport Link

Link-layer ARQs/ backoffs hurt TCP rate control

Highly Variable RTT

Link ARQ Link ARQ Link ARQ

Rate Control Rate Control

slide-9
SLIDE 9

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Hop: A Clean Slate Re-design End-To-End Packets Com plexity Hop-by-Hop Blocks Minim alism

slide-10
SLIDE 10

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Hop Design

Reliable Block Transfer ACK Withholding Micro-block Prioritization Virtual Retransmission Backpressure

Per-hop Multi-hop

slide-11
SLIDE 11

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Reliable Per-Hop Block Transfer

B-SYN Request for B-ACK B-ACK Packet bitmap Mechanism s

Burst mode (TXOP) Block ACK based ARQ

Benefits

Amortizes control overhead

CSMA TXOP

slide-12
SLIDE 12

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Hop Design

Reliable Block Transfer ACK Withholding Micro-block Prioritization Virtual Retransmission Backpressure

Per-hop Multi-hop

slide-13
SLIDE 13

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Mechanism

Leverages in-network caching Re-transmits blocks only when unavailable in cache

Benefits

Fewer transmissions Low overhead Simple

Virtual Retransm ission ( VTX)

B-SYN B-ACK

A B E C D

DATA

slide-14
SLIDE 14

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Mechanism

Leverages in-network caching Re-transmits blocks only when unavailable in cache

Benefits

Fewer transmissions Low overhead Simple

Virtual Retransm ission ( VTX)

A B E D

B-SYNVTX B-ACK

B-SYNVTX B-SYN with VTX flag set

VTX Tim er

E2E ACK

B-SYN

VTX Tim er

slide-15
SLIDE 15

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Hop Design

Reliable Block Transfer ACK Withholding Micro-block Prioritization Virtual Retransmission Backpressure

Per-hop Multi-hop

slide-16
SLIDE 16

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Mechanism

Limits # outstanding_blocks per-flow at forwarder

Backpressure

A B C D E Source Dest Slow Limit of Outstanding Blocks= 2

B-SYN Withhold B-ACK B-SYN B-SYN

slide-17
SLIDE 17

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Mechanism

Limits # outstanding blocks per-flow at forwarder

Benefits

Improves network utilization

Backpressure

A B C D E

Source Dest

F G

Dest

Aggregate goodput without backpressure: 6 Mbps

2 0 Mbps 1 0 Mbps 2 0 Mbps 2 0 Mbps 2 0 Mbps 1 Mbps

slide-18
SLIDE 18

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Mechanism

Limits # outstanding blocks per-flow at forwarder

Benefits

Improves network utilization

Backpressure

A B C D E

Source Dest

F G

Dest

Aggregate goodput with backpressure: 1 0 Mbps

2 0 Mbps 1 0 Mbps 2 0 Mbps 2 0 Mbps 2 0 Mbps 1 Mbps

Limit of Outstanding Blocks= 1

slide-19
SLIDE 19

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Hop Design

Reliable Block Transfer ACK Withholding Micro-block Prioritization Virtual Retransmission Backpressure

Per-hop Multi-hop

RTS/ CTS is overly conservative and incurs high overhead.

slide-20
SLIDE 20

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Mechanism :

Receiver withholds all but

  • ne B-ACK

Benefit:

Low overhead Less conservative Simple

Ack W ithholding

A C B

B-SYN B-ACK DATA B-SYN B-ACK DATA

W ithhold

B-ACK

slide-21
SLIDE 21

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Hop Design

Reliable Block Transfer ACK Withholding Micro-block Prioritization Virtual Retransmission Backpressure

Per-hop Multi-hop

slide-22
SLIDE 22

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Micro-block Prioritization

Mechanism s

Sender piggybacks small blocks to B-SYN Receiver prioritizes small block’s B-ACK

Benefits

Low delay for small blocks

SSH FTP

B-SYN B-ACK DATA B-SYN

6 4 B 1 MB Sender Sender Receiver

slide-23
SLIDE 23

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

2 0 nodes on the 2 nd floor of UMass CS building Apple Mac Mini Dual Core 1.8GHz, 2GB RAM, Atheros 802.11 a/ b/ g card

Testbed

slide-24
SLIDE 24

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Single-flow Single-hop Perform ance

Hop achieves significant gains over TCP

Hop Tcp

2 8 x 1 .6 x 1 .2 x

slide-25
SLIDE 25

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Single-flow Multi-hop Perform ance

Hop Tcp

2 .7 x 2 .3 x 1 .9 x

Hop achieves significant gains over TCP

slide-26
SLIDE 26

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Graceful Degradation w ith Loss

Em ulated link layer losses at the receiver

TCP drops to zero w ith m oderate losses

Hop Tcp

slide-27
SLIDE 27

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Scalability to High Load

3 0 concurrent flow s

Hop achieves m assive gains over TCP and is m uch fairer

Mean Goodput

Hop Tcp

Hop-by-hop TCP 1 5 0 x 2 0 x 2 x

slide-28
SLIDE 28

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Hop over W LAN

Hop im proves utilization over TCP+ RTS/ CTS

AP

Mean ( kbps) Median ( kbps) Hop 663 652 TCP 587 244 TCP+ RTS/ CTS 463 333

slide-29
SLIDE 29

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Low Delay for Sm all Transfers

AP

4 nodes perform large transfers, 1 node perform s sm all transfer

Transfer Size ( KB)

Hop low ers delay across all file sizes

slide-30
SLIDE 30

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Sum m ary of Other Results

Partitionable netw ork

TCP breaks down Hop significantly outperforms (TCP-based) DTN2.5

Netw ork and link layer dynam ics

Hop outperforms TCP under dynamic network conditions

Hop under 8 0 2 .1 1 g

Similar performance gains as in 802.11b

I m pact on VoI P

Hop impacts concurrent VoIP slightly more than TCP, but achieves significantly higher goodput.

slide-31
SLIDE 31

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Related W ork

Fixing E2 E rate-control

Separating loss/ congestion [ Snoop, WTCP, Westwood+ , ATCP, TCP

[ Snoop, WTCP, Westwood+ , ATCP, TCP-

  • ELFN]

ELFN]

Network-assisted rate control [ ATP, NRED, IFRC, WCP]

[ ATP, NRED, IFRC, WCP]

Hop circum vents rate control

Backpressure

ATM, theoretical work [

[ Tassiulas Tassiulas, ,… …] ]

Tree/ chain sensor data aggregation [ Fusion, Flush]

[ Fusion, Flush]

Reliable point-to-point transport [ RAIN, CXCC, Horizon]

[ RAIN, CXCC, Horizon]

Hop reduces backpressure overhead using blocks

Batching

Common optimization at link [ 802.11e/ 802.11n,

[ 802.11e/ 802.11n, WiLDNet WiLDNet , Kim08, , Kim08, CMAP] CMAP], transport [ Delayed [ Delayed-

  • ACK, DTN2.5]

ACK, DTN2.5], and network [ [ ExOR ExOR] ] layers

Hop leverages batching across layers

slide-32
SLIDE 32

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Sum m ary

Block sw itching > packet sw itching

Key abstraction: Reliable per-hop block transfer

Hop

Fast: Significant throughput, fairness, delay gains Robust: Degrades gracefully to challenged networks Sim ple: Minimizes complex cross-layer interaction Source code at http:/ / hop.cs.um ass.edu Can we have one transport protocol for diverse wireless networks? Yes, w e can!

slide-33
SLIDE 33

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

The End

Questions?

slide-34
SLIDE 34

U UNIVERSITY OF

NIVERSITY OF M

MASSACHUSETTS

ASSACHUSETTS, A

, AMHERST

MHERST •

  • Department of Computer Science

Department of Computer Science

Dynam ic Netw ork Conditions

3 0 concurrent flow s Auto Bit-Rate Control OLSR

Hop continues to significantly outperform TCP under dynam ic netw ork conditions