TLS in manet

1
Transport Layer for MobileTransport Layer for Mobile
Ad Hoc NetworksAd Hoc Networks
Prepared By :Prepared By :
Patel Jay CPatel Jay C
ME(EC)-140870705004ME(EC)-140870705004

2
The Quest for the Transport
Layer
 Assess the state of the art of transport protocols
 Target environment: mobile ad hoc nets (MANET)
 Which is the best TCP variant?
 Do we need a new transport protocol?
 The question is very timely…
Introduction

3
Network Architecture at a
Crossroads
 The community recognizes the need for change
 Wireline-centric network design is “obsolete”
 New network environments have emerged
 Ad hoc, sensors, consumer-owned, delay-tolerant
 New networking technologies have emerged
 UWB, cooperative approaches, MIMO, directed
antennas
Introduction

4
New Class of Networks
 Thousands of nodes, highly resource
constrained, highly unreliable wireless links,
low duty cycle (smartdust)
 Tens - thousands of nodes, Nano-sensors
 Hundreds of nodes, resource constrained,
unreliable wireless links (Sensors)
 Tens of nodes, resource constrained, wireless
links, charged every day (PDAs)
links, line powered (embedded devices)
links, line powered (computers)
Introduction

5
A New Era Has Begun
New
Machines
New Environments
Applications
New
Networks
Introduction

6
The Role of Networking is Central
Wireless
Networking
Embedded
Systems
Sensors
Embedded
Sensor
Applications
Introduction

7
Revisiting the Architecture
 The vision:
Wireless as an integral part of the network
Multiple wireless hops: not just the last mile
Pockets of wireless ad hoc connectivity
 A new protocol stack is required
Is TCP/IP capable of delivering?
Introduction

8
Revisiting The Hourglass
Introduction
Email WWW Voice...
SMTP HTTP RTP...
TCP UDP
IP
Ethernet PPP…
MultiAccess async sync...
copper fiber radio...
User Application
Application Protocol
Transport Protocol
Media Access Protocol
Media Sharing Principles
Physical System
Internet Protocol

9
Problem: Evaluate TCP
 Why does TCP perform poorly in MANETs?
Developed for wire-line networks.
Assume all losses are due to congestion.
 Many TCP variants have been proposed.
How good are they? Are they sufficient?
 Are there any other alternatives?
Are non-TCP protocols the solution?

10
Our goal
 Identify the problems of TCP in MANETs.
 Evaluate various major TCP variants.
12 TCP variants, 7 improvement techniques
 Observations:
Most TCP variants are NOT sufficient.
A new transport layer protocol is needed.

11
Overview of Results
 The best TCP variants:
 TCP-Westwood and TCP-Jersey seem the best.
 Both protocols estimate bandwidth more accurately.
 TCP mechanisms:
 Feedback from intermediate nodes leads to big gains.
 The best non-TCP approaches:
 Ad-hoc Transport Protocol (ATP) seems to address most issues
 Non-window based: estimates achievable rate periodically
 Split-TCP: promising new way of looking at transport layer
 Dynamically buffer packets mid-path
 Key: Separation of congestion control from reliability.

12
Roadmap
 Overview of TCP
 The problems of TCP over MANETs
 Overview of best transport protocols
 In depth
 Specific problems of TCP over MANETs
 Details of major TCP variants
 Discussion - other efforts
 Conclusion

13
Overview of TCP concepts
 Conventional TCP: Tahoe, Reno, New-Reno
 Sending rate is controlled by
 Congestion window (cwnd): limits the # of
packets in flight
 Slow-start threshold (ssthresh): when CA
start
 Loss detection
 3 duplicate ACKs (faster, more efficient)
 Retransmission timer expires (slower, less
efficient)
 Overview of congestion control mechanisms
 Slow-start phase: cwnd start from 1 and
increase exponentially
 Congestion avoidance (CA): increase linearly
 Fast retransmit and fast recovery: Trigger by
3 duplicate ACKs
Overview
Slow-start Congestion
avoidance

14
What is different in MANETs?
1. Mobility
 Route stability and availability
1. High bit error rate
 Packets can be lost due to “noise”
1. Unpredictability/Variability
 Difficult to estimate time-out, RTT, bandwidth
1. Contention: packets compete for airtime
 Intra-flow and inter-flow contentions
1. Long connections have poor performance
 More than 4 hops thruput drops dramatically
Overview

15
Overview of the Best Protocols
 TCP-Westwood
 Estimate bandwidth to alleviate the effect of wireless errors.
 TCP-Jersey
 Estimate bandwidth to alleviate the effect of wireless errors.
 Congestion warning assists the determination of packet loss
due to wireless error from congestion.
 ATP
 Rate based transmission, periodic rate feedback, no timeout
concept, reliability provided by SACK.
 Split-TCP
 Separating congestion control from reliability.
 Dropped packets are recovered from the most recent proxy
instead of the source.
Overview

16
Why does TCP fail in MANETs?
Specific problems are identified:
1. TCP misinterprets route failures as congestion
2. TCP misinterprets wireless errors as congestion
3. Intra-flow and inter-flow contention reduce
throughput and fairness
4. Delay spike causes TCP to invoke unnecessary
retransmissions
 RTO too small  unnecessary retransmissions.
1. Inefficiency due to the loss of retransmitted packet
 When retransmitted packet is lost timer expires  performance drops
Overview

17
Roadmap
 Overview of TCP
 The problems of TCP over MANETs
 Overview of best transport protocols
 In depth
 Specific problems of TCP over MANETs
 Details of major TCP variants
 Discussion - other efforts
 Conclusion

18
Specific problems of TCP over MANETs
 TCP misinterprets route failures as
congestion
Effects: Reduce sending rate
Buffered packets (Data and ACKs) at
intermediate nodes are dropped.
Sender encounters timeout.
 Under prolonged disconnection, a series of
timeouts may be encountered.

19
 TCP misinterprets wireless errors as
congestion
 Effects: Incorrect execution of congestion control
 Performance drops.
 Wireless channel is error-prone compared to wireline
 Fading, interference, noise

20
 Intra-flow and inter-flow contention
 Effects: Increased delay, unpredictability, and
unfairness.
 Inter-flow contention: contention of nearby flows.
 Intra-flow contention: between packets of the same
flow (e.g. forward data and reverse ACKs).
 Wireline: only packet on same link “compete”
Data stream
ACKs stream
Two nearby flows

21
 Delay spike causes TCP to invoke
unnecessary retransmissions
Effects: Performance drops and many
unnecessary retransmissions. [Ludwig & Katz]
Variability: Spikes are not uncommon here
Spikes throw off parameter estimation and
tuning
 RTO, window size, slow-start threshold

22
 Inefficiency due to the loss of
retransmitted packet
Effects: Performance drops significantly
under high loss environment (e.g. MANETs).
Losing a retransmitted packet hurts
 TCP can recover from one loss (fast retransmission)
Wired networks: packet loss rate is low.
Here, high packet loss makes the problem
significant

23
Classification of Transport
protocols
 TCP variants try to improve the performance
by the following ways:
 Estimating the available bandwidth
 Determining route failure and wireless error
 Reducing contention
 Detecting spurious retransmission
 Exploiting buffering capability
 New approaches: Non TCP variants
 Use rate based instead of window based approach
 Enable dynamic buffering (split TCP)

25
A Rate-based Transport Layer Protocol
 Ad-hoc Transport Protocol (ATP) [Sundaresan et. al.]
 Feedback from intermediate nodes
 path failure, queueing delay, periodic feedback on rate
 Rate based transmission
 Entirely rate-controlled. (no window concept)
 Evenly distribute transmissions over time. (reduce burstniess)
 Decoupling of congestion control and reliability
 Does not require the arrival of ACKs to clock out segment.
 Does not employ cumulative ACKs but solely relies on periodic SACK (with
20 SACK blocks) to identify losses.
 Pro: 1) Estimate rate accurately. 2) Reduce traffic on the reverse path. 3)
Recover more than one lost segment at a time.
 Con: 1) Incompatibility problem. 2) Require the assistance from the
intermediate nodes. 3) Fastest possible time to detect and recover packet lost
is 1 second.

26
Split-TCP: a new approach
 Split-TCP: work done at UCR [Kopparty et. al.]
 Setup proxies along the connection  many short TCP connections.
 Congestion control and reliability are separated.
 Proxies buffer packets from the previous proxy or the source.
 Any dropped packets are recovered from the most recent proxy but
not from the source.
 Pro: Enhance parallelism. Reduce bandwidth consumption on
retransmission.
 Con: Optimal frequency of proxy placement is not clear.
R SP P P

27
Other Transport Layer Efforts
Can FAST TCP and XCP work well over MANETs?
 Do not seem suitable for MANETs.
 Basic idea: React faster to change.
 Fast TCP [Jin et. al.]
 Determine equilibrium by queuing delay and loss information.
 cwnd far away from equilibrium?  Rapid (Large) change.
 cwnd approach equilibrium?  Small change.
 XCP [Katabi et. al.]
 Explicit congestion signaling.
 Intermediate nodes estimate spare bandwidth and generate
feedback to the sender.
 Neither protocol can deal with mobility.
 Mobility and route changes will throw off calculations.

TLS in manet

More Related Content

What's hot (20)

Viewers also liked (7)

Similar to TLS in manet (20)

More from Jay Patel (17)

Recently uploaded (20)

TLS in manet

Editor's Notes