Interactions of Link-Layer and Transport Protocols
This page is not a complete list of references to the related work.
We have tried to be selective in what papers to cite,
but there are surely good papers that we are not aware of or have
forgot about. We would be happy to be notified about such papers, so
please feel free to drop us a mail.
Maintained by
Pasi Sarolahti
and
Sally Floyd.
Table of Contents:
Related Web Pages
IETF History in this Area
End-to-End Performance Enhancements
Link Layer based Performance Enhancements
Cross-layer Communication Mechanisms
Mid Boxes
Security
Router Performance & Hardware Protocol Processing
Radicalities
General Background:
End-to-End Principle & Congestion Control
Quality of Service
Other Pointers to Literature
Related Web Pages:
-
References on Layering
and the Internet Architecture
-
TCP and other Transport Protocols over X
-
Other Changes Proposed to TCP, including Reordering,
Corruption, Link Outages and Intermittently-Connected Paths,
-
Cross-Layer
Protocol Engineering for Wireless Mobile Networks: Part 1
IEEE Communications Magazine, Vol.43 No.12, December 2005.
" Of late, there has been an avalanche of cross-layer design proposals
for wireless networks."
-
Cross-Layer
Protocol Engineering for Wireless Mobile Networks: Part 2
IEEE Communications Magazine, Vol.44 No.1, January 2006.
" Applications using traditional protocol stacks (e.g., TCP/IP) from
wired networks do not function efficiently in mobile wireless
environments. This is primarily due to the layered architecture and
implementation of protocol stacks. One mechanism to improve the
efficiency of the stack is cross-layer feedback, that is, making
information from within one layer available to another layer of the
stack."
IETF History in this Area
-
TCP Over Satellite Working Group (tcpsat)
This working group produced
two RFCs.
-
Performance Implications of Link Characteristics Working Group (pilc)
This working group produced
seven RFCs.
- Triggers for Transport BOF (trigtran)
(November 2002,
March 2003)
Expired drafts:
draft-dawkins-trigtran-probstmt:
"When transport protocols are deployed over a path including a wireless
sub-network, a wireless access device now has significantly better
knowledge of the wireless portion of the connection path characteristics
than one or both endpoints... This document will set the framework for
investigation of whether this special knowledge from wireless access
devices can be useful to endpoint transports."
draft-dawkins-trigtran-linkup:
"This document codifies the "Link Up Notification" for TCP."
- Access Link Intermediaries Assisting Services BOF (alias)
(July 2003
)
"This BoF investigates further
work in support of transport intermediaries that provide assistance to
access links, including (but not exclusively) wireless links, primarily
in the areas of security protocol interaction with transport intermediaries
and response to changing link conditions."
-
RFC 3077: A Link-Layer Tunneling Mechanism for Unidirectional Links
-
Architectural Implications of Link Indications,
B. Aboba, internet-draft, work in progress, 2007.
"This document describes the role of link indications within the
Internet Architecture. While the judicious use of link indications
can provide performance benefits, inappropriate use can degrade
both robustness and performance."
-
Combined Fixes for TCP over Dynamic Paths,
presentation at the TCPM working group, July 2006.
"Basic Idea: Have TCP respond to generic
connectivity-change indications from lower-layers."
-
Cross-layer Indications for Transport Protocols,
P. Sarolahti, S. Floyd, and M. Kojo,
Internet-draft draft-sarolahti-tsvwg-crosslayer-01.txt, work in
progress, March 2007.
For citations, see
TCP and Other Transport Protocols over X and
Other Changes Proposed to TCP.
- A. Singh, A. Konrad, and A.D. Joseph.
Performance Evaluation of UDP Lite for Cellular Video.
In Proceedings of NOSSDAV'01, June 2001.
Discusses using RTP over UDP lite as transport for transmitting
MPEG-4 or H.263+ video stream in GSM network. UDP lite is compared to
regular UDP in reliable and non-reliable GSM mode. Non-reliable GSM
mode reduces the delay in delivering datagrams and UDP lite reduces
the loss rate of datagrams. As a end result, it is concluded that UDP
lite results in best perceived image quality when the codec is
resilient to some amount of errors.
- S. Cen, P.C. Cosman, and G.M Voelker.
End-to-End Differentiation of Congestion and Wireless Losses.
IEEE/ACM Transactions on Networking, 11(5), October 2003.
Compares different schemes of separating congestion losses from
wireless losses at the end host. In addition to existing solutions,
authors propose a new solution and present a hybrid scheme that
heuristically switches between different algorithms.
- M. Allman, W. Eddy, S. Ostermann.
Estimating Loss Rates With TCP.
ACM Performance Evaluation Review, 31(3), December 2003.
This paper introduces mechanisms for estimating the real loss rate of
packets at the TCP sender with TCP Reno and with SACK option.
Simply counting the number of
retransmissions often gives wrong results because TCP senders do spurious
retransmissions for various reasons. The paper identifies some of the
reasons for unnecessary retransmissions.
Another class of performance enhancements are those that operate
purely on link layer. As with the above section, there are lots of
work done in this category, and only a few items are selected.
- H. Balakrishnan, S. Seshan, and R.H. Katz.
Improving Reliable Transport and Handoff Performance in Cellular
Wireless Networks.
ACM Wireless Networks, 1(4), December 1995.
Describes Snoop agent that caches unacknowledged TCP data at the
wireless base station and makes local retransmission over wireless link
while masking the wireless losses from the real fixed sender. Additionally,
the paper presents a solution supporting handoffs by using IP multicast.
- R. Ludwig and B. Rathonyi.
Link-layer Enhancements for TCP/IP over GSM.
In Proceedings of IEEE Infocom '99, March 1999.
Proposes Quickstart-PPP that reduces the number of round-trips
required in the PPP handshake. Additionally, the paper discusses the
possibility of making the link flow-adaptive by using different
reliability mode for TCP and UDP packets.
- J. Scott, G. Mapp.
Link Layer Based TCP Optimisation
for Disconnecting Networks.
In ACM SIGCOMM Computer Communication Review, 33(5),
October 2003.
This paper presents a link-layer enhancement for
improving TCP performance on link disconnections.
When link comes back up after disconnection,
link layer protocol retransmits the last packet to
kick start the TCP transfer. This avoids the delay
related to TCP retransmission timeout. Different
variants of packet retransmission are compared.
This scheme could also be seen as one kind of
implicit link layer trigger.
-
T. Duke, T. Henderson, and J. Meegan,
Experience with "Link-Up Notification" over a Mobile Satellite Link,
CCR, V.34 N.3, pp. 93-104, July 2004.
Papers that discuss some form of link-layer/transport
interaction.
- B.S. Bakshi, P. Krishna, N.H. Vadiya, and D.K. Pradhan.
Improving Performance of TCP over Wireless Networks.
In Proceedings of the 1997 International Conference on Distributed Computer Systems,
Baltimore, May 1997.
The paper studies TCP Tahoe performance over simulated wireless
links. In addition to varying the packet MTU, the effect of explicit
feedback mechanisms are studied. The paper observes spurious
reransmission timeouts due to persistently reliable wireless link. In
addition to discussing use of ICMP Source Quench to hold the sender
when link goes to bad state, the paper presents Explicit Bad State
Notification that makes the sender cancel the present retransmission
timer and and restart a new one. Thus many retransmission timeouts are
avoided. EBSN is sent by the base station after every unsuccesful
attempt to transmit a packet over wireless link.
- H. Balakrishnan, V.N. Padmanabhan, S. Seshan, and R.H. Katz.
A Comparison of Mechanisms for Improving TCP
Performance over Wireless Links.
IEEE/ACM Transactions on Networking, 5(6):756-769, December 1997
This paper compares various end-to-end, split connection, and
link-layer based mechanisms for improving TCP performance over
wireless link. Additionally, an explicit loss notification mechanism
is presented. The receiver checks the correctness of a packet, and if
it is corrupted, an ELN is sent with a duplicate acknowledgement. Use
of ELN improves the performance over the baseline TCP Reno.
- K. Chandran, S. Raghunathan, S. Venkatesan, and R. Prakash.
A Feedback-Based Scheme for Improving TCP Performance in Ad-Hoc
Wireless Networks.
Proceedings of the 18th International Conference on Distributed
Computing Systems (ICDCS), Amsterdam, May 1998.
Presents TCP-F protocol that delivers explicit route failure
notification to TCP sender when a route failure in Ad-Hoc network
occurs. When the TCP sender gets a route failure notification, it
stops sending new packets, and freezes its timers and other TCP
state. The sender then waits for route re-establishment notification
or route failure timer to expire in order to continue transmission
from the frozen state sending activity and waits for a route failure
timer before continuing the connection.
- H. Balakrishnan and R.H. Katz.
Explicit Loss Notification and Wireless Web Performance.
Proceedings of IEEE Globecom Internet Mini-Conference,
Sydney, Australia, November 1998.
This paper extends the ELN mechanism presented above by
combining a Snoop agent at the wireless base station that
is also capable of making ELN markings. The paper also
discusses a bit different set of TCP transfer scenarios
than the earlier paper of the same authors.
- S. Goel and D. Sanghi.
Improving TCP Performance over Wireless Links.
In Proceedings of TENCON'98, pages 332-335. IEEE, December 1998.
Suggests using ICMP messages for informing the sender about failure
to transmit a packet on wireless link. ICMP-DEFER message is used to
indicate that link layer retransmissions are in progress and the
sender should defer making TCP retransmissions. When receiving the
ICMP-DEFER message, the sender restarts TCP RTO. Additionally,
ICMP-RETRANSMIT is sent from wireless basestation to indicate that TCP
sender should retransmit a packet, i.e. the wireless base station is
not doing any more local retransmissions.
- G. Holland and N.H. Vaidya.
Analysis of TCP Performance over Mobile Ad Hoc Networks.
In Proceedings of the Fifth International Conference on Mobile
Computing and Networking,
pages 219-230. ACM/IEEE, Seattle, August 1999.
The paper describes Explicit Link Failure Notification used in the
context of Ad-Hoc networks. ELFN is sent when link or route
failure in Ad-Hoc network is detected. Using ICMP host
unreachable and DSR routing protocols "route failure"
messages to deliver the ELFN notification were studied.
- R. Krishnan, M. Allman, C. Partridge, J.P.G. Sterbenz.
Explicit Transport Error Notification (ETEN) for Error-Prone Wireless
and Satellite Networks.
Technical Report No. 8333, BBN Technologies, March 2002.
Discusses different types of Explicit Transport Error
Notifications: Backward ETEN, Forward ETEN and Cumulative ETEN that
indicates the probability of corruption error on the network path to
the transport end hosts. The report evaluates ETENs with TCP Reno and
SACK, and compares TCP Westwood performance to ETEN schemes. It also
discusses security issues involved with ETENs.
- A. Gurtov and R. Ludwig.
Lifetime Packet Discard for Efficient Real-Time Transport over
Celullar Links.
In ACM Mobile Computing and Communications Review, 7(4):32--45,
October 2003.
Proposes use of IP options to indicate maximum lifetime for packets
in real-time protocols that do not require definite reliablity from
the link layer. Instead of persistently retransmitting a packet, a
link-layer can discard the packet if its lifetime in IP option is
exceeded. Additionally, the paper presents Headercasting,
i.e. transmitting the headers for packets after being discarded in the
network. The receiver can send feedback to sender based on the header,
and this helps to avoid unnecessary congestion control after packets have
been discarded due to lifetime expiration.
- Y. Swami and K. Le.
Lightweight Mobility Detection and Response (LMDR) Algorithm for TCP.
Internet Draft "draft-swami-tcp-lmdr-02.txt", March 2004. Work In
Progress.
A proposal of using a bit in TCP header to indicate that an end host
has moved to different subnet. The other end can then use this
information in adjusting its congestion control parameters. This
method does not require support from intermediate hosts, but uses
local information from lower protocol layers to adjust transport
protocol state.
- A.Gurtov and J. Korhonen.
Effect of Vertical Handovers on Performance of TCP-Friendly Rate
Control. ACM MCCR, 2004.
Discusses the TFRC performance on vertical handovers between GPRS,
WLAN and UMTS. TFRC is observed to adapt slowly on suddenly changing
network conditions. The paper proposes Explicit Handover Notification
to allow resetting the TFRC congestion control state for the new
conditions after the handover.
- M. Welzl
Passing Corrupt Data Across Network Layers: An Overview of Recent
Developments and Issues.
EURASIP Journal on Applied Signal Processing, 2005, Issue 2.
An overview of transport protocols such as UDP Lite and DCCP
that are designed to allow link layers to pass corrupt data to
network layers and above. The paper includes a discussion of
possible mechanisms for transport protocols to inform link layers
that corrupted packets should be passed up to higher layers.
The effect of intermediate proxies, NATs and Firewalls need to be
discussed when designing enhancements to end-to-end communication.
- N. Freed.
Behavior of and Requirements for Internet Firewalls
RFC 2979, October 2000.
Describes characteristics of firewalls and lists some common types
of firewalls. The RFC gives requirements for firewalls and their
transparency. Examples of how firewalls should handle some specific
types of network traffic are given. Additionally, some guidelines for
application designers about dealing with firewalls are given.
- J. Border, M. Kojo, J. Griner, and G. Montenegro.
Performance Enhancing Proxies Intended to Mitigate Link-Related
Degradations.
RFC 3135, June 2001.
This RFC discusses the different types of proxies that are known to be
used to improve performance especially on environments with wireless links.
The document describes mechanisms used in proxies and gives specific examples
of proxies. Implications and disadvantages of having proxies in the
communication path are also discussed.
- S. Floyd.
Inappropriate TCP Resets Considered Harmful.
RFC 3360, August 2002.
The document was motivated by some firewalls that send TCP Resets to the
TCP source when rejecting TCP segments. The document lists the events that
have been observed to trigger sending a TCP RST at firewall. It gives a
specific example regarding Explicit Congestion Notification (ECN), where
the presence of ECN bits in TCP header causes a firewall to send RST
back to the sender. Finally, the document more generally discusses the
problems that broken firewalls cause for TCP development an proposes
alternative solutions to be applied at midboxes instead of sending TCP RST
packets.
Discussing security issues related to transport protocols and in- and
outband signaling.
Understanding performance issues on packet processing is important
when designing triggers that are meant to be processed at busy
routers.
- V. Srinivasan, G. Varghese, S. Suri, Marcel Waldvogel.
Fast Scalable Algorithms for Level Four Switching.
In Proceedings of ACM SIGCOMM '98, Vancouver, Canada, August 1998.
Discusses efficiency of lookup algorithms in layer 4 switches, that
use additional information such as transport protocol and port numbers
in network switches or routers. This information can be used for
example in routing decisions or firewall filtering. The paper presents
two algorithms for efficient lookup of routing filters. The paper also
demonstrates experimental results when applying the lookup algorithms
on Pentium class machine.
- I. Hadzic and J. M. Smith.
Balancing Performance and Flexibility with Hardware Support for
Network Architectures.
ACM Transactions on Computer Systems, 21(4), November 2003.
Presents The Programmable Protocol Processing Pipeline (P4) for
accelerating protocol processing functions on hardware. P4 can be
reconfigured in real-time locally or remotely, and thus it is able to
flexibly adapt on changing conditions. The paper presents on example
application of Forward Error Correction where the P4 node monitors the
bit error rate and applies FEC only when the error rate exceeds a
certain level.
These papers discuss revolutionary ideas on the Internet architecture.
- D. L. Tennenhouse and D. J. Wetherall.
Towards an Active Network Architecture.
ACM SIGCOMM Computer Communication Review, 26(2), April 1996.
Rather than presenting any analytical or experimental results, this
paper discusses various topics regarding Active Networks and lays
foundation on future work on more intelligent network. Different types
of active networking are discussed and some threads for future work
are given.
- D. Katabi, M. Handley, and C. Rohrs.
Congestion Control for High Bandwidth-Delay Product Networks.
In Proceedings of ACM SIGCOMM 2002, Pittsburgh, PA, USA, August 2002.
Paper presents Explicit Control Protocol (XCP) that extends the idea of
ECN by allowing the routers to report the degree of congestion in the
network in feedback packets. With the new control protocol the
utilization control is separated from fairness control at the routers,
and most of the congestion-related packet losses are avoided.
- R. Braden, T. Faber, M. Handley.
From Protocol Stack to Protocol Heap - Role-Based Architecture.
In ACM SIGCOMM Computer Communication Review, 33(1), January 2003.
(Proceedings of HOTNETS-1 Workshop).
This paper presents a new architecture for network
protocols. Instead of layering the protocols, this paper
proposes a more flexible way of organizing different protocols
used in (Internet) communication.
Classical Papers. To remind about the basic rules that have to be
dealt with.
- V. Jacobson and M.J. Karels.
Congestion Avoidance and Control.
Slightly revised from ACM Computer Communication Review: Proceedings of the SIGCOMM '88,
Stanford, CA, August, 1988.
The Congestion Control paper. Introduces congestion window, slow
start, congestion avoidance and discusses the round-trip time
estimation.
- S. Floyd and K. Fall.
Promoting the Use of End-to-End Congestion Control in the Internet.
IEEE/ACM Transactions on Networking, 1(4), August 1999.
Discusses the negative effects of not being responsive to
congestion, identifies different forms of congestion collapse, and
suggests incentives employed at routers to encourage end hosts to
apply a proper congestion control. The paper proposes methods for
identifying flows that are not TCP-friendly, unresposive flows, and
flows that use disproportionate share of bandwidth, and how to deal
with such flows. Additionally per-flow scheduling algorithms at
routers are discussed.
- S. Floyd.
Congestion Control Principles.
RFC 2914, September 2000.
This RFC discusses various issues regarding end-to-end congestion control.
It describes the history and the problems that lead to present
congestion control requirements. The document gives guidelines for authoring
standards and implementations to follow the congestion control principles.
QoS mechanisms such as DiffServ are one form of signaling between the
network (including link layer) and the transport end points.
- T. Aimoto and S. Miyake.
Overview of DiffServ Technology: Its Mechanism and Implementation.
IEICE Transaction on Information and Systems, Vol. E83-D, No. 5,
pp. 957--964, 2000.
Gives on overview of DiffServ, including discussion of techniques
on gigabit routers to support DiffServ, and management of DiffServ
policies in the network.
- L. Breslau, E.W. Knightly, S. Shenker, I. Stoica, H. Zhang.
Endpoint Admission Control: Architectural Issues and Performance.
In Proceedings of ACM SIGCOMM 2000, August 2000, Stockholm, Sweden.
Endpoint admission control mechanisms try to ensure a requested
quality of service by probing the network from the end hosts. This
kind of mechanisms would avoid the scalability problems of Integrated
Services, because they do not require state at the routers. This paper
discusses architectural and performance issues of different approaches
for Endpoint admission control. The paper concludes that although
out-of-band probing often results in better loss rate, it is not
always superior to in-band signaling. The identified disadvantages of
endpoint admission control techniques are the substantial delay they
cause on connection setup, the problems related to misuse of endpoint
admission control mechanisms, and the open issues regarding deployment
of endpoint admission control techniques.
Web pages that have more pointers to interesting papers.
To be added:
Pasi Sarolahti
Last modified: November 2007.