FINAL PROJECTS (alphabetical order):
Evaluation of different TCP congestion control algorithms by ns-2:
Presentation slides and final report (PDF files).
TCP congestion control was introduced into the Internet in the late 1980's by Van Jacobson. Immediately preceding this time, the Internet was suffering from congestion collapse. The algorithm for TCP congestion control is the main reason we can use the Internet successfully today despite resource bottlenecks and largely unpredictable user access patterns.
TCP congestion control lies in Slow-Start, Additive Increase Multiplicative Decrease (AIMD), Fast Retransmit and Fast Recovery. There are different implementations among which are Tahoe TCP, Reno TCP, New Reno TCP, SACK TCP and and TCP Vegus. In the project, we will evaluate the performance regarding throughput, delay, and loss. Various parameters will be used.
Simulation will be done using Network Simulator (NS-2)
References:
Forward Error Control in Wireless LANs:
Presentation slides and final report (PDF files).
Project Description:
Forward error control (FEC) is not specified in the current 802.11b Wireless Local Area Network (LAN) standard.
This project is to investigate the application of FEC to archive better TCP performance over wireless LAN.
Simulation will be carried out in OPNET to model the FEC performance without implementing the actual coding
algorithms.
Reference:
Generating Network Topologies with Highly Optimized Tolerance (HOT):
Presentation slides and final report (PDF files).
Due to the complexity of today's network, simulation tools are often used to evaluate new systems and protocols. In order to better evaluate the performance, the topology models used in these tools are required to be as realistic as possible. Recent empirical studies have shown that Internet topologies exhibit power law statistics[1]. One theory which tries to explain this phenomenon is "self organized criticality" (SOC)[2]. Jean Carlson and John Doyle proposed another theory[4], called highly optimized tolerance (HOT), which is a new mechanism used to generate power law distributions. Compared with SOC, HOT focuses on systems which are optimized, either through natural selection or engineering design, to provide robust performance despite uncertain environments. In this project, we will implement a topology generator based on HOT theory and if time permits, we will compare our computational results with some other topology generators.
The project will be done using Java programming language, ns-2 and some other topology generators.
References:
Route optimization of Mobile IP over IPv4:
Presentation slides and final report (PDF files).
1. Abstract
The support of mobility in the modern communications networks is becoming
essential and important with the booming development of mobile devices.
Mobile IP, built on IPv4, was designed by IETF to serve the needs of
supporting portable IP addresses on Internet.
In the basic mobile IP protocol, datagrams going to the mobile node have to travel through the home agent when the mobile node is away from home. On the other hand, the datagrams sent from the mobile node to other wired nodes can be routed directly. This asymmetric routing, called "Triangle routing", is generally far from optimal, especially when the destination node is close to the mobile node.
Eliminating the "Triangle routing" problem, in order to improve network efficiency, is one appealing topic in mobile IP. IETF proposed extension part of the basic mobile IP, called "Route Optimization". Actually, in the next generation of the Internet Protocol - IPv6, "Route Optimization" is integrated as a fundamental part of the mobility support.
However, IPv4 has already been widely deployed and will continuously dominate the Internet for a long time. Therefore, the study of Route Optimization is still of interest to us, although it is no longer a problem in IPv6.
2. Project Plan
Tool used in the project: ns-2
In the project, we plan to implement the Route Optimization extension of mobile IP over IPv4 in ns-2, since the basic mobile IP has already been implemented by the Monarch project of CMU in ns-2. Also, the mobile IP over IPv6 in ns-2 has been almost implemented by MobiWan project of MOTOROLA Labs and INRIA Rhtne-Alpes PLANETE.
We need to extend the current modules of mobile IP in ns-2. After extension, we will compare the usual routing scheme in mobile IP without optimization to the routing with optimization in terms of the network utilization and end-to-end delay etc. If time permits, we would like to compare other route optimization approaches with the IETF proposal, to get hands-on experience and better idea about them.
3. Reference
An Analysis of Constraint-based Routing in MPLS:
Presentation slides and final report (PDF files).
Project Description:
1. Introduction
Multiprotocol Label Switching-MPLS has now become a fundamentally important technology in the internet. Several of the largest Internet service providers have deployed MPLS in their production networks to solve problems such as traffic engineering and to offer IP services efficiently over ATM backbone networks. [5]
As currently specified in RFC 3036 [3], the LDP protocol for MPLS does not support signalling of the MTU for LSPs to ingress LSRs. This functionality is essential to the proper functioning of RFC 1191 [2]. Without knowledge of the MTU for an LSP, edge LSRs may transmit packets along that LSP which are, according to [4], too big. Such packets may be silently discarded by LSRs along the LSP, effectively preventing communication between certain end hosts. [1]
The solution proposed in [1] enables automatic determination of the MTU for an LSP with the addition of a TLV to carry MTU information for a FEC between adjacent LSRs in LDP Label Mapping messages.
2. Project plan
In this project, I will first try to implement MTU Signalling extension for LDP [1] in OPNET. Then I will compare the model of extension with the one from the OPNET MPLS model suite .
Reference:
Presentation slides, final report and addendum (PDF files).
Overview:
IEEE 1394b is a proposed serial-bus interface and protocol that performs much like a packet switched network. The existing IEEE 1394a-2000 is a 400 Mbps technology, with IEEE 1394b extending that all the way up to 3.2 Gbps. More interestingly for us, IEEE 1394b also extends the link level protocol to support packet transmission or bus-arbitration immediately after an ACK, instead of waiting for the DELAY_TIMEOUT.
Presently this technology shows up in computers, video cameras, stereo systems, high-speed peripherals, and more. It supports peer-to-peer communications (unlike USB which requires a single controlling host), and auto-detects the spanning tree for its network layout.
Both IEEE 1394 technologies support two delivery modes:
1) guaranteed delivery (asynchronous).
2) guaranteed delay (isochronous).
Isochronous transfer provides a fixed bandwidth channel over fixed time intervals. This mode is ideally suited to time-based multi-media applications such as video and audio.
The goal for my project has two parts:
1) Understand IEEE 1394b sufficiently to implement its PHY (equivalent to Ethernet's MAC) and Link layer protocols in NS-2. 2) Explore the bandwidth utilization of several IEEE 1394b networks of various sizes and loads, comparing the asynchronous and isochronous transfer modes.
While such a comparison is interesting, it is not as applicable as I previously hoped. Since the isochronous transfer will not re-transmit lost packets, it is not suitable for all tasks (such as file transfers); further, IEEE 1394 uses bus arbitration (command packets) to determine who can transmit data next (unlike Ethernet, where simultaneous attempts at transmission will consume bandwidth). Since application needs are the primary constraint in choosing the transfer mode, it doesn't seem that isochronous transfer is a universal solution to improving efficient allocation of bandwidth. However, determining the overhead imposed by both transfer modes under a variety of network loads, is still interesting. It may even be possible to get a reliable isochronous transfer with a higher level protocol (e.g., using two isochronous channels -- one very low bandwidth one used for the ACKs).
I already have NS2 installed, compiling and working in preparation for adding the IEEE 1394 related classes.
References:
On the high variability of AS sizes in Internet Topology:
Presentation slides and final report (PDF files).
There has been a significant increase in research activities related to modeling and analyzing the Internet topology in the past three years. The results of such studies may facilitate many tasks such as designing more efficient protocols, creating realistic models for simulation, and speculation on the growth process of the Internet. In particular, we focus on the toplogical features at the domain, or Autonomous System (AS) level, where an AS is a connected subnetwork administered by a specific authority. We plan to investigate the high variability in AS sizes (the number of routers in an AS) and how it is related to AS degrees.
A recent study by Faloutsos et al. [1] used topology information collected using Border Gateway Protocol (BGP) routing tables obtained from the Oregon route server [2], and showed that graph distributions followed power-laws. An explanation of that phenomenon, the BA model, was presented by Barabasi and Albert [3]. However, it has been shown that the BGP data may provide only a very sketchy picture of the complete inter-AS connection in the actual Internet [4]. Using more complete data sets, Chang et al. showed that power laws do not hold, however the distributions are highly variable [5]. A recent paper by Tangmunarunkit et al. argued that the BA model can not be a valid explanation for the connectivity evolution in AS topology [6]. They used the BGP data to support their alternative explanation.
The objectives of this project are to:
Presentation slides and final report (PDF files).
Project Objective: To model a system that will transport a RTP packet stream over ATM Adaptation Layer 5.
Real-Time Transport Protocol (RTP) is an Internet protocol used for transmitting real-time data such as audio and video. It typically runs on top of the User Datagram Protocol (UDP) protocol, although the specification is general enough to support other transport protocols. UDP runs on top of IP networks, offering a direct way to send and receive datagrams over that network.
The Type 5 ATM adaptation layer (AAL) enhances the services provided by ATM, participating in the segmentation of data into 48-byte frames. The breakdown of the AAL type 5 framework consists of a common part and a service specific convergence sublayer, which may be defined to support specific user services. The adaptation layer supports the non-assured transmission of user data frames: it assumes that error recovery is provided by higher layers.
In this project, we propose to utilize the network modeling application Opnet Modeler 8.0 to model an RTP over ATM network. The network will initially compose of six unique nodes:
Our network will begin as a simplex connection, sending packets from Node (1) to (2) to (3) ... to Node (6). As our project progresses, the functionality of the RTP/UDP/IPv4 header compressor and decompressor will be joined into a single node type, as will the functionality of the RTP-AAL5 compressor and AAL5-RTP decompressor. Similarly, a single node type will generate and sink RTP packets. At this point, a duplex connection would consist of the following nodes:
The compression of RTP/UDP/IPv4 headers will follow the algorithm described in RFC2508 [2]. The packet converter algorithm adheres to the ideas written in the Encapsulation of Real-Time Data Including RTP Streams over ATM [3].
Time-permitting, we will also add a process in the RTP/AAL5 packet de/compressor to detect whether the incoming packets contain compressed or uncompressed RTP/UDP/IP headers. Uncompressed-header packets will then be allowed and transmitted over ATM in a similar fashion as described above.
Our purpose in choosing to do a project on RTP over ATM is due the team members' interest of both RTP and ATM standards and protocols.
References:
Improving TCP Performance with Periodic Disconnections over Wireless Links:
Presentation slides, demo slides, and final report (PDF files).
In theory, transport protocol such as TCP should be independent of the technology of the underlying network layer. In practice, it does matter because most TCP implementation have been carefully optimized based on assumptions that are for wired netwrok but which fail for wireless networks. The principal problem is the congestion control algorithm. Nearly all TCP implementations nowadays assume that timeout are caused by congestion, not by lost packets. Consequently, when a timer goes of, TCP slows down and sends less vigorously. The idea behind this approach is to reduce the network load and thus alleviate the congestion. But wireless transmission links are highly unreliable. They lose packets all the time. Many papers have been written proposing methods for improving TCP performance over wireless links, such as Berkley Snoop, Indirect TCP, WTCP etc. All these papers have, however, concentrated on only one problem associated with wireless links --- a perceived high BER over the wireless links.
While a high BER has significant implications for protocol performance, other limitations of the wireless environment are equally or more important than high BER. Frequent disconnection is one of these problems which are caused mainly by handoff, and physical obstacles blocking radio signals.
A solution proposed by Kevin Brown and Suresh Singh, M-TCP, was designed to work well in the presence of frequent disconnection events and over low bit-rate wireless links subject to dynamically changing bandwidth. In addition, it maintains end-to-end TCP semantics. The protocal structure may be viewed as a three-level hierarchy. At the lowest level are the mobile hosts(MH) who communicate with mobile support station(MSS) nodes in each cell. Several MSSs are controlled by a machine called the Supervisor Host(SH). The SH is connected to the wired network and it handles most of the routing and other protocol details for the mobile users. In additon it mantains connections for mobile users, handles flow-control and is responsible for maintaining the negotiated quality of service.
We'll simulate this M-TCP solution on OPNET and compaire its performance with other TCPs.
[References]
Implementation and Analysis of Megaco/H.248 Protocol Using OPNET:
Presentation slides and final report (PDF files).
[Description] Megaco/H.248 Protocol is a media gateway control protocol. Media gateway control protocols were established for the need of IP networks to interwork with traditional telephony systems and provide support for large-scale phone-to-phone deployments. Media gateway control protocols specify a master/slave architecture for decomposed gateways, in which Media Gateway Control (MGC) is the master server, and one or more Media Gateways (MGs) are the slave clients that behave like simple switches. One MGC can serve many MGs. While some other multimedia over IP protocols like Session Initiation Protocol (SIP) and H.323 are based on a peer-to-peer architecture.
Currently the best-known media gateway control protocols are Media Gateway Control Protocol (MGCP) and Megaco/H.248. MGCP was published as informational RFC2705 and has been widely deployed. As an evolution of MGCP conceptually, Megaco/H.248 is expected to win wide industry acceptance as the official standard because it is a collaborative effort of the Internet Engineering Task Force (IETF) and International Telecommunication Union (ITU), following an agreement by both bodies to cooperate on a single unified protocol.
In our project, we will implement an environment using OPNET to simulate Megaco/H.248. One MGC and two MGs will be in it. Also, we will implement some Megaco/H.248 commands on top of User Datagram Protocol (UDP) between MG and MGC that are required for basic call flows to demonstrate the functions of Megaco/H.248. The voice traffic between two MGs will be simulated using Real-time Transport Protocol (RTP). Finally, we will give out some statistical analysis on various network performance parameters, such as call setup time and average delay per call experienced by media packets.
[References]
Implementation of Start-time Fair Queuing Algorithm in OPNET:
Presentation slides and final report (PDF files).
Project Description:
In 1985 Nagle identified the need to guarantee network performance due to the potential for ill-behaved sources to dominate bandwidth [5]. Since then, Fair Queue (FQ) algorithms have been widely adopted as one of the keys methods to regulate traffic at packet forwarding-agents, such as IP routers [2]. A variety of algorithms based on the original FQ concept have been conceived, with different levels of fairness and complexity; one such algorithm is "Start-Time Fair Queuing" (SFC), which provides good fairness without great complexity. The SFC solution is of particular interest because it provides good performance for real-time traffic (which is very sensitive to delay [6]), something that many earlier techniques did not address [4].
In this project we intend to model SFC in Opnet. In addition, a simple network model using SFC will be built and simiulations will be run on it with different types of traffic to evaluate the behaviour of the algorithm, and to compare its performance with other scheduling schemes such as Weighted Fair Queue [3], Priority Queues and Vitual Clock [1].
Simulation will be done using OPNET
References:
TCP performance over satellite link:
Presentation slides and final report (PDF files).
Over the past few years, the demand to use satellite devices to access the Internet is growing because satcom can deliver Internet services to consumers and institutions in remote areas of the world not covered by good terrestrial connectivity. It is expected that TCP protocol will be frequently used over the Satellite network in near future.
Our project will focus on studying congestion control of variant types of TCP over satellite network. As satellite network is a typical long delay network, we prospect this project can fit most general long delay and error-prone link.
We will evaluate the impact of different parameters such as window size, file size toward throughput on TCP Vegas, Reno, Tahoe, and Sack both in error-free and error-prone links.
In addition we will try to analyze TCP's congestion avoidance algorithm which may result in drastically unfair bandwidth allocations according to multiple connections with different RTTs(rount trip time).
We plan to use ns-2 to implement and simulate above methods.
References:
Performance evaluation and enhancement of the wireless local area networks:
Presentation slides, demo slides, and final report (PDF files).
Unlike the large bandwidth that can be achieved by the wired networks, the bandwidth of wireless network is much more limited due to the ``air'', which is the physical medium used to transfer signal by WLAN, and it is not only error prone but also very ``expensive''. Hence, improving the performance of the WLAN is a very important topic to research.
In this project, I will explore several important issues related to improving the performance of the WLAN from the Link Layer (Media Access Control Protocol) to the Network Layer (Transfer Control Protocol). First, I will implement several types of back-off algorithms used in the MAC layer, and compare them with the standard back-off algorithm described in the IEEE802.11 reference. Then I will implement a link-layer protocol that is TCP-aware (such as SMART-snoop protocol). Finally, I will analyze the data gathered in the previous two steps to get the conclusion.
OPNET 8.0 is used to do the simulation.
References:
Comparison of Queuing Algorithms using OPNET Modeler (Smart Queuing: An Adaptive Approach):
Presentation slides, demo slides, and final report (PDF files).
Congestion control and Quality of Service (QoS) provision are important issues in todays high-speed networks. Packet scheduling can provide users with different QoS as well as ensure that the network is running efficiently. There are many packet scheduling or queuing algorithms, each has its own advantages and disadvantages. We investigated the performance of several different queuing mechanisms (FIFO, PQ, WFQ, CQ) using OPNET network simulation tool. A set of traffic parameters is identified that determines which mechanism will optimize the network performance in terms of throughput, delay, and loss rate. From this characterization, we introduced a smart queuing mechanism, one that adapts to the current traffic situation by dynamically changing between the different algorithms.
REFERENCES
Comparison of different congestion control algorithms (AIMD, TFRC and TCP):
Presentation slides, demo slides, and final report (PDF files).
Abstract
Congestion control in packet networks has proven to be a difficult problem in general. Properties. However, this problem is particularly challenging in the Internet. There are different congestion control algorithms for different requirements, TCP is used for reliable data transmission, and it can also be regarded as AIMD(1,1/2), which is a special kind of AIMD(a,b), TFRC is more suitable for transmission of data like voice and video which are sensitive to the variation of transmission rate.
In this project, we compare the performance of AIMD(a,b), TFRC and TCP under different conditions, with the performance index like transient response, traffic smoothness, throughput ratio and friendliness metrics in the scenarios of short-term congestion, long duration congestion and variation of background traffic. Through the statistics collected from these experiments, we can reach some conclusions on the advantages, disadvantages of these congestion control mechanisms.
Reference:
[1] Jamal Golestani, A Class of End-to-End Congestion Control Algorithms
for the Internet , Proceedings of ICNP, 1998.
[2] S. Kunniyur and R. Srikant, "End-To-End Congestion Control: Utility
Functions, Random Losses and ECN Marks", Longer version of the paper that
appeared in Proceedings, INFOCOM 2000, Tel-Aviv, Israel, March 2000.
Also submitted to IEEE Transactions on Networking
[3] S. Kunniyur and R. Srikant, "Fairness of Congestion Avoidance Schemes
in Heterogeneous Networks", Proceedings, International Teletraffic
Congress-16, Edinburgh, Scotland, 1999
[4] Yair Bartal, J. Byers and D. Raz, Global Optimization using Local
Information with Applications to Flow Control, STOC, October 1997.
[5] TCP Friendly Rate Control (TFRC): Protocol Specification,
Handley, M., Pahdye, J., Floyd, S., and Widmer,
J. Internet draft draft-ietf-tsvwg-tfrc-02.txt, work in progress, May 2001.