I have created this page for keeping track of my reading materials. Here is the bib file .

Research Papers

[+] The Congestion Managers - Internet Drafts

Summary The draft describes the Congestion Manager (CM) module that uses the concept of Ensemble TCP sharing described in RFC 1379 - TCP Control Block Interdepence - to share the states (i.e congestions properties) amongst multiple connections from a sender side going to the same receiver, and thus perform congestion control. It describes about an API that enables application to know about the network states, share information with each other, pass information to CM, and uses a scheduler to distribute available bandwidth amongst the multple flows.

Reference H. BalaKrishnan, S. Seshan, “The Congestion Manager”, Internet Request For Comments, RFC 3124, June 2001

[+] An Integrated Congestion Manager Architecure for Internet Hosts

Summary Balakrishnan. et al presented a framework for network congestion management. Their network Congestion Manager(CM) maintains network congestion states, and provides an API for applications to learn network states, share among each other, and schedule data transmission. The API provides some functions such as query path status, data transmission scheduler, notification to CM upon successful transmission, states update upon congestion or successful transmission, and callbacks to application upon rate change. )

Reference H. BalaKrishnan, H. Rahul, S. Seshan, “An Integrated Congestion Manager Architecure for Internet Hosts”, Proc. ACM SIGCOMM, Cambridge, MA, September 1999

[+] Effects of Ensemble - TCP

Summary

Eggert. et al presented Ensemble TCP (E-TCP) where it utilized the concept of TCB information reusing and sharing among other connections. It has been designed to show the aggregated network transmission behavior of an ensemble (parallel TCP connections) as a single TCP/Reno connection. They also proposed that it can be used with other TCP extensions. Authors also compared their proposed E-TCP with persistent HTTP 1.1, and showed some benefits. Their architecture, however, did not discuss the issue of reusing the cached information when the network is idle after sometime, as the network properties might change during the idle period.

Reference

L. Eggert, J. Heidemann, J. Touch, Effects of Ensemble TCP, USC/Information Sciences Institute, December 1999

[+] TCP Control Block Interdependence


Summary

TCP is connection oriented protocol on top of IP, a connection-less best-effort delivery protocol, to provide reliable data dreams support. Each TCP connection maintain state of per-connection in a data structure called Transport Control Block (TCB) that holds information such as connection state, current round trip times and congestion control information. Below you will see a summary of the TCB. Local Process State holds the Pointers to send and receive buffers, Pointers to retransmission queue and current segment and Pointers to IP PCB.

Per Connection Shared State

1. Macro State – describes the protocol for establishing and maintaining shared state, and holds information such as: Connection state, timers, flags, and local and remote host numbers and ports

2. Micro State is used to maintain the reliability and congestion control when the connection has established. Send and receive window state(size*, current number), RTT and variance, cwd size, csw size ssthresh, max window seen*, MSS

TCB interdependence can be shared in two ways, 1) Temporal Sharing 2) Ensemble sharing.

1) Temporal sharing can be used to cache states of a closed connection, and this previous connection state can be used to instantiate a similar connection to avoid inefficiencies.
2) Ensemble sharing occurrs when an active host opens another concurrent connection. Touch [1] proposes that TCB states can be shared across multiple concurrent connections for improving the connection behavior.

T/TCP has implemented the temporal sharing [2] where RTT and MSS have been cached, but they suggested to put congestion control parameters for effective use of channel resources.

For ensemble sharing TCB information such as MSS and RTT can be shared by copying and congestion window sharing is done by giving fair share to the pending connection. If old sending congestion window is oldsend-cwd then new TCB is proposed in [1] : oldsend-cwd/N+1 and then subtract oldsend-cwd/N+1/N from each concurrent. However, the current implementation says new TCB Cwnd should start with 1 segment, whereas T/TCP suggested copying the cached oldsend-cwd.
Reference

1. J. Touch, “TCP Control Block Interdependence”, Internet Request For Comments, RFC 2140, April 1997
2. R. Braden, Transaction—“TCP concepts”, RFC 1379, USC/Information Sciences Institute, September 1992.

[+] Video Conferencing Tool (Vic)

Summary Video Conferencing Tool (Vic) is a real time multimedia video conferencing tool that supports various codecs such as H.26x, DV, MPEG etc. Vic is a single threaded systems where set of basic objects implemented in C++ are coordinated and controlled using TCL/TK. [1]

At the sender side, video device grabs frames, and then the encoder performs encoding – block based image compression, RTP framing and packet transmission. Congestion control module is sits in between of sender and receiver in order to provide the calculated sending rates. To regulate the overall transmission of Vic, it has two implementations of congestion control mechanism – (1) TCP Friendly Window based Congestion Control (TFWC) [2] uses cwnd and (2) TCP Friendly Rate Control (TFRC) uses inter packet arrival rate. Vic has two separate data channels: DataHandler – to convey RTP data packets, and ControlHandler – to convey RTCP control packets over separate UDP ports. It uses RTCP Extended Report (RTCP XR) to carry congestion control information. Diagram below shows the overview of vic architecture:

At the receiver side, it stores the arrived packets in the play-out buffer, de-frame RTP packets, and finally decodes the image frames.

References
[1] S. McCanne, V. Jacobson, “vic: A Flexible Framework for Packet Video”, ACM Multimedia'95, November 1995, San Francisco, CA, pp. 511-522
[2] S. Choi, “Congestion Control for Real-time Interactive Multimedia Streams”, University College London, Department of Computer Science, September, 2010.

[+] TCP-Friendly Rate Control for Bursty Media Flows

Summary This paper analyses the performance of two new TFRC improvements which were designed to provide better support for bursty media flows, and found that the revised TFRC can substantially improve the performance of other traffic that shares the same congested path. Moreover, This also removed the incentive to use padding to gurantee the required media rate by RFC 3448 which unnecessarily consumes the capacity, hence is undesirable for other flows sharing the same bottleneck

Reference
Sathiaseelan, Arjuna, and Gorry Fairhurst. "TCP-Friendly Rate Control (TFRC) for bursty media flows." Computer Communications 34.15 (2011): 1836-1847.

Important references
1. Handley, M., Floyd, S., Padhye, J., & Widmer, J. (2003). TCP friendly rate control (TFRC): Protocol specification.

[+] Analysis and Performance Evaluation of the EFCM Common Congestion Controller for TCP Connections

Summary Based on E-TCP, Savoric et al. proposed an Ensemble Flow Control Mechanism (EFCM) where a controller actively probes for the information from the flows, and calcualtes new rate for the flow by aggragating congestion properties (e.g. RTT, congestion window). They showed that EFCM considerably increases the throughput and fairness for the flows sharing the same bottleneck. However, their method does not provide any techniques for shared bottleneck detection.

Reference
Savoric, M., Karl, H., Schlager, M., Poschwatta, T., & Wolisz, A. (2005). Analysis and performance evaluation of the EFCM common congestion controller for TCP connections. Computer Networks, 49(2), 269-294.

[+] A Survey on Lower-than-Best-Effort Transport Protocols

Summary .......
References

[+] Less-than-Best-Effort service: A Survey of End-to-End Approaches

Summary .......
Reference .......
Important references .......

[+] Accurate Shared Bottleneck Detection Based on SVD and Outliers Detection

Summary .......
Reference .......
Important references .......

[+]

Summary .......
Reference .......
Important references .......