Overview

The Internet Real-Time Lab (IRT) conducts research in the areas of Internet and multimedia services: Internet telephony, wireless and mobile networks, streaming, quality of service, resource reservation, dynamic pricing for the Internet, network measurement and reliability, service location, network security, media on demand, content distribution networks, multicast networks and ubiquitous and context-aware computing and communication.

Research slides for 2006, 2002, 2001, 2000, 1999. Old overview../Overview slides.

Session Initiation Protocol (SIP)

SIP is a transport-independent application-layer signaling protocol for creating, modifying and terminating sessions (e.g., Internet conferences, IP telephony, events notification, instant messaging and presence) with one or more participants. It supports user mobility by proxying and redirecting requests to the user's current location.

Columbia InterNet Extensible Multimedia Architecture (CINEMA)

CINEMA is a flexible architecture that allows you to build several clients and servers for Internet multimedia services using Session Initiation Protocol (SIP), Real-time Transport Protocol (RTP) and Real Time Streaming Protocol (RTSP). We are setting up an Internet telephony infrastructure that will replace the traditional analog phones in our department. It provides a variety of interesting services like multi-party conferencing, voice and video mail, instant messaging, programmable servers, interworking with H.323, interactive voice response, interoperability with existing telephone networks, instant messaging and presence, SNMP based monitoring and control, customer billing and accounting, and so on. We are collecting performance metrics for sipd (proxy), sipconf (conferencing), rtspd (media server) and our SIP user agent stack to define benchmarks for large-scale IP telephony installations. Our software runs on variety of platforms, Unix as well as Windows. SIP stack is available in C/C++ and, in future, Java.

Intelligent multi-function SIP endpoint

sipc is a SIP user agent for Internet telephony calls, presence information retrieval, instant messaging, shared web browsing, Internet TV and network appliance control. It supports audio, video, text and white board, and can be extended easily for additional media types. sipc can support programmable services (using CPL and SIP CGI) such as time-of-day or presence based call handling. We are designing an XML-based Language for End System Services (LESS) specifically for end system service creation.

Call processing language (CPL)

CPL can be used to describe and control Internet telephony services. It is designed for SIP proxy, redirect and registrar (network) servers as well as user agents (end-points). It is simple, extensible, easily edited by graphical clients, and independent of operating system or signaling protocol.

Emergency services on Internet

We are building 911 and emergency notification systems for IP telephony. When a user dials 911 from regular phone or "sip:sos@domain" from IP phone, the outbound telephony server finds the caller's physical location and puts it in the call initiation message sent to the nearby Public Safety Answering Point (PSAP). SIP event notification framework is used to provide emergency notification to end-points, for instance, by flashing a light to help hearing impaired. We also plan to incorporate multimedia communications research for remote medical monitoring of heart patients.

Context aware ubiquitous computing

In a step towards ubiquitous computing we are building a context aware architecture in which user presence is detected using swipe card, iButtons, or IR/RF transceivers and the environment is modified as per user preference. For instance, when a user enters a room the stereo may start playing her favorite music, and the SIP server may forward her calls to the room telephone.

Mobile and ad hoc networking

We are working on mobile and ad hoc networks, in particular, peer-to-peer collaborative caching and delay tolerant ad hoc networking. The 7DS architecture for peer-to-peer data exchange is particularly useful when the mobile devices experience intermittent connectivity to the Internet. When a host cannot access the Internet to acquire the information, it queries other peers in close proximity over wireless LAN. Some of these may serve as ad-hoc, temporal gateways to the Internet. 7DS can enhance the collaboration by either content pre-fetching or query on demand, and thus creates an ad-hoc network to spread information. Particularly, for location-dependent data (e.g., weather or traffic reports) and popular data (music files, news) that do not change very rapidly. We have implemented 7DS and investigated the effects of the wireless coverage range, network size, query mechanism, peers' cooperation strategy and their power conservation mechanism on the information dissemination.

QoS measurement for VoIP

We have investigated the characterization of packet loss and delay in the Internet, comparing various loss models (Gilbert, extended Gilbert, trace-based model), and how they affect the perceived quality for VoIP systems. Our new perceived quality estimation method is based on automatic speech recognition. We studied the network aggregation behaviors of voice traffic using modern voice codecs and silence detectors, and found that aggregation efficiency is worse than assuming the commonly used exponential on/off voice traffic model. This helps to properly provision the VoIP network. We also found that forward error correction (FEC) has better end-to-end perceived quality than redundant codec based scheme. We have devised a way to optimize the FEC quality (listening quality and delay impairment) and studied the trade-off between loss-robust codec versus simpler FEC scheme. We are investigating the quality and availability of VoIP service using various test points on the public Internet.

Scalable network architecture for adaptive QoS support

The new value-added Internet services requires that the providers are able to add new services quickly and efficiently, and equip the network to meet the high quality and reliability expectations and diverse requirements. We have designed and implemented a scalable service framework that supports dynamic resource negotiation between users and network, as well as between peering network domains. It enables short-term resource reservation by the network and demand adaptation by user applications, as well as the pricing of network resources based on usage, QoS and user demand. The Resource Negotiation and Pricing (RNAP) protocol is particularly efficient in supporting services where resources are scarce, such as access networks, bottleneck web servers and wireless air interfaces. It provides performance benefits under high or bursty loads and provides QoS for wide range of user applications. It improves network utilization and user connectivity. It increases user benefit and network revenue, and reduces service-blocking rate.

QoS for large scale networks

Resource reservation must accommodate the rapid growth and increasing service diversity of the Internet. Border Gateway Reservation Protocol (BGRP) provides a distributed architecture for inter-domain aggregated resource reservation for unicast traffic that scales well, in terms of message processing load, state storage and bandwidth. BGRP relies on Differentiated Services for data forwarding. We use soft state to maintain reservations. In contrast to RSVP, refresh messages are delivered reliably, allowing us to reduce the refresh frequency. YESSIR (Yet another Sender Session Internet Reservations) is a new reservation mechanism that simplifies the process of establishing reserved flows while preserving many unique features introduced in RSVP. Senders generate reservation requests to reduce the processing overhead. It builds on top of RTCP, uses soft state to maintain reservation states, supports shared reservation and associated flow merging and is compatible with the Intserv models. It extends the all-or-nothing reservation model to support partial reservations that improve over the duration of the session.

Mesh-enhanced SLP

mSLP enhances Service Location Protocol with a fully-meshed peering Directory Agent (DA) architecture and provides reliable and consistent directory service. Peer DAs exchange service registration information and maintain the same consistent data for shared scopes. It also greatly simplifies service agent registrations in systems with multiple DAs. It is backward compatible with SLPv2 and can be deployed incrementally.

Analysis of LDAP performance

We have measured and analyzed the performance of Light-weight Directory Access Protocol for latency and scalability under various access patterns and investigated mechanisms for improving performance.

MarconiNet/Wireless IP telephony

MarconiNet defines a streaming architecture for next generation networks built upon standard IETF protocols SIP, SAP, SDP, RTP/RTCP, RTSP and IP Multicast. It deals with various operational issues (e.g., mobility, handoff, QoS, Channel management) associated with building a true IP based streaming network over a mobile Internet. It focuses on local server based program management, fast-handoff of multimedia stream, local advertisement insertion for real-time streaming traffic, and application layer mobility for interactive real-time traffic for a highly mobile environment including military type ad-hoc networks.

Interactive voice response

VoiceXML is used to create interactive voice dialogs (e.g., tele-banking, or voice mail access) that feature synthesized speech, digitized audio, recognition of spoken and DTMF key input and recording of audio for telephony applications. Voice applications can be written using existing tools like servlets and CGI that can be accessed via a telephone using our IP-telephony Voice-XML browser. We are developing applications to join an authenticated conference, voice mail access, and generic conference control in our IP-telephony test-bed.

SIP hardware phone

E*phone (Ethernet phone) is a replacement for your telephone handset in an IP-telephony environment. It uses Texas Instruments DSP and CRTX real-time operating system. Both hardware and software were built in our lab.

SIP-H.323 interworking

H.323 is an ITU-T's recommendation for multimedia conferencing over packet-based networks. For those who believe that both SIP and H.323 will co-exist in future it is important to define an interworking standard between the multi-stage call signaling of H.323 and simple request-response architecture of SIP.

Multi-Layer Utilization Maximal

(Receiver driven adaptive multimedia sessions) Typical fairness definitions for multicast networks involve sending data to all receivers at the same rate. In contrast, multi-rate approaches are more realistic in a heterogeneous Internet as they allow receiving data at different rates, based on receiver capability and network links' capacity. To accomplish intra-session and inter-session fairness in the presence of multi-rate sessions, we present a fairness policy based upon the number of receivers in a session. This improves bandwidth usage and provides an incentive for using multicast. We have implemented a distributed protocol to accomplish a fair bandwidth distribution between concurrent multicast trees that allows the receivers to adapt to network changes based on the fair rate of the multimedia session they belong.

IP multicast fault recovery

We have analyzed the multicast failure recovery problem (in PIM over OSPF) with simulation and test bed. This is useful for critical distributed multimedia applications that require high availability.

update my research information >>