Time Synchronization for VoIP Quality of Service
IEEE Internet Computing
Minimizing Transmission Costs through Adaptive Marking in Differentiated Services Networks
MMNS '02 Proceedings of the 5th IFIP/IEEE International Conference on Management of Multimedia Networks and Services: Management of Multimedia on the Internet
TCPtransform: property-oriented TCP traffic transformation
DIMVA'05 Proceedings of the Second international conference on Detection of Intrusions and Malware, and Vulnerability Assessment
| Hi-index | 0.00 |
In recent years real-time multimedia services have become increasingly popular over the Internet, in particular voice over IP (VoIP). Nevertheless, several issues remain outstanding before the wide acceptance of these services. First of all, due to the Internet's best-effort nature, users may experience less than desirable quality and sometimes intermittent service. For example, Internet phone calls are often less clear than calls carried over the public switched telephone network (PSTN). Second, despite increasing popularity of consumer VoIP services, wide-area enterprise deployment remains rare. Third, for a service as critical as VoIP, the provider must offer a high level of service availability, where availability refers to how often the service is available, that is, a call can be successfully made and maintained, as opposed to the voice quality of a call. We tackle the first issue in two parts. In Part I of this dissertation, we present our measurement of Quality of Service (QoS) for Internet real-time multimedia services, which is important for both customer-provider service level agreement validation and for providing feedback to adaptive applications. First we discuss the analytical modeling of objective QoS metrics (loss, delay, jitter) and the techniques for measuring them. Second, we investigate the relationship between objective QoS and subjective quality, notably the Mean Opinion Score (MOS). This is done through a series of MOS listening tests, particularly on the effect of loss burstiness on subjective quality. Third, because MOS listening tests are time-consuming and labor intensive, we will introduce and evaluate an automated MOS estimation algorithm that we designed based on automatic speech recognition (ASR). In Part II, we analyze and compare QoS improvement measures. We first examine end-to-end based approaches, with an original study comparing the subjective quality and trade-off of loss recovery methods based on forward error correction (FEC), low bit-rate redundancy (LBR) and packet loss concealment (design of more error resilient codecs). Next, we study network-based QoS improvement approaches, specifically how the on-off patterns of human speech produced by modern silence detectors differ from the traditionally believed exponential model, and how this affects voice traffic aggregation performance. In Part III of this dissertation, we investigate the state of the art of VoIP in three aspects. First, since the quality of a service is no better than that of its weakest component, we evaluate VoIP end-points (IP phones and software clients), and discuss the level of quality that existing products provide. Second, to address the issue of enterprise VoIP deployment, we present a case study of IP telephony deployment in the department of Computer Science, particularly in the aspect of PSTN-IP integration. Finally, to investigate the last issue, we will describe an Internet measurement experiment we have conducted, and assess through various metrics the service availability of VoIP in today's Internet.