Quality of Service (QoS) measures are the network parameters; delay, jitter, and loss and they do not reflect the actual quality of the service received by the end user. To get an actual view of the performance from a user's perspective, the Quality of the Experience (QoE) measure is now used. Traditionally, QoS network measurements are carried on actual network components, such as the routers and switches since these are the key network components. In this thesis, however, the experimentation has been done on real video traffic. The experimental setup made use of a very popular network tool, Network Emulator (NetEm) created by the Linux Foundation. NetEm allows network emulation without using the actual network devices such as the routers and traffic generator. The common NetEm offered features are those that have been used by the researchers in the past. These have the same limitation as a traditional simulator, which is the inability of NetEm delay jitter model to represent realistic network traffic models, such to reflect the behaviour of real world networks. The NetEm default method of inputting delay and jitter adds or subtracts a fixed amount of delay on the outgoing traffic. NetEm also allows the user to add this variation in a correlated fashion. However, using this technique the outputted packet delays are generated in such a way as to be very limited and hence not much like real internet traffic which has a vast range of delays. The standard alternative that NetEm allows is generate the delays from either a Normal (Gaussian) or Pareto distribution. This research, however, has shown that using a Gaussian or Pareto distribution also has very severe limitations, and these are fully discussed and described in Chapter 5 on page 68 of this thesis. This research adopts another approach that is also allowed (with more difficulty) by NetEm: by measuring a very large number of packet delays generated from a double exponential distribution a packet delay profile is created that far better imitates the actual delays seen in Internet traffic. In this thesis a large set of statistical delay values were gathered and used to create delay distribution tables. Additionally, to overcome another default behaviour of NetEm of re-ordering packets once jitter is implemented, PFIFO queuing discipline has been deployed to retain the original packet order regardless of the highest levels of implemented jitter. Furthermore, this advancement in NetEm's functionality also incorporates the ability to combine delay, jitter, and loss, which is not allowed on NetEm by default. In the literature, no work has been found to have utilised NetEm previously with such an advancement. Focusing on Video On Demand (VOD) it was discovered that the reported QoE may differ widely for users of different age groups, and that the most demanding age group (the youngest) can require an order of magnitude lower PLP to achieve the same QoE than is required by the most widely studied age group of users. A bottleneck TCP model was then used to evaluate the capacity cost of achieving an order of magnitude decrease in PLP, and found it be (almost always) a 3-fold increase in link capacity that was required. The results are potentially very useful to service providers and network designers to be able to provide a satisfactory service to their customers, and in return, maintaining a prosperous business.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:766205 |
| Date | January 2018 |
| Creators | Roshan, Mujtaba |
| Publisher | Queen Mary, University of London |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://qmro.qmul.ac.uk/xmlui/handle/123456789/44049 |
Page generated in 0.0027 seconds