Adult Web Sites and Hosting

280 Wikstrand Copyright 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. Bianchi, G. (2000). Performance analysis of the IEEE 802.11 distributed coordination function. IEEE Journal on Selected Areas in Communication, 18(3), 535-547. Bocheck, P., Campbell, A., Chang, S. -F., & Liao, R. -F. (1999). Utility-based network adaptation for MPEG-4 systems. Proceedings of the 9th International Workshop on Network and Operating System Support for Digital Audio and Video (pp. 279- 288). Florham Park, NJ: AT&T Labs. Research. Bryant, J., & Raney, A. A. (2000). Sports on the screen. In D. Zillman & P. Vorderer (Eds.), Media entertainment - the psychology of its appeal (pp. 153-174). Mahwah, NJ: Lawrence Erlbaum Associates. Cronin, E., Filstrup, B., Kurc, A. R., & Jamin, S. (2002). An efficient synchronization mechanism for mirrored game architectures. Netgames 02: Proceedings of the 1st Workshop on Network and System Support for Games (pp. 67-73). New York: ACM Press. Curcio, I. (2002). Mobile video QoS metrics. International Journal of Computers & Applications, 24(2), 41-51. de Waard, D. (1996). The measurement of drivers mental workload. Unpublished doctoral dissertation, University of Groningen. Diot, C., & Gautier, L. (July-Aug. 1999). A distributed architecture for multiplayer interactive applications on the internet. IEEE Network, 13(4), 6-15. Fodor, G., Eriksson, A., & Tuoriniemi, A. (2003, July). Providing quality of service in always best connected networks. IEEE Communications Magazine, 41(7), 154- 163. Fong, A. C. M., & Hui, S. C. (2001). Low-bandwidth internet streaming of multimedia lectures. Engineering Science and Education Journal, 10, 212-218. Ghinea, G., & Thomas, J. P. (2000). Impact of protocol stacks on quality of perception. Proceedings of the International Conference on Multimedia and Expo (pp. 847- 850). New York: IEEE. Gutwin, C., Benford, S., Dyck, J., Fraser, M., Vaghi, I., & Greenhalgh, C. (2004). Revealing delay in collaborative environments. CHI 04: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 503-510). New York: ACM Press. Hands, D., & Wilkins, M. (1999). A study of the impact of network loss and burst size on video streaming quality and acceptability. In M. Diaz, P. Owezarski, & P. S nac (Eds.), IDMS 99 (pp. 45-57). Berlin, DE: Springer Verlag. Hazas, M., Scott, J., & Krumm, J. (2004). Location-aware computing comes of age. Computer, 37(2), 95-97. Jin, J., & Nahrstedt, K. (2004). QoS specification languages for distributed multimedia applications: A survey and taxonomy. IEEE Multimedia, 74-87. Kies, J. K., Williges, R. C., & Rosson, M. B. (1997). Evaluating desktop video conferencing for distance learning. Computers & Education, 28, 79-91. Knecht, R. S., & Zenger, B. R. (1985). Sports spectator knowledge as a predictor of spectator behavior. International Journal of Sport Psychology, 16, 270-279.

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Clan Web Hosting services

A Three-Layer Framework for QoS-Aware Service Design 279 Copyright 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. Summary This paper has presented a three-layer model of networked services and applications. The model is based on three layers: usage, application, and network. By setting targets for and measuring utility in appropriate layers of the model, it becomes possible to consider new ways to solve problems. Changes can be made in or between each of the levels as shown through examples. The examples show that a problem in the network layer, which unaddressed would have to be dealt with in the usage layer, can be addressed in the network layer itself, in the application layer, or even in the usage layer. The next step is to use the model for run-time adaptation. Adaptation can be performed in some measure in each layer of the model, and might also include negotiations between the layers. The main contribution of this paper is the presentation of the three-layer model and its generalization to different applications and usages. References Aggarwal, S., Banavar, H., Khandelwal, A., Mukherjee, S., & Rangarajan, S. (2004). Accuracy in dead-reckoning based distributed multi-player games. Proceedings of ACM SIGCOMM 2004 Workshops on Netgames 04 (pp. 161-165). New York: ACM Press. Barber, P. J., & Laws, J. V. (1994). Image quality and video communication. In R. I. Damper, W. Hall, & J. W. Richards (Eds.), Proceedings of the IEEE International Symposium on Multimedia Technologies and Future Applications (pp. 163-178). London: Pentech Press. Figure 5. Effects on mental effort of getting feedback about the delay

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Clan Web Hosting services

278 Wikstrand Copyright 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. between different entities by using time buckets or breathing time buckets (Diot & Gautier, 1999; Steinman, 1993). In this type of solution, each message is delayed at the receiving end for a period of time, which is dependent on the known delay between the sending and receiving processes, and until it is safe to assume that all recipients have received all the messages for that time step. Each of the three different types of solutions creates problems in the usage layer while attempting to solve problems in the network layer. Dead reckoning leads to inconsistencies when it turns out that the simulated track differs from the actual track. Time warp creates huge inconsistencies when the game suddenly moves back to a previous state. Time buckets introduce additional delay but mostly avoids inconsistencies. Proposed Solution We studied a solution which made a change in the application layer (Wikstrand, Schedin, & Elg, 2004). It was aimed at making the user in the usage layer more adapted to the delay problems in the network. In a simple two-player game, pong, we introduced a delay as well as some sort of feedback to the user about the level of the delay (Figure 4). We performed an experiment where we let two users play pong against each other for about sixty seconds per round. We then varied the level of the delay and how the gravity of the delay was presented to the players. It turns out that having the feedback does let the user adjust to the level of the delay in a better way, that is, the mental effort was more adapted to the delay and the user was more tolerant to the delay (Figure 5). Other researchers have performed a similar study regarding collaborative virtual environments (Gutwin, Benford, Dyck, Fraser, Vaghi, & Greenhalgh, 2004). Figure 4. The pong game used in the project; a faint shadow on the left provides feedback to the user about the delay

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Clan Web Hosting services

276 Wikstrand Copyright 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. would be through a different kind of network, for example, WCDMA or another 3G technology. What we saw as the biggest problems here were access cost and service level variability from the network. As the user moved around, and as other conditions changed in the network, different bandwidth would be available to the user at different costs. We saw a similar problem in the usage layer. The user would have different levels of interest during a game, and also would not be interested in paying for maximum bandwidth throughout a game, as that would be prohibitively expensive. Proposed Solution We devised a solution in the application layer which would meet the requirements posed by the network in the form of variable service levels and from the user in terms of different cost as a function of time. Our proposed solution was to code the game both as video (Figure 2(a)) and as positions (Figure 2(b)). The positions would then be rendered as simple graphics in the user terminal. Streaming the positions would require less than 1 kbps while the video might go as high as 350 kbps. Together the two different modes provided a wide range of requirements on the network. The system would then switch between different levels of video quality and the simple position-based graphics, based on input from the user as well as network conditions and the cost of available bandwidth. We performed experiments based on short (about 90 seconds) sections of a game which we had encoded in three different ways: as animations, as low-quality video, and as high quality video. The experiments were based on users watching and rating the clips as described above. There were a total of 89 participants in the experiment. We also tested the users levels of sports fandom (Wann et al., 2000) and their level of knowledge about the sport (Knecht & Zenger, 1985). Both of these variables have been shown to affect how people react when spectating sports. It turned out that the position-based animations were more popular than we had imagined, especially among users who were not as knowledgeable in the target sport in comparison to low bandwidth (20 kbps) video. The animations were easier to understand while the Figure 2: (a) A video frame from the experiment and the corresponding animation; (b) frame

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Inexpensive Web Hosting services

A Three-Layer Framework for QoS-Aware Service Design 277 Copyright 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. video was more emotional. Figure 3 illustrates how a secondary task affected the perceived emotional effectiveness and the video quality. Video quality was judged to increase with low bit-rate video when the secondary task was harder. (Wikstrand & Eriksson, 2002; Wikstrand, Eriksson, & stberg, 2003; Wikstrand & Sun, 2004) Example: Networked Games A basic problem in networked games is that one or more users might be delayed in relationship to each other or to a central server if there is one. This will have an effect on the usage layer, as users will try to compensate for the resulting lack of responsiveness and inconsistent behavior which results (Vaghi, Greenhalgh, & Benford, 1999). Many researchers have looked at the gravity and impact of the problem and on how to compensate for it on the network and the application layer. Possible Solutions Solutions on the network layer are mostly concerned with or based on providing or obtaining differential quality of service or guaranteed quality of service. On the application layer, many different solutions have been suggested. The most common might be so called dead reckoning (Aggarwal, Banavar, Khandelwal, Mukherjee, & Rangarajan, 2004). Other solutions are based on being able to go back to a previouslyknown safe state, for example, time warp and trailing state synchronization (Cronin, Filstrup, Kurc, & Jamin, 2002). The first solution is based on retracing the execution back to a safe point. The second solution is a simplified version of the first. In it, a delayed version of the game is run as a special process. When a problem occurs, its state is copied to all participating entities. Yet another type of solution is to avoid inconsistency Figure 3. Significant interactions between coding and secondary task

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Inexpensive Web Hosting services

A Three-Layer Framework for QoS-Aware Service Design 275 Copyright 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. is arguably a usage layer characteristic, and perceived video quality. This shows that the designer must also consider the context of usage. It might not be possible to provide an exhaustive list of all interesting interactions between the usage layer and the other two layers, and it shall not be attempted here. Suffice it to say that just as it was with objectives and measures as discussed above, so is it here. The specifics of the usage domain will have a vast influence on what interactions are possible and important. In the following, I shall give examples of how the three layer model was used in projects in which I were involved. Example: The Arena Project I was working on a project5 where we were investigating how to get an interactive, multimedia service distributed to a huge number of users in a sports arena. Each user would have a hand-held computer connected through a wireless local area network. We realized that the biggest bottleneck was the network layer. Performance in an IEEE 802.11 network will decrease dramatically with an increasing number of users, both in terms of system throughput and user throughput (Bianchi, 2000). Proposed Solutions On the usage layer, we wanted to give the user the impression that he or she had a fully interactive experience with instant access to live and recorded video from the game. The live video could be provided through multi-casting a video stream. The recorded video was trickier. We decided that we would have to find a solution on the application layer, and we devised a distributed pre-caching scheme (Svanbro, Lindahl, Jonsson, Rosenqvist, & Wikstrand, 2003). This scheme also relied on multi-cast. The scheme was evaluated through simulations and by building a prototype, but because of the nature of how it would be used, with a large number of users, it was not deemed practical to perform a full scale user test. Through trials and simulations, we discovered that multi-cast in an IEEE 802.11 network suffers from a huge drawback. Packets will often collide with other traffic and be lost. We devised a media access algorithm which would solve the problem by changing the way that multi-cast worked in such networks (Nilsson, Wikstrand, & Eriksson, 2002). The solution was tested in simulations but not on real users. So in summary, the usage situation with a lot of users in a limited area who wanted to use bandwidth-intensive applications which would overload the network demanded a solution where we worked with both the application and the network layer. Example: The Bastard Project An unsolved problem from the previous project was how to deal with a situation where a user wanted to access the same services but away from the arena where the access

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Inexpensive Web Hosting services

274 Wikstrand Copyright 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. problems in the network. These examples might be seen as far-fetched and maybe even as some kind of cheating . Their main purpose is to get the reader started in thinking in new ways to take advantage of the interactions between the layers.4 Interactions with the Network Layer Interactions between the network and the application layers have been described in much detail elsewhere. The reader should start by looking at the references provided earlier in this chapter. In general, the idea is that the application layer and the network layer will affect each other. If, for instance, the bit-rate goes down at the network layer, the application layer might adapt in various ways. One way might be to demand of the network layer that it again increases the bit-rate, for example, by switching to a different network provider. A more realistic approach in most cases might be to make changes at the application layer. This could be done by changing, for example, the codec, the frame rate, or the spatial resolution. Anyone who has used a mobile phone might have noticed another way to deal with the connectivity problem at the network layer through interactions with another layer, in this case the usage layer. An icon advises the user of reception conditions. The user might then use this information to first of all understand that connectivity is non-existent or poor and might then use this information to take action, for example, defer calling, finding another way to communicate, and so forth. See below for a similar solution for networked games. Interactions with the Application Layer The interactions between the application layer and the usage layer are perhaps the most important and interesting ones in the context of this chapter. Above, it was demonstrated that the content might be more important to the user than the application layer quality. That does not mean that if all else is equal, that the application layer quality is unimportant. On the contrary, as shown in the introduction, more is indeed generally better. In Mueller and Agamanolis (2005), the information being transmitted between the two sites is exactly the same, how a ball travels through a room and hits a screen, but in one case the user interacts with the application through a traditional computer interface, and in the other case through a physical exertion interface. It turns out that not only did the users have more fun and get to know the other player better when using the exertion interface, they also thought that the transmitted audio and video quality was better. This is an example of how changes at the application layer (the interface) lead to changes in the usage layer (the interaction), which in turn leads to changes in how the application layer quality is perceived. But there are simpler interactions between the two layers. In the Bastard project (see below) we were able to show a significant interaction between secondary task load, which

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Inexpensive Web Hosting services

A Three-Layer Framework for QoS-Aware Service Design 273 Copyright 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. Mental workload was measured using the Rating Scale Mental Effort (RSME) (Ziljstra, 1993). The RSME has been validated and found to be accurate for workloads of around 60 90 seconds (Verwey & Veltman, 1996). Task performance was measured objectively where possible, that is, through a quiz after each clip in the Bastard project (see below) and as the score in the networked games project (also below). Enjoyment was measured through self-report. Certainly, a host of other measures have been proposed ranging from measuring hormone levels to observations of laughs per hour . It was felt unnecessary to go to such lengths for this research. Design Tool The important issue is to be able to use the model as a design and analysis tool. Its justification lies in its ability to serve as a useful tool to those who want to build an optimal service. Finding Out What is Important First of all, just like in any human-computer related activity (and as explained above), it is necessary to find out what is important at the usage layer. This chapter is not intended as a full fledged guide on design methodology, but I shall nevertheless mention some activities which I have found to be useful here. Reading scholarly works concerning what is important to persons involved in the activity: In the case of the Arena and the Bastard projects (see below), one of the main sources was Wann, Melnick, Russell, and Pease (2000). Interviewing current and potential users of the application: Prototypes at various levels of sophistication might be useful to focus the discussion. Involving a subject-matter expert in the design team: In the Bastard project, one of the team members was a player in the Swedish national football team. Looking at similar work as presented in scholarly or commercial contexts: This might show what is possible and desirable, and might also give new ideas. Interactions Between Layers One of the purposes of the model is to make us consider each of the three layers and the interactions between them. For instance, we can drop application layer quality if the home team is doing badly so that the user does not have to see them crushed . Or we can implement a synchronization algorithm in the application layer to compensate for

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Inexpensive Web Hosting services

272 Wikstrand Copyright 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. Some measures try to predict, based on objectively measurable features of the video, and so forth, how a human judge would subjectively rate the service level. Such measures include the moving pictures quality metric (Verscheure et al., 1998) and the perceptual distortion metric (Winkler, 2001). Usage Layer It is under this heading that we should find the mythical utility mentioned above. However, no such single variable can be identified, and we shall have to be content with considering several which will be used together in various combinations to form the overall utility. In traditional Human Factors work, three important objectives are usually considered. The worker should be as productive as possible while minimizing any harm done to him or her. He or she should also be able to find some kind of intrinsic motivation or enjoyment in performing the assigned task. The balance shifts between these three objectives, depending on the situation. For instance, a person who might not want to risk his or her health in the least way in the line of duty might happily go parachuting or swimming with sharks in his or her spare time. In the usage layer, the most meaningful measure of the harm done to the user might be task-induced stress or mental workload (de Waard, 1996; Wickens, 1992; Wilson & Sasse 2000). Task performance will obviously be measured in different ways depending on the actual task and usage situation. For video, examples include the ability to read lips, recognize faces and emotions, understanding of the presented material, and recall of things presented in the session (Barber & Laws, 1994; Procter et al., 1999; Kies et al., 1997). Specifically for conversational multimedia measures might include how well meaning is conveyed and agreed upon (Veinott, Olson, Olson, & Fu, 1999). For games, usage layer quality measures can include the level of fun and enjoyment as well as the ability to win the game or get a high score. In any case, it is crucial that the designer understands the usage area and what might be the relevant factors which influence the three main objectives in that particular context. Unlike in the other two layers where most of the same measures can be used regardless of what the actual context is, here it will be more specific. It is also important to consider how pre-existing differences between users will affect the actual measurements. For instance, it will be shown below how users subject-matter expertise affected their preference for different application layer encodings. Measures Used by the Author As we shall see in the next section, the author has used this model in a variety of contexts. In all these cases, a basic model for measuring usage and, to some extent, application layer quality has been used. The model is largely based on self-report measures, as these are easier to administer and more likely to be possible to use in a semi automated process.3

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Inexpensive Web Hosting services

A Three-Layer Framework for QoS-Aware Service Design 271 Copyright 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. Quality of Service (QoS) is a complex concept in many ways. On one hand, it refers to the quality of service as offered by a layer to the layer above it and as judged by the upper layer. On the other hand, it refers to the mechanism used to provide that quality of service and is measured in terms relevant to the layer offering the service (Pulakka, 2003). Because of this dichotomy, it is important to know what we talk about when we talk about quality of service. Who is the judge of the quality and on which scale is it measured? How can measures used in one layer be compared to measures used in another layer? What is necessary to do in each layer in a given situation to achieve the intended goals? In this section, a brief review of quality of service measures which have been proposed or used by other researchers is presented. Placing the quality measures in the model is important because it allows us to decide which measures are more important in a given application.2 Network Layer In the network layer mostly objective performance measures are used. The most basic concepts are delay and bit rate (sometimes called bandwidth) (Wang, Clayppol, & Zuo, 2001). Delay variability is called jitter and is sometimes more important than the delay itself (Wang et al., 2001). Especially in wireless networks, packet loss is an important measure which affects both of the previously-mentioned variables (Fong & Hui, 2001; Ghinea & Thomas, 2000; Hands & Wilkins, 1999; Koodli & Krishna, 1998; Patel & Turner, 2000; Verscheure, Frossard, & Hamdi, 1998). Obviously, if a packet is lost, the bit-rate goes down, and in certain protocols there will be a delay when the packet is retransmitted. Application Layer On the application layer, there is a rich flora of service level measurements. In streaming video and conversational multi-media, for instance, there are a host of measures, both objective and subjective, regarding the fidelity of the video compared to an imagined original which was not transported over a network. There are many subjective fidelity measures. In the double stimuli continuous quality scale, a human judge expresses their level of preference between an original and a degraded version (Winkler, 2001). In many other approaches, the judge is asked to compare the degraded version with an imagined broadcast quality or the live event (Kies, Williges, & Rosson, 1997; Procter, Hartswood, McKinlay, & Gallacher, 1999). In cases where only a small level of degradation is to be expected, an error detection and annoyance level approach might be feasible (Moore, Foley, & Mitra, 2001; Steinmetz 1996). There are also many objective measures of service levels here. For video, peak signal to noise ratio is often used, for example, when comparing video codecs (Verscheure et al., 1998; Winkler, 2001). Other common measures are bit-rate, frame rate, spatial and chromatic resolution, and so on (Wang et al., 2001).

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Clan Web Hosting services