J. Padhye and J. Kurose. An empirical study of client interactions with a continuous media courseware server. In Proc. of ACM NOSSDAV, July 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....result in further duplication of message data. Thus storage resources, which include active disk space and their protective back up systems, are needlessly consumed with redundant static data. Wasted Resources Recipients of CM content will often only want to render small fractions of messages [10]. Recipients will question paying for the delivery and storage of entire CM messages when they only render small fractions of messages they receive. Furthermore, in the case of distribution lists, a recipient may not render any of the CM content for certain messages, particularly for messages ....

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media course- ware server. In Proceedings of NOSSDAV '98, Cambridge, UK, Jul 1998.

....workloads is crucial to properly designing and provisioning current and future services. Recently, there have been several studies attempting to uncover the multimedia workloads characteristics. However, most of the studies are devoted to the analysis of workloads for educational media servers [1,2,3,12,13,16]. One recentstudy[9]characterizes the workload of a media proxy of a large university. Our paper presents and analyzes the enterprise media server workloads based on the access logs from two different media servers in Hewlett Packard Corporation. Both logs are collected over long period of time ....

....high temporal locality of accesses, the special client browsing pattern showing clients preference to preview the initial portion of the videos, and that rankings of video titles by popularity do not fit a Zipfian distribution. Recent studies on client access to MANIC system audio content [16] and low bit rate videos in the Classroom2000 system [13]provide the analysis of accesses to educational media servers in terms of daily variation in server loads, distribution of media session durations, and some clientinteractivity analysis. Extensive analysis of educational media server ....

J.Padhye, J.Kurose. An Empirical Study of Client Interactions with Continuous-Media Couseware Server. Proc. 8th Int'l. Workshop on Network and Operating System Support for Digital Audio and Video (NOSSDAV 1998.

....to post on servers or to redistribute to lists, requires prior specific permission and or a fee. NOSSDAV 02, May 12 14, 2002, Miami, Florida, USA. Copyright 2002 ACM 1 58113 512 2 02 0005 . 5.00. most of the studies are devoted to the analysis of workloads for educational media servers [1, 2, 3, 10, 11, 14]. One recent study [9] characterizes the workload of a media proxy of a large university. This paper presents and analyzes the enterprise media server workloads based on the access logs from two di erent media servers in Hewlett Packard Corporation. Both logs are collected over long period of time ....

....observed high temporal localityof accesses, the special client browsing pattern showing clients preference to preview the initial portion of the videos, and that rankings of video titles by popularity do not t a Zip an distribution. Recent studies on client access to MANIC system audio content [14] and low bit rate videos in the Classroom2000 system [11] provide the analysis of accesses to educational media servers in terms of daily variation in server loads, distribution of media session durations, and some client interactivity analysis. Extensive analysis of educational media server ....

J.Padhye, J.Kurose. An Empirical Study of Client Interactions with Continuous-Media Couseware Server. In Proc. of the 8th Int'l. Workshop on Network and Operating System Support for Digital Audio and Video, July 1998.

....Multimedia Presentation Techniques In the course of developing our six instructional modules as well as the nationwide movement towards instructional technology, we have explored a wide variety of multimedia presentation techniques. Much of our initial content was presented using the MANIC [3] courseware, developed here at UMASS with NSF support. MANIC uses a Web based format that delivers audio (and limited video) using the Real Media format synchronized to HTML pages for the presentation graphics. Features include the ability to search through the HTML pages as well as highlighting ....

J. Padhye, J. Kurose, "An Empirical Study of Client Interactions with a Continuous-Media Courseware Server", Network and Operating System Support for Digital Audio and Video, 1998.

....and synthetic generation of streaming access workloads of fundamental importance in the evaluation of Internet and streaming delivery systems. Over the last few years, there have been a small number of studies that attempted to characterize streaming media workloads [1] 2] 3] 11] 21] [26]. However, to our knowledge, all these studies targeted pre recorded, stored streams (e.g. news clips, film trailers, educational clips) and none has considered the characterization of live streams (e.g. camera feeds) This paper provides such a characterization for a unique data set capturing ....

....relevant to some aspects of our work) is outside the scope of this paper. Thus, in the remainder of this section, we restrict our coverage of related work to studies of streaming media workload characterization and synthesis. Streaming Media Access Characterization: Several previous studies [26], 18] 2] 11] 3] have characterized workloads of pre recorded media object access primarily from media servers for educational purposes. We summarize these efforts below. Padhye and Kurose [26] studied the patterns of user interactions with a media server in the MANIC system. They ....

[Article contains additional citation context not shown here]

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proceedings of NOSSDAV, June 1998.

....request an object from the beginning and play it without interruption to the end. It is unknown how these techniques scale in a non sequential access environment, in which clients may request the segments of an object. Indeed, recent studies on the characterization of streaming access workloads [34, 25, 12, 4] have revealed that client access is seldom sequential due to frequent client inter activity. While several studies have tried to minimize the bandwidth requirement for non sequential access in Video on Demand servers [5, 31, 32, 1, 11, 35] it is still unknown what are the potentials and ....

....the beginning of an object. After receiving a segment of the object (the ON segment) the client skips a portion of the object (the OFF segment) This process repeats until a request or a jump goes beyond either end of the object. In prior studies that characterized streaming access workloads [34, 4], it has been observed that the distributions of ON segments tend to be heavy tailed. In particular, the Pareto distribution was found to be a close fit. Thus, in our simulations, we generated requests that exhibited such properties. For ON periods, we used a Pareto distribution with parameter # ....

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proceedings of NOSSDAV, June 1998.

....relevant to some aspects of our work) is outside the scope of this paper. Thus, in the remainder of this section, we restrict our coverage of related work to studies of streaming media workload characterization and synthesis. Streaming Media Access Characterization: Several previous studies [26, 18, 2, 11, 3], have characterized workloads of prerecorded media object access primarily from media servers for educational purposes. We summarize these efforts below. Padhye and Kurose [26] studied the patterns of user interactions with a media server in the MANIC system. They characterized session length ....

....characterization and synthesis. Streaming Media Access Characterization: Several previous studies [26, 18, 2, 11, 3] have characterized workloads of prerecorded media object access primarily from media servers for educational purposes. We summarize these efforts below. Padhye and Kurose [26] studied the patterns of user interactions with a media server in the MANIC system. They characterized session length and user activity within a session. A session was considered a sequence of alternating ON periods (when the user is retrieving the media) and OFF periods (when no media is being ....

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proceedings of NOSSDAV, June 1998.

....when swapping is disabled, as well as when type D2 thresholding is employed. We resort to simulation to explore these remaining cases. As before, we assume arrivals are described by a Poisson process. Service times are described by a lognormal distribution, as has been observed in practice by [1, 30, 31, 32]. The probability density function of the lognormal distribution is defined as f lognormal (x) x p 2 exp( log(x) 2 ) where log(x) is the natural logarithmic function and and are the standard parameters used within the lognormal distribution. We used the mean and ....

J. Padhye and J. Kurose, "An empirical study of client interactions with a continuous-media courseware server," IEEE Internet Computing, Apr. 1999.

....request an object from the beginning and play it without interruption to the end. It is unknown how these techniques scale in a non sequential access environment, in which clients may request the segments of an object. Indeed, recent studies on the characterization of streaming access workloads [34, 25, 12, 4] have revealed that client access is seldom sequential due to frequent client interactivity. While several studies have tried to minimize the bandwidth requirement for non sequential access in Video on Demand servers [5, 31, 32, 1, 11, 35] it is still unknown what are the potential and ....

....session from the beginning of an object. After receiving a segment of the object (the ON segment) the client skips a portion of the object (the OFF segment) This process repeats until a request or jump goes beyond the last byte of the object. In prior studies that characterized streaming access [34, 4], it has been observed that the distributions of ON segments tend to be heavy tailed and that the Pareto distribution was found to be a close t. Thus, in our simulations, we generated requests that exhibited such properties. For ON periods, we used a Pareto distribution with parameter = 2 and ....

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proceedings of NOSSDAV, June 1998.

....multicast streaming methods for stored content. 1. INTRODUCTION This paper provides an analysis of the server log data for two media servers in use at major public universities in the United States. Both servers contain higher quality videos than servers or client workloads previously analyzed [1,9,12,14,17]. The eTeach system [21] was introduced in September 2000 to deliver the lectures and laboratory demos for a computer science course with an enrollment of 280 students. A key feature of this course is that there are no classroom lectures; the students obtain all course content from the server. ....

....not media files or were not found on the servers were removed from each log. 2.2 Related Work There have been several studies of Web workloads [2,3,4,5,6,7,8] but those workloads did not include significant access to media content. Recent studies of client access to MANIC system audio content [17], low bitrate videos in the Classroom2000 system [14] a mix of education and entertainment videos in the mMod system [1] and education content on an internal server of a large international corporation [12] have provided data on particular aspects of those media server workloads (e.g. example ....

[Article contains additional citation context not shown here]

J. Padhye and J. Kurose, An Empirical Study of Client Interactions with a Continuous-Media Courseware Server, Proc. NOSSDAV '98, July 1998.

....to indicate that the framework was effective. In the first two years where MANIC courseware was available, we found that on campus students primarily used MANIC materials for review and while working on homework assignments. This observation is supported by our analysis of the MANIC logs [11] that show students average 45 minutes per MANIC session; 50 used the text only features (not the audio track) and the average time on play out was 7 minutes. Anecdotally, the students found the search index capability of MANIC particularly useful in locating materials related to ....

Padhye J. and Kurose, J. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. IEEE Internet Computing (April

....as well as the variable size variablecost nature of Web objects. The emergence of streaming media applications on the Internet have resulted in an increased interest in e ective streaming media delivery techniques. Recent studies have focused on streaming media workload characterization [2, 4, 5, 11, 23] and synthesis [18] as well as caching techniques [3, 19, 28, 32, 33] Acharya et al.: characterized streaming objects [2] and user access patterns [4] Their work revealed several observations, including: the highly variable object sizes, the skewed popularity of objects, and the existence of ....

....popularity pro le that ts a Zipf like distribution. Almeida el al: 5] analyzed workloads from two media servers used for educational purposes. They studied request arrival patterns, skewed object 17 popularity, and user inter activities extending results obtained earlier by Padhye and Kurose [23] on user inter activities with a media server. Several studies proposed caching techniques for streaming media objects. Aiming to reduce network bandwidth requirements, Wang et al.: 33] proposed that proxy servers cache the part of a stream with high bit rate. Their scheme is called video ....

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proceedings of NOSSDAV, June 1998.

....and streaming delivery systems. Over the last few years, while many studies have considered the characterization of HTTP workloads [6, 7, 9, 11, 14, 15, 19, 22, 28, 30] and synthesis of HTTP request streams [8, 10, 34, 35] only very few studies focused on characterizing streaming media workloads [1, 3, 5, 13, 29], and none has tried to generate representative streaming media workloads. Because HTTP requests and streaming accesses are di erent, HTTP request generators are not suitable for generating streaming access workloads. These di erences include the duration of the accesses, the size of the objects, ....

....Almeida el al: 5] analyzed workloads from two media servers for educational purposes. They studied the request arrival patterns, skewed object popularity, and user inter activity times. Examples of characterization e orts targeted at non web environments include the work of Padhye and Kurose [29], which studied the patterns of user interactions within a media server, and the work of Harel et al.: 20] which characterized a workload of media enhanced classrooms, and observed user inter activity such as jumping behavior . In Section 3, we incorporate many of these characteristics in the ....

[Article contains additional citation context not shown here]

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proceedings of NOSSDAV, June 1998.

....experience. For example, data collected from earlier MANIC courses had suggested that most students would view a course only for about 30 minutes before switching to another activity, and that they often used the material in a nonlinear way, focusing on topics of interest and skipping others [10]. Working from this information, we designed shorter, tightly focused segments for the presentation and made use of the keyword search feature to let students easily locate topics of interest. E mail feedback suggests that users of the tutorial find it helpful and easy to use, and our usage logs ....

J. Padhye and J. Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. In IEEE Internet Computing (April 1999).

....recordings of the lecture. The captured and processed class materials can later be accessed by students to review the contents of a lecture, view a missed lecture, study a difficult concept or prepare for an examination. Media enhanced classrooms are becoming increasingly common in universities [11, 7, 15, 14, 4]. In addition, several companies such as StudentU [20] and Study24 7 [21] have begun to offer students manually generated, on line class notes. Where media enhanced classrooms are available, students have expressed a preference for these classrooms [3] Finally, similar applications that capture ....

....issues are ignored, it may not be possible to generalize conclusions drawn about access patterns in the Classroom 2000 server to other media enhanced classroom systems. 5. 2 Study Sessions One way of reducing the impact of user interface design is to focus on student behavior in study sessions [13, 14] rather than counting the number of individual server accesses. A study session is a continuous period of interaction with the Classroom 2000 server. Unfortunately, since the HTTP protocol does not support connections, it is not possible to explicitly identify these continuous periods of ....

[Article contains additional citation context not shown here]

Jitendra Padhye and Jim Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. IEEE Internet Computing, April 1999.

....issues are ignored, it may not be possible to generalize conclusions drawn about access patterns in the Classroom 2000 server to other media enhanced classroom systems. 5. 2 Study Sessions One way of reducing the impact of user interface design is to focus on student behavior in study sessions [13, 14] rather than counting the number of individual server accesses. A study session is a continuous period of interaction with the Classroom 2000 server. Unfortunately, since the HTTP protocol does not support connections, it is not possible to explicitly identify these continuous periods of ....

....the Classroom 2000 server. Unfortunately, since the HTTP protocol does not support connections, it is not possible to explicitly identify these continuous periods of interaction. Instead, we estimate session length using a session termination threshold. We use the method described by Padhye et al. [13, 14] to identify sessions: a session is considered closed if a user has not accessed the system for a period longer than a specified threshold. This definition of a study session requires that we be able to identify particular users and their accesses. To do this, we restrict our analysis of study ....

[Article contains additional citation context not shown here]

Jitendra Padhye and Jim Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. Technical Report UM-CS-1997-056, University of Massachusetts, 1997.

....Classroom 2000 in Spring 1998. If a student accessing a continuous media stream did not explicitly terminate the access, it was difficult to determine a session end point. The MANIC system at the University of Massachusetts at Amherst encountered a similar difficulty in determining session length [4]; we followed their example and used a cutoff threshold of 30 minutes to determine session length. If a student has not made an explicit request for a period longer than the specified threshold, the session is considered to be ended. Any future accesses from the same network address are considered ....

....on such jumps is not currently captured, and thus is not represented here. Jumps were divided fairly evenly between forward and backward jumps. Average jump lengths were 7:5 minutes for forward jumps and 8:6 minutes for backward jumps. One interesting note is that the MANIC workload study [4] found relatively few backward jumps, while in our workload, we saw approximately 50 backward jumps. This discrepancy may be an artifact of the different user interfaces of MANIC and Classroom 2000. The Classroom 2000 server fetches all slide and penstroke images associated with a lecture the ....

[Article contains additional citation context not shown here]

Jitendra Padhye and Jim Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. Technical Report UM-CS-1997-056, University of Massachusetts, 1997.

....value. l app is used as input to their formula to estimate the throughput of a TCP connection under observation. When a non TCP flow is regulated using the formula, it is not possible to compute l app since the flow may use a different window size. Instead, Handley et al. 5] and Padhye et al.[4] estimate l act on the non TCP flow by dividing the number of loss events by the total number of packets transmitted. Loss events are registered as follows. The first packet loss is counted as a loss event. Following this, there is a back off for the duration of an RTT during which no packet loss ....

J. Padhye, J. Kurose, D. Towsley, and R. Koodli, "An Empirical Study of Client Interactions with a Continuous-Media Courseware Server," IEEE Internet Computing, April 1999.

....no way of indicating when a study session ended, we assumed that if there had been no activity from a student browser for over 30 minutes, the study session had ended. This threshold value was calculated using a technique described in the MANIC project at the University of Massachusetts at Amherst [19]. The top graph in Figure 8 shows a single student s access behavior for one course (an undergraduate course on software engineering taught by the author) The vertical axis represents lecture dates. This class met on Mondays, Wednesdays and Friday, and horizontal lines are used for clarity to ....

....2000, mainly because their systems were not engineered to facilitate the automated collection and post production of materials as was the case for Classroom 2000. Projects that are related to the educational application of Classroom 2000 include the MANIC system at the University of Massachusetts [19], the DEPEND system at the University of Oslo [20] the CHITRA project at Virginia Tech [5] and the Lecture Browser system in the Project Zeno effort at Cornell [21] There are also companies emerging to facilitate post production of recorded meetings and lectures, such as the company Eloquent, ....

Jitendra Padhye and Jim Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. Technical Report UM-CS-1997-056, University of Massachusetts, 1997.

....put multimedia display and capture equipment into classrooms. However, few of these efforts have attempted a workload model of the kind presented in this paper. Most closely related to our modeling efforts is the characterization of the MANIC project at the University of Massachusetts at Amherst [4]. The MANIC system is a classroom content delivery system built around the mixed media facilities from RealNetworks [3] The study focuses on a one semester senior level course, in which 200 users (identified by unique log in IDs) accessed the audio and text content associated with the course. ....

....is difficult in Classroom 2000: if a student accessing a continuous media stream does not explicitly terminate the access, it is difficult to determine a session end point. The MANIC system at the University of Massachusetts at Amherst encountered a similar difficulty in determining session length [4]; we followed their example and used a cutoff threshold to determine session length. If a student has not made an explicit request for a period longer than the specified threshold, the session is considered to be ended. Any future accesses from the same network address are considered to be part of ....

Jitendra Padhye and Jim Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. Technical Report UM-CS-1997-056, University of Massachusetts, 1997.

No context found.

J. Padhye, J. Kurose, "An Empirical Study of Client Interactions with a ContinuousMedia Courseware Server", Proc of Network and Operating System Support for Digital Audio and Video, 1998.

No context found.

Padhye, J., and Kurose, J., "An Empirical Study of Client Interactions with a Continuous-Media Courseware Server," Proceedings of NOSSDAV '98, Cambridge, UK, July 1998.

No context found.

J. Padhye, J. Kurose, "An Empirical Study of Client Interactions with a Continuous-Media Courseware Server",Proc of Network and Operating System Support for Digital Audio and Video, 1998.

.... multimedia WWWbased tutoring system that was originally developed in 1997 and was used by more than 200 users during the Spring 1997 semester to listen to, and view, the stored audio lectures and lecture notes for a full semester seniorlevel Networking course at the University of Massachusetts [9]. Previous work has been done to analyze and improve the performance of MANIC from both system s and users perspective [9] providing empirical and analytical characterizations of observed user behavior in MANIC. The work in [9] focused on studying the sessionlevel behavior (e.g, the length of ....

.... Spring 1997 semester to listen to, and view, the stored audio lectures and lecture notes for a full semester seniorlevel Networking course at the University of Massachusetts [9] Previous work has been done to analyze and improve the performance of MANIC from both system s and users perspective [9] providing empirical and analytical characterizations of observed user behavior in MANIC. The work in [9] focused on studying the sessionlevel behavior (e.g, the length of individual sessions) and interactive user behavior (e.g. the time between starting stopping pausing the audio within a ....

[Article contains additional citation context not shown here]

Jitendra Padhye and Jim Kurose. An empirical study of client interactions with a continuous-media courseware server. In NASSDAV, 1998.

No context found.

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous media courseware server. In Proc. of ACM NOSSDAV, July 1998.

No context found.

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous media courseware server. In Proc. of ACM Workshop on Network and Operating System Support for Digital Audio and Video (NOSSDAV), July 1998.

No context found.

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous media courseware server. In Proc. of ACM NOSSDAV, July 1998.

No context found.

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proceedings of International Workshop on Network and Operating System Support for Digital Audio and Video (NOSSDAV), June 1998.

No context found.

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proceedings of NOSSDAV, June 1998.

No context found.

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proceedings of NOSSDAV, June 1998.

No context found.

J. Padhye and J. Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. In Proc. NOSSDAV, Cambridge, UK, July 1998.

No context found.

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous media courseware server. In Proc. of ACM Workshop on Network and Operating System Support for Digital Audio and Video (NOSSDAV), July 1998.

No context found.

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous media courseware server. In Proc. of ACM Workshop on Network and Operating System Support for Digital Audio and Video (NOSSDAV), July 1998.

No context found.

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proceedings of NOSSDAV, June 1998.

No context found.

J. Padhye and J. Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. In Proc. NOSSDAV, Cambridge, UK, July 1998.

No context found.

Jitendra Padhye and Jim Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. In Proceedings of NOSSDAV, June 1998.

No context found.

J. Padhye and J. Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. In Proceedings of NOSSDAV, June 1998.

No context found.

J. Padhye and J. Kurose. An empirical study of client interactions with a continuous-media courseware server. In Proc. Inter. Workshop on Network and Operating System Support for Digital Audio and Video, 1998.

No context found.

Jitendra Padhye and Jim Kurose. An Empirical Study of Client Interactions with a Continuous-Media Courseware Server. IEEE Internet Computing, April, 1999.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC