This year as last year, the Media Performance Group attended the ACM MMSys conference. This year the conference was hosted at University of North Carolina at Chapel Hill. This year we had one dataset paper accepted (Network Traffic from Anarchy Online: Analysis, Statistics and Applications) and one full-paper accepted to the MoVid 2012 workshop (A Comparison of Quality Scheduling in Commercial Adaptive HTTP Streaming Solutions on a 3G Network) which was co-located with ACM MMSys. All these papers should now be available in the ACM Digital Library and our web-pages.

Breakfast meeting at SAS Lounge

The four of us participating at the conference started the trip quite differently. Tomas, Carsten and me (Håkon) left Monday the 20th of February. Tomas flew from Gardermoen to New York and down to Raleigh/Durham, While me and Carsten flew Oslo via Copenhagen to Washington D.C Dulles (Airport code IAD ;) ) and finally to Raleigh/Durham.

So the three of us all started the day at the SAS Scandinavian Lounge at Gardermoen with our traditional “Monday Morning Breakfast Meeting”. Andreas flew out of Oslo on Tuesday the 21st of February. While we waited for Andreas to arrived we had the time to look around at the UNC campus. The fist thing we experienced was that it was not easy to get a paring space, and to get any parking at campus we needed one of many permits. After we found a parking space off campus and had a lunch, we visited Ketan Mayer-Patel, a professor at UNC, and the local arrangement chair of this years MMSys.

Wednesday 22nd of February was the first day of the conference. After a brief introduction from the organizing committee, we started with the fist session which was about quality for HTTP streaming. The session was followed by the first keynote, by Deepak S. Turaga from IBM, titled Automating Large-Scale Multimedia Exploration for Smarter Planet Applications. The keynote was very interesting, and relevant for our groups work in distributed processing. Later in the first day there were a session on signal processing and one on emerging topics before Andreas presented his paper in the datasets session. The first day ended with an open ended demo and posters session, where all the authors at MMSys were invited to bring posters of their work, and it was possible to walk around and have discussions.

Presentation of MMSys 2013 in Oslo

Presentation of MMSys 2013 in Oslo

On the second day (Thursday 23rd) of the MMSys conference we started with a session on supporting 3D content, before we had the second keynote of the conference. This keynote was held by Leonidas Kontothanassis fron Google on the topic Content Delivery Considerations for Different Types of Internet Video. One of the things we learned about in this keynote was Googles challenges with YouTube. Later in the day we had sessions on adapting to mobility, and adapting to the infrastructure. Before the end of the day me and Carsten had a short presentation about next years MMSys, which will be held in Oslo, with MoVid and NOSSDAV as co-located workshops. After the second day finished we all went over to Ackland Art Museum at campus for a tour of one of their exhibitions called “The Spectacular of Vernacular”. After a nice tour of the museum we had a reception at one of the bars located right outside campus called “The Back Bar”, we all had a great evening (and night).

On Friday (the last day of the conference) it was time for MoVid, one of the co-located workshops to MMSys. MoVid started with a session on quality assessment and application, before we had a keynote by Jim van Welzen from Nvidia titled Challenges Leveraging the Burgeoning Power of Mobile Multimedia Architectures for Innovative Commercial Multimedia Applications. Here he talked about some of the challenges you have when developing applications for mobile platforms when the APIs constantly are evolving. After the keynote, we had a panel discussion, and the last of the workshop called multimedia transport. In this session Carsten presented Haakon Riiser’s paper (he could unfortunately not come). MMSys 2012 was a very interesting confernece, we had many interesting presentations and discussions, and we all look forward to next year when we will be in Oslo.

We all left Chapel Hill on Saturday. Andreas headed back to Norway, and me, Carsten and Tomas headed up to New york, where we spent the weekend before Carsten and me traveled up to Amherst, Massachusetts were we visited Michael Zink at University of Massachusetts (UMass).

The underlying network statistics of a virtual worldThe research community needs data to analyse and understand how different networked applications behave. In cooperation with iAD-partner Funcom, we have made a server-side trace from the MMORPG “Anarchy Online” available to the public. The dataset paper is presented at the MMSYS 2012 conference that is held in Chapel Hill, North Carolina this week.

The network traffic from an MMORPG is driven by events; the players move, attack, trade and perform other actions. Each time an action is perfomed, the client dispatches a data packet to the server. The server processes the actions and distributes the necessary information to other affected players. When playing a game, you don’t want to wait for your actions to register. Lag is despised in games. Therefore, the data is dispatched onto the network as soon as it is generated. An effect of event-driven applications with time-sensitivity is that the data-streams have small payloads and the time between packets is relatively high. This dataset is a core example of such network traffic. It is, as far as we know, the first server-side trace from a commercial MMO that has been made available to the public.

From the dataset, we do a network analysis of the key properties of a server-side trace from “Anarchy Online” from Funcom. You can extract many interesting properties like payload sizes, packet rates, data delivery latencies, retransmission statistics, loss rates and stream correlation.

Examples of applications of the dataset may be to replay the data to generate realistic traffic for testing, create loss/delay parameters for network emulation, analyse to tweak protocol development for interactive applications and probably much more.

Now that the dataset is available, we hope interested parties, including academia and industry, will use it to push research forward in the field of system support for interactive applications.

For more statistics and analysis of the dataset, you can download the paper here.

The February 2012 breakfast meeting at the iAD Lab was about our view and experience around social and collaborative training- and learning environments – and how technology plays a vital part in this (r)evolution. About 30 persons showed up at the iAD Lab @ IFI on a early Wednesday morning to participate in this meeting!
Participants at the Meeting

Participants on the meeting

Lars Sverre Gjølme from Accenture shared his experience based on work done with various clients, highlighting some of the issues we believe are critical for the better,smarter and faster eTraining. Eilin Schjetne – head of Business implementation in the Business Development unit of DNB did a presentation on how they recently developed a brand new digital training platform. The system was also demoed for the audience.

Professor Pål Halvorsen from the University of Oslo commented on how search technology can play a vital part when “learning objects” becomes more multimedia oriented, and also gave the participants a brief introduction of iAD’s vision of the next generation platform for search and collaboration. Slides from the presentations will be available for the iAD webpage.



On Friday 27 January, Kristian R. Evensen defended his PhD thesis Aggregating the Bandwidth of Multiple Network Interfaces to Increase the Performance of Networked Applications.

Kristian Evensen

Devices capable of connecting to two or more different networks simultaneously, known as host multihoming, are becoming increasingly common. For example, most laptops are equipped with a least a Local Area Network (LAN) and Wireless LAN (WLAN) interface, and smartphones can connect to both WLANs and HSDPA-networks. Being connected to multiple networks simultaneously allows for desirable features like bandwidth aggregation and redundancy.

Enabling and making efficient use of multiple links requires solving several challenges related to deployment, link heterogeneity and dynamic link behavior. Even though multihoming has existed for a long time, for example routers require being able to connect to different networks, most existing operating systems, network protocols and applications do not take host multihoming into consideration. The default behavior is still to use a single link for all traffic. Using a single link is, for example, often insufficient to meet the requirements of popular, bandwidth intensive services like video streaming.

In this thesis, we have focused on bandwidth aggregation on host multihomed devices. Bandwidth aggregation is to merge physical links into one logical link that offers a higher bandwidth. Even though bandwidth aggregation has been a research field for several years, the related work has failed to consider the challenges present in real world networks properly, or does not apply to scenarios where a device is connected to different networks.

In order to solve the deployment challenges and enable the use of multiple interfaces in a way that works in a real-world network environment, we have created a platform-independent framework, MULTI. MULTI works in the presence of NAT, automatically detects and configures the device based on changes in link state, and notifies the application(s) of any changes.

MULTI was used as the foundation for designing transparent and application-specific bandwidth aggregation techniques. A transparent bandwidth aggregation technique is a technique that requires no modifications to the application, nor the operating system at either the sender or receiver. An application-specific technique, on the other hand, is an extension to an existing application or application type.

The application-specific bandwidth aggregation technique presented in this thesis, improve the performance of quality adaptive video streaming. The technique was evaluated with different types of streaming in both a controlled network environment and real-world networks. Adding a second link gave a significant increase in both video and playback quality. The technique is not limited to video streaming or the system/protocols we have worked with. As long as two common requirements are met, the technique can be applied.

In many cases, it is not possible to extended the application directly with multilink support. Working on the network-layer allows for the creation of transparent bandwidth aggregation solutions. A transparent, network-layer bandwidth aggregation solution must cater to the behavior of the different transport protocol in order to achieve efficient bandwidth aggregation. The transparent bandwidth aggregation techniques introduced in this thesis are targeted at Universal Datagram Protocol (UDP) and Transmission Control Protocol (TCP), the two most popular transport protocols in the internet today.

In December 2011 the World Opera in collaboration with University Tromsø, Kulturhuset Tromsø and the Verdione project produced their first 2-site distributed opera performance ‘La Serva Padrona’. The script was based on an opera from the 17th century written by Giovanni Battista Pergolesi. The production explored the space between the spaces, an artform that links multiple performance spaces in realtime through the net.

Art.on.Wires has streamed all three performances live to the Internet. Watch a short time-lapse clip of our preparations and rehearsals below:

Credits

Harald Bakkeby Moe – Baritone
Berit Nordbakken Solset – Soprano
Jenny Svensson – Dancer
Bjørn Andor Drage – Keyboard
Yuko Kawami – Violin
Bernt Simen Lund – Cello
Nasra Ali Omar – Drums
Niels W. Lund, General Director
Bjørn Andor Drage, Musical Director
Jason E. Geistweidt, Technical Director
Rakel Nystabakk, Production Manager

This is a short 9min documentary we made last year about the second Art.on.Wires Media Arts Festival. You can learn more about what happened at the festival here.

The SPIE Electronic Imaging Conference 2012 has accepted a paper written in collaboration between our group and University College Gjøvik. The paper describes part of our work on color calibration for multi-camera arrays. We measured the spectral responses of two otherwise identical Gigabit Industry vision cameras containing a single-chip CCD sensor over a large spectrum of illuminations and exposure times. What we found is that even cameras of the same model and from the same production line exhibit variances in color responses that are significant for multi-view video reconstruction. Based on this observation and our measurements we developed a model to characterise the response of a camera. Assuming the characteristics of all cameras are known, we are able to color-calibrate all images across the entire array.

Abstract—The advance and rapid development of electronic imaging technology has lead the way to production of imaging sensors capable of acquiring good quality digital images with a high resolution. At the same time the cost and size of imaging devices have reduced. This has incited an increasing research interest for techniques that use images obtained by multiple camera arrays. Use of multi-camera arrays is attractive because it allows capturing of multi-view images of dynamic scenes, enabling the creation of novel computer vision and computer graphics applications, as well as next generation video and television systems.

There are additional challenges when using a multi-camera array, however. Due to inconsistencies in the fabrication process of imaging sensors and filters, multi-camera arrays exhibit inter-camera color response variations. For the majority of applications, which use multi-view images obtained from multi-camera arrays, it is insufficient to assume that the different camera’s response can be considered the same without prior verification. Therefore, it is necessary to characterize the response of the different cameras in the array.

SIGMM Records (3):4


The December issue of the SIGMM Records is out. They can be found here. We can announce the first call for nominations for ACM SIGMM’s new award, the Nicolas Georganas TOMCCAP Award, which will be awarded for the best paper that was published in ACM TOMCCAP in 2011. You can also find the call for another award, the ACM History Fellowship award.

We are pleased to introduce to you a new columnist for the Records. Pablo Cesar will from this issue on report about an Open Source project in every issue. This issue talks about LIRe, an image retrieval library.

Christian Timmerer reports from the 98th MPEG meeting, which concluded some important work and is preparing for new challenges.

The education column presents a book on image and video processing that you may find interesting in your teaching. You can also find a report from Ed-Media, a conference that discusses the use of multimedia for education.

You can also read the PhD thesis summaries provided by three candidates who have recently passed their doctoral exams, you find pointers to the latest issues of TOMCCAP and MMSJ, and several announcements from ACM and SIGMM.

On Wednesday 11 January, Dominik Kaspar defended his PhD thesis Multipath Aggregation of Heterogeneous Access Networks.

Dominik Kaspar

In his thesis, Dominik Kaspar explores different ways to increase the speed and reliability of Internet connections by transmitting data over several networks simultaneously. One of the proposed solutions, which can be used by Web browsers and other HTTP-based applications, resulted in a U.S. patent. This method uses logical file segmentation and request pipelining to split a file over numerous Internet paths and recombine it at the destination. Experiments with live video streaming showed that this approach is highly efficient and can be deployed in a lightweight manner.

Furthermore, the thesis investigates the negative effects of data reordering. When data packets are sent over several Internet connections that differ in speed, the packets are likely to arrive out of order. Especially for the TCP communication protocol, which transports the majority of Internet traffic, packet reordering is known to be very destructive. To reduce packet reordering, a scheduler was designed that keeps track of the connection speeds and smartly sends packets so that they always arrive in order at the destination. In addition, a large set of experiments was carried out in which TCP was exposed to various degrees of packet reordering. Thereby, several parameters were identified, which, when correctly tuned, lead to improved TCP performance over multiple networks.

While related research often introduces complexity to third-party servers or to long-established network protocols, the proposed solutions target an easy deployment. Implemented and tested in real-world networks, the thesis shows the benefit of simultaneously using multiple Internet connections to achieve a higher data throughput.

Two articles that were published on the Norwegian website forskning.no provides some coverage of MPG’s projects Romus and Verdione.

The article “Nettvideo hakket videre” talks a lot about results on P2P streaming with network coding, but a lot of the information contained in the articles is based on Pengpeng’s work on video quality assessment. Including a snapshot of the famous squirrels from Big Buck Bunny in two qualities.

The other article, “Video-overført verdensopera”, is about Verdione. It is based on Norunn Torheim’s telephone interviews with Niels Lund and Carsten, and provides a summary of the challenges that Verdione faced in actually bringing our technology to the artists.

© 2011 Media Performance Group Blog Suffusion theme by Sayontan Sinha