A Dataset for Cloud Live Rate-Adaptive Video

There is an audience for non-professional video "broadcasters", like gamers, online courses teachers and witnesses of public events. To meet this demand, live streaming service providers such as ustream, livestream, twitch or dailymotion have to find a solution for the delivery of thousands of good quality live streams to millions of viewers who consume video on a wide range of devices (from smartphone to HDTV). Yet, in current live streaming services, the video is encoded on the computer of the broadcaster and streamed to the data-center of the service provider, which in most cases chooses to simply forward the video it get from the broadcaster. The problem is that many viewers cannot properly watch the streams due to mismatches between encoding video parameters (i.e. video rate and resolution) and features of viewers’ connections and devices (i.e. connection bandwidth and device display).

To address this issue, adaptive streaming working along with cloud computing could be the answer. Whereas adaptive streaming allows managing the diversity of end-viewers requirements by encoding several video representations at different rates and resolutions, cloud computing provides the CPU resources to live transcode all these alternate representations from the broadcaster-prepared raw video.

It is well known that the QoE of an end-viewer watching a stream depends on the encoded video and the parameters values used in the transcoding. But, in this new scenario in the cloud, we also need to consider the transcoding CPU requirements. In the “cloud video” era, the selection of video encoding parameters should take into account not only the client (for the QoE), but also the data-center (for the allocated CPU). To set the video transcoding parameters, the cloud video service provider should know the relations among transcoding parameters, CPU resources and end-viewers QoE, ideally for any kind of video encoded on the broadcaster side.

We would like to announce the publication of a dataset containing CPU and QoE measurements corresponding to an extensive battery of transcoding operations in http://dash.ipv6.enstb.fr/dataset/transcoding/ with the purpose of contributing to research in this topic. Most of the credits for this work (and so this post) have to be given to Ramon Aparicio-Pardo.

To elaborate the dataset, we have used four types of video content, four resolutions (from 224p up to 1080p) and bit rates values ranging from 100 kbps up to 3000 kbps. Initially, we have encoded each of the four video streams into 78 different combinations of rates and resolutions, emulating the encoding operations at the broadcaster side. Then, we transcode each of these broadcaster-prepared videos into all the representations with lower resolutions and bit rates values than the original one. The overall number of these operations, representing the cloud-transcoding, was 12168. For each one of these operations, we have measured the CPU cycles required to generate the transcoded representation and we have estimated the end-viewers’ satisfaction using the Peak Signal to Noise Ratio (PSNR) score). We depict a basic sketch of these operations for one specific case where the broadcaster encoded its raw video with 720p resolution at 2.25 Mbps and we transcode it into a 360p video at 1.6Mbps.

We give below an appetizer of how these CPU cycles and satisfaction decibels vary with transcoding parameters. They show some examples of the kind of results that you will find in the dataset, here a broadcaster-prepared video of type “movie,” 1080p resolution and encoded at 2750 kbps. If you wonder how the rest of figures look like, 558 curves and their corresponding 12168 measurements of cycles of hard CPU work and decibels of viewers’ satisfaction are waiting for you in http://dash.ipv6.enstb.fr/dataset/transcoding/

Toward a new public higher education system

My previous post was quite harsh about the way the French government addresses the MOOC phenomenon. I would like now to be more constructive (and also to demonstrate that I am not only a moaner). So, basically, what would I do if I was French ministry of Higher Education! In short:

• I'd shut down FUN. To be competitive, such project requires investment an order of magnitude greater than the planned fundings. When the objective is to attract tens thousands of students, there is no room for small players. 
• I'd stop fundings through call for proposals. These calls grant people who know how to write proposals and who, in the best case, release results years later. Moreover, and most importantly, these calls do not give the sense of responsibility to university managers. French higher education institutions have to learn how to promote their best professors and to make them "MOOC-able" instead of begging to the government as if "make a MOOC" was a right.
• I'd massively invest on French-friendly start-ups. The focus should be on three main domains where the position of France is today weak: an European-scale portal, tools for scalable learning, online student evaluation. The investment can be leaded by a structure such as BPI France.

In the following, I give my personal analysis of the context. I first decompose the traditional functions of a higher education institution, and analyze the challenges.

• Define the topic of the courses. In France, the institutions conceive curriculum, which are then checked by academic accrediting agencies like ABET in US, or CTI and AERES in France. Shortly put, the curriculums target young people (named students) and aim at developing their employability. Several courses form a consistent curriculum. As for the MOOCs, students are mainly workers, with a large diversity of motivations. The course is a unit, which should be independent. The topics are focused. It is thus quite different, but not fundamentally challenging.
• Select the students. This is the main asset of the Grandes Ecoles. However, MOOC are (expectedly) scalable, so you can teach an unlimited number of students. The question is no more to filter the best students before the course. The aim now is to have the right audience for the course: as many students as possible, with a high motivation for the topic and the right background. As said in my previous post, portals like Coursera are far better than any French higher education institution. 
• Build the course. Every MOOC creators agree that building a scalable online course is quite different from a traditional course for a small, on-site, population. MOOCs require new categories of workers. But the role of the teacher is still prominent. So far, the teachers have worked in traditional higher education institutions.
• Deliver the course. A building full of classrooms is useless. What is needed is a great, scalable, full-featured learning management tool. Moreover, you need a competitive team of developers to implement online exercices which have an added value and increase the student experience. Here, again, I don't think that any traditional french higher education institution can compete in providing such tool. Only a team of excellent super-committed software developers can do it.
• Assist students during their learning experience. The challenge of MOOCs is to provide the same kind of assistance as for a traditional course with one professor and a dozen of students, although the number of students is in the order of thousands. The power of community is the lever.
• Evaluate the students. When students are spread all over the world, it is impossible to organize exams the usual way. Companies like ProctorU have developed offers, where either exam rooms are available anywhere in the planet, or specific, secured, online tools allow anybody to be monitored as if she was on-site.

In the traditional model, all these functions are fulfilled by higher education institutions. In the new model related to MOOC, I foresee that traditional institutions will be outperformed by start-ups on a subset of functions: create a portal to attract students, develop a scalable learning platform, and evaluate students worldwide. These functions require strong skills in software development, in empowering a community of open-source developers, in promotional activities and marketing, in worldwide staff management, in agile development, in reliable online infrastructure, in website design. My claim is that neither universities nor public structures have any of the above skills.

Instead, I suggest to give a special mission to BPI France to make sure that funding goes to the most brilliant European start-ups related to education, in particular on the aforementioned functions (attract students worldwide, develop scalable learning platforms, evaluate students). By investing on European SMEs, the emergence of a champion is possible. And if the public force is one of the main investors, it may also ensure some of the "public missions" (e.g. almost free access to knowledge). Examples of such brilliant European start-ups include OpenClassroom, Iversity and FutureLearn.

On their side, the traditional French higher education institutions have to evolve. I like the analogy between MOOC and scientific books. Not all professors write books. Not all institutions ask their teaching staff to write books. Excellent professors (experts in some area, extremely brilliant as teachers) attract editors because the books they may write can become a success. It is thus up to the institutions to decide whether they should promote their excellent professors so that they may be detected by editors. Being "MOOC-able" is now a criteria for hiring professors in EPFL according to its director. This is the kind of shift French institutions have also to embrace.

The misconceptions behind the French FUN-MOOC portal

It is frequent that bloggers start their controversial posts with a disclaimer about how their personal opinion is not necessarily endorsed by their employers. In the case of this post, it is one step further: I am afraid that my opinion is the opposite of my employers' one.

I would like to talk about MOOC, you know, this innovation that may disrupt the higher education. My colleagues and I have been quite active in this area for the past couple of years, with a MOOC open on Spring 2013, and a contribution to two successful MOOCs.

One year ago, the French government decided to be pushy in this area, and thus to build from scratch a website named France Université Numérique (sorry no english translation for the wikipedia page yet), which aims at gathering MOOC in french from French higher education institutions in a free online portal. I summarize below some of the profound, symptomatic, critical misconceptions about what is innovation and Internet that this project demonstrates:

• This public (state-funded) project emerges although some private French start-ups (e.g. OpenClassrooms and Unow) were just kick-starting in the MOOC area. For the set of young entrepreneurs who were trying to gain reputations and to convince universities and Grandes Ecoles to join the MOOC movement, the arrival of such a competitor changed the game. FUN is a de facto incumbent since higher education entities are also funded by government. FUN is a public non-profit action, so it is completely free without need of any business model. These start-up have managed to find their place in a new ecosystem nonetheless, but, in my opinion, FUN-MOOC did not help. When it will be time to cry about the lack of "French Coursera" (like most cry today about the lack of French Google), we shall not forget that the government actually prevents the raise of such a possible French success story by entering the market like a bull in a china shop. 
• I am always sad when I realize that our political leaders still consider in 2014 that it is trivial to build a popular 24/7 Internet full-featured portal, and that it is trivial to manage a sophisticated professional online tool such as a massive, social, online course platform. It seems that the numerous failures of the public French websites have been quickly forgotten. FUN-MOOC was, and it is still now, a complete disaster. Typically, the portal has been shut down during two entire days in September 2014 for software upgrades. As can be expected from projects that are managed by people who know very few about the Internet and software, numerous shocking mistakes have been done, e.g. considering INRIA and a so-called SSII as a good team for the development of a software, and forking from the main open source online course platform project although the developer community was vivid and active.
• The branding does not look like an important matter for those who initiated this project. The acronym is FUN (I guess there is a couple of references when you type FUN on google). The full name is French oriented, which is good in France and in some places in Africa, but is bad when you consider any other part of the world. It is hard to know whether it is a consequence, but we have almost no Swiss, nor Quebec students registered in our courses in french. More generally, if one wants to build a popular website, the branding is key. It looks like the government did the same mistake as the creator of lescopainsdavant. When you compete versus Coursera, Udacity and EdX, a FUN name is not a gift.

I was very doubtful about this initiative, and I publicly claimed it. I told some friends that the government would try to shut down FUN-MOOC in less than two years when the MOOC bubble would go down and when our political leaders would realize that FUN-MOOC is expensive and not necessarily good for the economic sector as a whole. Well, I was wrong, it took them only nine months to realize. Unfortunately, we don't have our happy end yet. Indeed, the government asks whether some higher education entities would be happy to maintain FUN-MOOC on behalf of the government. The very sad point about it is that a consortium of various French entities (including Institut Mines-Telecom) is a candidate. Let me continue the misconception list:

• When it is time for innovation in general, a consortium of bureaucratic state-funded education entities is not the right vehicle. Exploring new business models, breaking the rules and embracing disruption are not in the DNA of public French universities, are they?
• A consortium of universities is no better than a government for managing a 24/7 full-featured online portal nor a sophisticated professional online tool. Universities struggle to have decent websites and learning software. I don't see any reason for a success in such a project, whatever the funding.
• The business model is, well, it is complicated, but with high probability it will be to complain that the government is not giving enough money for FUN-MOOC to work properly.
• The management is typical of the crazy French higher education system, with consortium of consortiums of entities that do not like each other, a probable series of rebranding operations just to be sure that everybody gets lost, weird processes where nobody really knows who is in charge of what, especially about critical points like the promotion of the portal and the development of new features, and, most of all, the promises of hours-long meetings.

It seems to me that all the main mistakes that a higher education ministry and a set of public universities can do are being done. Hopefully, some will eventually succeed in stoping this crazy bureaucratic counter-productive process. I failed.