Tokyo Video Tech #2 Report (en)

Published in

Tokyo Video Tech

8 min readFeb 25, 2019

Reported by Takesato Hayashi
Photograph by Katsuyuki Sakai, Takashi Okawa

On 20th February 2019, 2nd meetup event Tokyo Video Tech #2 UNTITLED (UNTITLED covering various topics related to online video technologies) sponsored by J-Stream was held. This time we had waiting list and we adjusted conference room in order to 52 attendees to be gathered. There were three speakers, and one special guest from Taiwan. They provided presentations and a lot of Q&A sessions. After the sessions, everyone enjoying “Networking with Food & Beverage” time together, it was very blazing time for online video engineers.

Event page：https://www.meetup.com/Tokyo-Video-Tech/events/258662769/

# Video Tech Alliance — Opening remark

Organizer Takesato Hayashi(Takesato) taking Opening remark part.

Last time, Takesato introduced magic number 11, the total number of Video Technology Meetups around the globe. Now it was changed to 15 which includes newly formed Norway Video Technology. In the same time, thanks to the organizer of Taipei Video Technology Shuen-Huei (Drake) Guan, he worked with Demuxed Team, Taipei and Tokyo are finally listed on Meetups around the world page<https://demuxed.com/meetups/>.

After the announcement of #1 Unite the world, Takesato and Drake started communication in Slack. When Drake came to Tokyo in this January and heard about upcoming #2, he said “I really want to participate”. Then he became a special guest speaker tonight.

Then Takesato announced Video Tech Alliance with Tokyo and Taipei. We believe that we can make use of geographical closeness to exchange ideas and speakers.

And finally, he revealed upcoming Tokyo Video Tech #3 will be held at Fastly Japan on Wednesday, 24th April, 2019. Details will be posted at Meetup.com and the Slack workspace <https://bit.ly/video-dev-jp>, so please look forward to it.

# How do we start Taipei Video Technology

First speaker Shuen-Huei (Drake) Guan, Taipei Video Technology

He talked in English. Like other Video Tech meetups, Tokyo Video Tech welcomes English sessions by overseas guests.

It has been 3 years since he started thinking about making Streaming Meetup(formerly known as). During AWS Summit Taipei 2016, several video experts gathered and talked casually. He thought “let’s have our meetups”. But it won’t happen until 2018. The first meetup was held on March, 2018 at KKStream. The 2nd meetup was held at 泛科學院 on August, 2018.

Then he explained overview of their meetup topics as follows. We realized that interests and the challenges we face are very common in every region.

1. Streaming as a linear channel (VOD2Linear)
2. Streaming Media West report
3. AV1 The next generation video codec
4. Next Generation Encoding Solution (Multi-cloud, multi-codec, per-title-encoding)

They use Meetup and Slack for member communication. If you’re interested in Taipei’s video streaming circumstances or want to join the meetups, you’re welcomed to join by the following URLs. The next meetup will be on March 13rd, 2019.

Meetup: https://www.meetup.com/Taipei-Video-Technology/
Slack: https://bit.ly/video-dev-tw

# Leverage multi CDNs in video streaming

Next speaker Go Kobayashi(Go), IoTv Center, TV Asahi Corporation.

In the session “Leverage multi CDNs in video streaming”, he explained the background/current status of Multi CDN, and future plans.

Go has one big theme “Stabilize viewing experience for all users”, and the CDN takes an important role for that. If you use single CDN, it has been found that the throughput of the CDN sometimes becomes unstable depending on the location of users or ISPs. In the same time, from his IoTv Center (considering with simul broadcasting) point of view, they’re anxious for stability of Live Streaming when it’s delivered by single CDN. By leveraging Multi CDNs, he thought there would be following merits, and he started using Multi CDNs.
1. Throughput is stabilized by redundant CDN.
2. Easy to add and delete CDNs.
3. Become easy to negotiate delivery costs with CDN service provider.

He started using Multi CDN for the production services and he saw the difference between each regions, ISPs and the cache hit ratio by CDNs. Unfortunately, the slides and detailed information was disclosed only in venue. So, I skip further details of his presentation. Very useful real stats information was disclosed for engineers who experienced the problems related to CDN.

After using Multi CDN, the services gained following merits.
1. High performance CDN can be selected by region and/or ISP.
2. Prevent network outage in advance.
3. Easy to add and delete CDNs.

For the future plan, he will use 3 to 4 CDNs in total, and will use Multi CDNs in Live Streaming actively.

# How we’ve solved the OVP codebase bloat

Next speaker Kentaro Nishimuta(Kentaro), Leader, Engineering and Laboratory, CDN & Video Platform Business Unit, J-Stream Inc.

The session titled “How we’ve solved the OVP codebase bloat”. As an engineer’s stand point of view, he explained what he experienced, struggled and want to change during Online Video Platform(OVP) development and management.

Kentaro express the term “Bloat” is different from “Expanded” in which various things expanding as a service grows. He said “Bloat” is, the system configuration becomes a mess and getting difficult to grasp the whole picture.

Common examples are,
1. Documents are not up-to-date.
2. Fabulous gurus and person dependency (afraid of the key person will quit someday…).
3. Manual deployment.
4. Lack of extensibility.
5. Uncertainty of scale of refurbishment and scope of impact.
6. Mysterious function (should not be used by anyone).
7. No test code.
8. No Staging Environment.

So, Kentaro said that he is working on the following items in order to prevent “Bloating” in the OVP renewal project.
1. Updating documents. (GitLab for ticket management. Using Markdown + mermaid. No more Excel and PowerPoint world.)
2. Scalable. (Loose coupling between system components. Creates micro services as a single function.)
3. Deployment automation.
4.Test automation (Deployment testing environment on the cloud).

Kentaro said, “It is inevitable to get “Bloating”, but as you know it is difficult to manage the system manually, we should be kept away from people’s hands by utilizing automation.” Although this is not limited to OVP, it seems that there are many engineers who have experienced similar issues in the online video world where standard technology continues to be born and a short release cycle is always required. I am impressed, there were many participants listening to his story while nodding at the venue.

# The challenge of video distribution. Advantages and disadvantages of CMAF

Last speakers of today is Masashi Ito (Maitou), team leader, research and development, technical department, Fuji Television Network, Inc.

The session titled “The challenge of video distribution. Advantages and disadvantages of CMAF”. He explained CMAF, standardized media file format, with lots of slides and demos.

Maitou’s problem is that the media distribution method differs between TV broadcasting and IP communication. TV broadcasting using “Hybridcast” to provide media data with MPEG-DASH(ISOBMFF fMP4). IP communication (PC browser and smartphone) using HLS(MPEG-2 TS). To adapt both platforms require some modification to media container and manifest file itself. In order to minimize such difference, he tried to use Common Media Application Format (CMAF). It was standardized media file format published on January 2018 as ISO/IEC 23000–19, uses ISOBMFF fMP4 for container.

System requirements for CMAF support are,

Players

iOS 10 or later.
PC web browser with Media Source Extensions (MSE) support and hls.js and libraries.

Encoding / Packaging

[MPEG-DASH]

Ffmpeg + MP4Box, AWS MediaConvert and etc.

[HLS]

AWS MediaPackage, AWS MediaConvert and etc.

Maitou demonstrated CMAF video playback on TV/PC/iPhone/mac. The video is converted by AWS MediaConvert.

“This is my personal impression” he said. Then he listed advantages and disadvantages of using CMAF.

Advantages

improvement of the CDN cache hit ratio by reducing the cost of one media file.
reduction of the storage cost.

Disadvantages

device dependencies of HLS native player.
tons of pre MPEG-2 TS encoded media asset needs to be re-packaged.
Unified standard of AD insertion method can not be seen.

Finally, he explains possibilities of reducing media distribution latency of Live Streaming by using CMAF chunk of CMAF media object model (compatible with MPEG-DASH fMP4). As an example actually used, he played Akamai’s demonstration content of the live stream.

TEST PLAYER FOR CHUNK-ENCODED CHUNK-TRANSFERRED DASH by Akamai
<http://mediapm.edgesuite.net/will/dash/lowlatency/low-latency-public-example.html>

Finally,

Tokyo Video Tech #2 UNTITLED last session, The Networking with Food & Beverage. All speakers and attendees were really enjoyed delicious foods and drinks together. Talking about today’s topics and media technologies. This time we had first overseas guest speaker and the member from Streaming Conference. I felt strongly that new network expands among speakers and attendees as they work in the same online video industry.

I would like to take this opportunity to thank the speakers first. Awesome team of J-Stream, preparations from the planning, setting of the venue and installation of Food & Beverage.
Thank you very much.