To enhance the DCP transmission efficiency, Ali engineers actually do it!

Author | Ali entertainment technology

Zebian | Hu Weiwei

Exhibition | CSDN (ID: CSDNnews)

background

DCP spelling is Digital Cinema Package, Chinese digital cinema package for digital image storage and conversion of audio, video and data streams, is cinema projection equipment used in the media package. An ordinary 2D movie DCP size generally between 40G ~ 60G, a general 3D movie times multiplied by 2, or if it is 4k IMAX film, the size of the DCP is normal to reach more than 200G.

Secret oversized paper over 200G Digital Cinema Package by TMS how efficient transmission to each auditorium. Now copy the DCP to each theater auditorium mainly through the use of TMS (theater projection management system) video transmission function

Energy transmitted from TMS responsible for the DCP to each auditorium, but the efficiency of such transfer is not high, TMS data source only, so that each of the auditorium to be copied video TM pulling on S, the bandwidth will become a bottleneck.

TMS models currently used to transport the DCP playback server

P2P transmission mode using the class

As can be seen from Figure 1, after the cinema hall playing server pull DCP, their bandwidth is in an idle state, then we can use similar P2P (peer to peer network transmission) transmission solution to, so that you can use existing facilities improve the distribution efficiency within the DCP theater, can lower the effect of the present extract efficiency.

P2P schematic

According to the above P2P manner, the transformation of the theater within the transmission DCP models:

TMS improved playback mode to transfer the DCP server

FIG mainly describes the transmission coordinator coordinates the respective transmission source Looking Office Movies pull DCP process. Features of this refinement is to increase the transmission of a coordinator state machine as a shared, coordinated path DCP copies each auditorium, the transmission path is calculated optimal solution. Transport coordinator core functions are:

1) were collected hall playing server network conditions;

2) where each flag in a storage hall DCP playback server;

3) According to the network, the distributed computing and transmitting server tasks to play and offices. And the use of the transmission of the program, the transmission efficiency will be greatly improved. For example there is a total file size of DCP

Is 400GB, a total of 10 video hall, bandwidth 1000Mbps≈125MB / s, then the TMS using conventional transmission mode, at least

To 400102410 ÷ 125 ÷ 60 ÷ 60 = 9.10 h, almost a day's time. If using a new way, you need only three hours, you can increase three times, and with the increase auditorium, efficiency gains exponentially.

A further development

From Figure 3 that, although the above scheme to solve the problem of wasted bandwidth, but architecturally there are some problems: transport coordinator is a single point, it hung up, transmission on the problem; data can have only one data source source hung up, transport stopped, and re-transmission to start from scratch. This time we have to how to solve?

These two issues can use the transport coordinator decentralized deployment and sub-file to download to solve fragment resume point:

1. Transport coordinator decentralized deployment

We can transport coordinator deployed to various video hall, each transmission coordinator is a virtually stateless nodes, without any information synchronization between nodes, the DCP transmission whenever a cinema hall completed, broadcast to all transport coordinator in. But sending a copy of the instruction the way you need to transform the look, unified by the TMS sends instructions to download the video transmission coordinator of each auditorium, and then transport coordinator is responsible for asking whether neighboring nodes have downloadable DCP, exists to download, no longer TMS transport needs coordinator designated download address for the auditorium. New network topology is as follows:

After further improve network topology of FIG.

2. Files divided fragments point resume download DCP

On a program, the server copies a DCP or native instructions playback operation, a large restriction, does not support multiple data sources and documents slicing. Since the playback server does not support, then we need to develop their own an intermediary role, the need to support multiple data sources and file fragmentation operation, and have a copy of DCP to the ability to play server's hard drive, and deployment just can bear the transport coordinator auditorium this responsibility. Source files and multiple data slicing scheme diagram is as follows:

Scheme schematic sheet document of multiple data sources and

The above-described two technical solutions can be combined. Summary: Although this approach does not improve much speed, but has improved in terms of system fault tolerance, the user experience better.

to sum up

It can be seen from the above section, by fusing we P2P, FTP, HTTP, file fragmentation techniques, resulting in a dedicated LAN transmission DCP technical solution. We draw on ideas P2P, to achieve the sources auditorium shared within the local area network effect, to overcome the shortcomings of traditional TMS transport DCP single data source; use FTP as a means of transmission, the transmission mode is compatible with the existing auditorium; use breakpoints resume, a file fragment that enhance fault tolerance of the system.

This program is actually very typical combination of innovative methods, technologies are already used and is very classic, but by re-sort them integrate, making it a qualitative change in performance, in order to produce new value. This article is a program born using this method, the design of this program in the process, I also learned some fur combination of innovative methods, we still continue to work hard to learn this method.

【End】