This document discusses using grid technology for distributed media processing tasks like video transcoding. It presents the MediaGrid concept of sharing heterogeneous storage and computational resources across organizations. Test results show distributing video transcoding across multiple servers can significantly reduce processing time. Simulation results indicate total job time is highly dependent on available WAN bandwidth when outsourcing to remote resource providers. The conclusions are that grid technology is viable for media production tasks by enabling parallelism, but technical limitations exist when using remote resources over insufficient network connections.
4. Originating problems
Tape-based media to file-based media
Multitude of file-based media transfers and processing
Storage / retrieval / transfer of media
Conforming
Transcoding
Upscaling
Editing
Geographically disperse facilities / resources / media storage
4
5. Grid technology as solution?
Grid technology
a Grid is a distributed processing architecture where heterogeneous
resources are shared between different participating organizations, across
an interconnecting network
Resources
Storage (media archive, temporary storage, etc.)
Computational (rendering farm, work stations, etc.)
Specialized (broadcasting, ingesting, etc.)
High speed interconnecting network (1-10 Gbit/s)
5
10. Grid technology proof-of-concept
Investigated the viability of Grid technology for processing tasks
in media production / distribution companies
Transcoding of media
Upscaling of media
Video transcoding deals with converting a video signal into
another one with different format, such as different bit rate,
frame rate, frame size, or even compression standard
Video transcoding is a resource intense process
I/O
Processing needs
10
11. Need for transcoded / rescaled video
VRT online media YouTube
http://www.deredactie.be http://www.youtube.com
11
12. Distributed video transcoding
How can we accelerate this process?
Server 4
Server 3
Server 1 Server 2
00:00:00 00:51:53
00:00:00 00:13:15 00:13:15 00:26:30 00:26:30 00:39:45 00:39:45 00:51:53
12
20. Setup overview
… TORQUE
… with GPFS cluster as media storage
… Java distributed transcoding front-end
… on each computational resource Transcode libraries
… the will to transcode in a distributed fashion
20
24. Discussion
Old version
Video files were physically split
Split / merge step could introduce artifacts
Current version
File is inspected and navigation file created allowing for easy frame-
addressing
Audio ripped and transcoded in separate step
No artifacts
Less media-transfers than in previous versions
Future version
Pre-fetching / replication of media to remote sites
24
27. Test results
Input media
Vob file
MPEG-2 video encoding
AC3 audio encoding
Size: 1,64 GB
Output media
Avi file
Xvid video encoding
MP3 audio encoding
Size: 700 MB
Currently no HD video input modules!
Not the most optimized video transcoders
Focus on measuring benefits of distributing
27
34. Video (up)scaling
Video scaling is converting video signals from one size or resolution to another: usually
quot;upscalingquot; or quot;upconvertingquot; a video signal from a low resolution (e.g. standard definition)
to one of higher resolution (e.g. high definition television).
00:00:00 00:51:53
00:00:00 00:51:53
720X576 984x752
34
42. Simulation results
Simulations provide very accurate total job turnaround times
Real-life transcoding behaves erroneously when
interconnecting GPFS with computational resource provider by
means of WAN link lower than 35Mbit/s
Control Traffic
Control Traffic
Click Router
Data
Data
Simulation results show what would happen to job turnaround
GPFS
time for lower WAN interconnections
42
46. Conclusions
Grid technology is a viable technology for dealing with media
production / distribution tasks
Inherent support for parallelism can seriously decrease the total
processing time
Need for adaptation of media tasks
Grid overhead is no issue
Outsourcing task processing to remote resource providers
Viable when interconnection is sufficient
Technical limitations (e.g. GPFS time-outs)
MediaGrid simulator can provide accurate performance
predictions
46
47. Questions ?
Feel free to e-mail: Bruno.Volckaert@intec.UGent.be