cloud computing online video
TRANSCRIPT
Some History
• Studied High School & Computer Engineering Degree in Sydney.
• Started developing HTML in 96.
• My first job in the web industry in 2001.
• Started Rorcraft in Oct 06.
• Started Ankoder in Dec 07.
It can mean many things.
• Originate from network diagrams
• Marks a new era of computing
• Marketing Jargon (similar to web2.0)
Computing Eras
• Mainframe (centralised) - 60s-70s
• PC (decentralised) - 80s-90s
• Internet Cloud Computing (centralised) - today.
• No local software
• No local hardware, only access devices
• Everything through the internet cloud
Is it SaaS?
• SaaS - Software as a Service (Gmail, Google Docs)
• PaaS - Platform as a Service (APIs)
• HaaS - Hardware as a Service (Amazon)
Difference is vague
Is it Scale on Demand?
• Yes, but Amazon’s EC2 is not automatic.
• Google Apps Engine is automatic.
• Trade off between automation and flexibility
Traditional way of scaling
• Buy Hardware
• Buy routers, load balancers, firewall
• NAS Network Attached Storage
• Database Cluster or Master/Slave
• High Availability (IP redirect)
Virtualisation
• Xen - http://www.xen.org/
• open source industry standard for virtualization
• Dedicated Servers =>Virtual Private Server
• Enterprise implement their own virtualised infrastructure
Amazon’s Web Services
• S3 = Simple Storage System
• EC2 = Elastic Cloud Computing (Elastic = Scale up and down)
• SQS = Simple Queue System
• SimpleDB = Key-Value Database
Google Apps Engine
• Run time environment - currently Python only.
• Auto scaling (no preset load)
• BigTable - distributed database system.
BigTable
• Based on Google’s own distributed file system
• Data Redundancy (Fault tolerant)
• Served from memory.
• Open source implementation:Hadoop - distributed file systemHBase - database on top of Hadoop
Videos...
• Problem 1: Too many formats
• Containers:avi, asf, mp4, ogg, flv, mov, 3gp
• Codecs:mpeg-2, wmv, divx, h.264, theora, quicktime, sorenson
Videos...
• Problem 2: Too many sizes and devices
• Full HD, HD, HD Ready, Apple TV, web, laptop, netbook, ipod/iphone, portable media devices, mobiles
• Different bitrate, frame rate.
Most popular
• Flash: Flv (sorenson), MP4 (h.264)
• Mobile: 3GP ( MP2 or MP4)
• Firefox 3.1: Theora (open source)
Video Conversion...
• Problem 3: Transcoding is CPU intensive and takes alot of time
• What if you need to transcode 100 videos at the same time?
• If each video takes 5mins, that’s 500mins! (8hrs+)
Ankoder: Solving the problem
• Leverage computing power from Amazon on demand.
• Platform as a Service
• Reliability
• Scalability
• At no upfront cost.
Our target clients
• Videos are not part of their core business.
• Large user base, high volume video upload.
• Some examples:- online auctions- educational, tutorials- online dating- yellow pages (alibaba, openrice etc)
How we transcode.
• FFmpeg - open source tool
• Flvtool2, MP4Box, ffmpeg2theora
• Quality = bitrate : resolution : frame rate
Next steps
• Real time video transcode
• Analytics - tracking video usage
• Ad network - serving relevant ads to video
• Video Streaming - seek to begin