Bhupendra Singh SlideShare and AWS SlideShare and AWS
Bhupendra Singh
SlideShare and AWSSlideShare and AWS
Prologue: About UsPrologue: About Us- Founded in 2006- First and The Largest Presentation Sharing Site- Based in San Francisco / Delhi- 25 Employees- Flash based client- RoR Backend
Last Chapter: TrafficLast Chapter: TrafficApproximately 10k Slideshows are added daily, which requires- additional more than 100 GB storage per day- computation power to convert them into various formats
daily peak number of uploads per hour can be 25x daily low
Chapter 1: Slideshare Chapter 1: Slideshare and S3and S3- Since beginning (2007)- SlideShare stores almost everything on S3 i.e. ppt, pdf, swf etc.- Wildly Popular- Both secure and public access- lots of support (we use right_aws)- makes embeds available while SlideShare is having a downtime
Chapter 2: Conversion Chapter 2: Conversion SystemSystem- FreeBSD Boxes- Configuration process was painful- Unstable system due to fragile nature of software- Older boxes were more unstable- required restart of conversion process in every 2-3 days- Scalability issues
Chapter 3: Red Army Chapter 3: Red Army and 2008and 2008
Chapter 5: Beginning of Chapter 5: Beginning of EC2EC2- Conversion System on EC2- Configure, bundle and howzat!- kill unstable system and fire a new one- Precautionary Measure: Kill every system after 24 hours and start fresh ones- Scalability, Bazzinga!
Scalability and EC2Scalability and EC2- Regular Conversion Instances- Suicide Bombers- Gather, analyze and be proactive- Peak handling + Cost Control- Free EC2 to S3 transfer and vice-versa- Less bandwidth cost
Chapter 4: SQSChapter 4: SQS- After moving to new Hosting Service (2008)- Glue between new Host and previous Host (conversion system was still there)- Amazingly Simple- Reliable ?? Who cares??- Messaging Channel between new Host and EC2 conversion cluster- Inexpensive- We are using it HEAVILY!
Chapter 6: Spot Chapter 6: Spot InstancesInstances- Latest feature from Amazon- less than half price than regular instances- Suits us very well- We’re on it!
Elastic Computing Cloud
S3 S3
AWS
User
SlideShare Server Cluster
SQS
Thanks