PutPlace's use of Amazon Cloud Services
Download as PPT, PDF1 like405 views
PutPlace provides an online backup service using Amazon Web Services including EC2, S3, and Elastic IPs. They have been using AWS in production since January 2008 and currently back up around 15,000 user files daily. Their monthly AWS costs are approximately $1,200 for EC2 and $470 for S3, which is doubling each month. To cut costs, they partition their database across smaller EC2 instances and use compression and deduplication to reduce storage needs.
PutPlace's use of Amazon Cloud Services
- 1. Secure ● Organise ● Share …your Digital Lifetime
- 2. PutPlace Online Backup Service Mac and Windows Clients Provide Sharing integration with Flickr/YouTube (Real Soon!) Original BP planned to spend 1m on data-centre in 2007
- 5. Key Stats Been using Amazon EC2/S3 in production since Jan ‘08 Public paid service launched in Oct ’08 Backing up around 15000 user files everyday Spend EC2 – USD $ 1200 a month S3 – USD $ 470 a month (S3 costs approximately double each month) Elastic IP – Use it but its really slow to respond No Simple DB (we use Postgres) No SQS (slow, non-determininistic, pricey) No EBS ( waiting and seeing how others perform) Keen on CDN!
- 6. Things We Use Perlbal – Load balancing Postgres – Database Django – App Servers and Task Managers Nagios – Internal Monitoring Monitis – External Monitoring Hamachi – Virtual Private Network (but no more free ) Hudson – Build Service Lots of Hosted Services Monitis DynDNS GoDaddy Google Docs Google Sites Google Email
- 7. How we cut costs Big chunk of EC2 cost is DB large instance Partition DB and run on smaller instances Storage is a variable cost compression and deduplication help We get hammered for puts, posts, lists Switch to MogileFS Dirty little secret of online backup – nobody downloads Bandwidth is a marginal cost Automation Can bring up (and down) a new grid in 10 minutes
- 8. The Future Gearman for Task Management CDN for static content Choice of Storage/Grid in Europe or US More use of regions for fault tolerance More automation around monitoring and deployment More stats Archive Service?
Editor's Notes
- #2: Welcome to our Investors briefing,