GoSquared Presentation at AWS for Startups

Presented by Geoff Wagstaff (CTO of GoSquared) on 13th February to the TechHub folk at Google Campus, London.

Transcript

1. Real-Time Web Analytics
   On
   Amazon Web Services
   @TheDeveloper
   Google Campus
   February 2013
   

   View Slide

2. Real-time web analytics
   

   View Slide

3. James Gill
   CEO + Co-founder
   Geoff Wagstaff
   CTO + Co-founder
   James Taylor
   Developer + Co-founder
   

   View Slide

4. View Slide

5. LiveStats
   

   View Slide

6. View Slide

7. In 2009, our Server melted.
   

   View Slide

8. Conventional Hosting
   • Contractual lock-in. Fixed term subscription
   • Laborious and slow to provision resources
   • Restricted
   • Dedicated, but not bespoke
   • Boring
   

   View Slide

9. Business constraints
   • Experimental technology
   • Agility
   • Low budget
   • Inexperience
   • High barrier to entry in market
   

   View Slide

10. GET TO THE CLOUD!!!11!!one!
    

    View Slide

11. Why AWS?
    Diverse + growing range of services
    EC2
    S3
    CloudFront
    CloudFormation
    CloudWatch
    RDS
    Route53
    IAM
    CloudSearch
    ElasticBeanstalk
    ElastiCache
    DynamoDB
    SES
    SNS
    SQS Glacier
    RedShift ElasticTranscoder
    Data Pipeline
    StorageGateway
    VPC
    SWF
    

    View Slide

12. Why AWS?
    Flexible & Scalable
    5 x c1.xlarge
    2.5m ELB requests
    1.4TB S3 storage
    4 Route53 record sets
    32 EBS volumes
    500 DynamoDB Units
    3 SQS queues
    

    View Slide

13. Why AWS?
    Pay-as-you-use Pricing
    On-demand
    Spot (market) pricing
    Reserved
    Per unit
    Per 1,000 Requests
    Free Tier
    Per resource
    Heavy
    Medium
    Light
    

    View Slide

14. Why AWS?
    Platform agnostic
    APIs
    SQL server
    Windows Server
    Linux
    Ubuntu
    Fedora
    CentOS
    Ruby
    PHP
    Python Node.JS
    Perl CLI tools
    

    View Slide

15. Why AWS?
    Reputable
    

    View Slide

16. Challenges
    • More moving parts
    • Daunting. Requires new skills and role changes
    • Architecting for cloud
    • Deployment - resources, software, data
    • Automation - scaling, failover, backup
    

    View Slide

17. Advantages
    •Full control - freedom, bespoke
    •Fosters good system design paradigms & avoids complacency
    •Abundant tools, services, community + documentation
    •Lowered barriers to entry: infrastructure for anyone
    •Reduced time to market
    

    View Slide

18. Gap Year project: Trends
    

    View Slide

19. Angels (2010)
    Series A (2011)
    

    View Slide

20. 6 Team Members
    Average age: 21
    

    View Slide

21. 25,000 sites and counting
    

    View Slide

22. Traffic
    10s Thousands
    100s Thousands 100s Millions
    99.9%
    concurrent visitors pageviews/month
    Requests/second Uptime
    

    View Slide

23. Our Stack
    Load Balancer
    Reverse Proxy
    Services
    Data Stores
    ELB
    Redis, MongoDB,
    MySQL (RDS), EBS, S3
    Service-oriented architecture:
    Tiered, decoupled, load-balanced
    clusters
    Fault tolerance, failover, backups
    Auto-scaling. Spot instances
    Collectors, processing, API,
    Traffic
    AWS Services: EC2, S3, RDS, Route53, CloudWatch, SNS, IAM
    Applications, Website
    Route53
    

    View Slide

24. #ProTips
    Monitoring
    Ganglia - http://ganglia.sourceforge.net/
    Graphite - http://graphite.wikidot.com/
    ServerDensity
    CloudWatch
    PagerDuty
    https://github.com/etsy/statsd/
    

    View Slide

25. Deployment
    SCM + hosting
    https://github.com/gosquared
    Deliver
    https://github.com/gosquared/deliver
    Continuous Integration
    http://jenkins-ci.org/
    Installation + con guration - http://www.opscode.com/chef/
    #ProTips
    

    View Slide

26. Scaling
    EC2 Clusters. Auto-scaling. Spot instances
    Messaging - RabbitMQ
    S3, CloudFront, Route53
    Producer
    C C C
    Redis - http://redis.io
    https://github.com/gosquared/convoy
    #ProTips
    https://github.com/maritz/nohm
    

    View Slide

27. Destroy
    Everything
    

    View Slide

28. Thanks!
    Geoff Wagstaff
    @TheDeveloper
    https://engineering.gosquared.com
    

    View Slide