Software Engineering in Startups
Download as PPTX, PDF1 like1,039 views
The document discusses essential strategies for software engineering in startups, emphasizing the importance of building the right product and efficient collaboration tools. Key points include prioritizing product discovery, continuous deployment, measuring performance, and maintaining a strong focus on hiring and company culture. It highlights that the core business of startups is finding scalable business models rather than solely concentrating on technology.
Software Engineering in Startups
- 1. Software Engineering in Startups @dusano TSstartup, 2012-10-23
- 2. mag. Dušan Omerčevid, UDI RI VP of Engineering @zemanta @dusano dusan.omercevic@zemanta.com restreaming.wordpress.com linkedin.com/in/dusanomercevic github.com/dusano skram.si
- 3. Outline • 15:10 – 16:00 Engineering in Startups • 16:10 – 17:00 Tools • 17:10 – 18:00 Technology • 18:10 – 19:00 Growing Systems • 19:00 - Beer(hunt)
- 5. Build the right product vs. build the product right?
- 7. Google's Way Source: http://www.quietaction.com/blog/inspired-d3-talk-tom-chi.html
- 9. Great Companies build Great Products
- 10. Courage Respect Communication Commitment Focus
- 11. Yishan's Way Yishan Wong, CEO of Reddit, Director of Engineering at Facebook, Senior Engineering Manager at PayPal (http://algeri-wong.com/yishan/engineering-management.html) 1. Make hiring your number one priority, always! 2. Let process be implemented by those who practice it 3. Promotion from within 4. Tools are top priority 5. Technical Leaders
- 12. PM 101 1. Product Owner & Product Backlog 2. Retrospectives 3. Team & Scrum master 4. Daily stand-up meeting & Demos 1. Iterations (Scrum & XP) or limited work-in- progress (Kanban)
- 13. Tools
- 14. Collaboration Tools • In person meetings & Whiteboard
- 15. Collaboration Tools #2 • Trello • Google Apps
- 16. Collaboration Tools #3 • IRC • Blog!
- 17. Collaboration Tools #4 • Code style guidelines • Git • Code reviews – see Code reviews at Khan for guidelines
- 18. Continuous Deployment • Make Deployment a non-event • If something is painfull, do it more often until the pain goes away! • Ideal: the new employee should be able to deploy on the first day! • Prerequisite: Unittests & TDD • Dark releases, Blue-green deployment, Canary releasing
- 19. Technology
- 20. VC warning! The core business of startups is searching for a repeatable and scalable business model and not playing with (unreliable) technology!
- 21. Operation System There's only one!
- 22. Programming Languages • JavaScript • Java • PHP • Python • C/C++ http://en.wikipedia.org/wiki/Programming_languages_used_in_most_popular_websites
- 23. Measure Anything, Measure Everything http://codeascraft.etsy.com/2011/02/15/measure-anything-measure-everything/
- 24. Measuring as Testing • Unittests are only run upon deploy, not periodically • You should have HealthChecks for all your applications, that are run upon deploy and then periodically • Have error inducing functionality to test your error detection functionality
- 25. Storage & Data Processing • PostgreSQL vs. MySQL • MySQL vs. Cassandra – The Immutable state of MySQL • Hadoop/Hive • Splunk • Lucene/Solr • RabbitMQ
- 26. Storage Latencies • Memcached: <1 ms (mainly network latency) • Cassandra: ~1 ms • Oracle for simple queries: 1-5 ms • DynamoDB: ~5 ms • SimpleDB: >10 ms
- 27. Cloud Computing AWS: Hetzner EC2 Large instance (4 EC2 Quad-Core i7-2600, 16 GB compute units/virtual cores, RAM, 2 x 3 TB disk: 7.5 GB RAM memory, 850 GB • 50 euro/month storage) • On-demand: 250 euro/month • Reserved: 600 euro + 35 euro/month • Spot: 15 euro/month
- 28. Amazon Web Services • Cloudfront ($0.12 / GB / month) • Elastic MapReduce • DynamoDB • S3 ($0.12 / GB / month) • Glacier ($0.01 / GB / month) • Mechanical Turk • Alexa Web Information Service
- 29. Growing Systems
- 31. Estimates in Software Development • 1 minute -> 2 hours (120x) • 1 hour -> 2 days (48x) • 1 day -> 2 weeks (14x) • 1 week -> 2 months (8x) • 1 month -> 2 years (24x) • 1 year -> forever (infinity)
- 32. Programming in Startups • Code is Liability • Supporting Throw Away Software • Don't be afraid of the if statement! • Develop mainly in master branch • Use feature toggles
- 33. Availability • 99% ~ 3 days • 99.9% ~ 9 hours • 99.99% ~ 1 hour • 99.999% ~ 5 minutes of downtime/year! Optimize also for MTTR not just MTBF!
- 34. Unittests • A Unittest should take at most 10ms to Run • 1054 unittests covering Zemanta's code • Create mock-ups! Don't do integration testing with unittests!
- 35. Principles of Software Documentation 1. Source code and running services are the ultimate documentation! 2. No documentation is much much better than misleading/outdated documentation! 3. Out of sight, out of mind! Documentation should be close to the source code and the running services! 4. Document only what’s absolutely necessary (data structures, development environment setup, release and deployment procedures) 5. Documentation should be part of company culture and engineering pride!
- 36. Take Home Message • The core business of startups is searching for a repeatable and scalable business model, not technology! • Measure Anything, Measure Everything • Do Code Reviews! • Be stupid!