Scalability Best Practices

**Scalability** rears it's ugly head as your startup starts it's growth trajectory from thousand to millions of users.

The talk is about the best practices that developers can start applying when they have their MVP validated. Best practices for not losing your peace of mind and save $$$ as your app grows.

In a scalable architecture, resource usage should increase linearly (or better) with load, where load may be measured in user traffic, or data volume. Based on his past experience working at Google, Mathify (Techstars '13), and now Appbase, Siddharth Kothari will be sharing some of the hard learned scalability lessons.

Transcript

1. Scalability Challenges and best practices Pain points of growing from 1 to 1 million users Siddharth Kothari @sids_aquarius siddharth@appbase.io
2. Siddharth Kothari Founder, Appbase.io Visiting Scholar, SCS@CMU Backend Engineer, Mathify (Techstars ‘13) And I know R + L = J
3. From Obscurity and irrelevance to...
4. Hard to ignore
5. What being “Hard to ignore” means
6. What it can also do
7. What it can also do • Unwanted attention • Need “High Availability” • Costs $$$$
8. Unwanted attention
9. High Availability
10. Costs $$$$
11. “Friendster’s service performance was great when it had 100,000 users, but it degraded to the point of being unusable when it had several million users.” - WSJ article Although, 70% startups fail due to premature scaling
12. Best practices from my experience • Scaling Mathify to 3M monthly unique visitors. • Appbase.io has handled up to 500,000 daily API requests. • Scaling Appbase.io to handle 100M monthly API requests.
13. http://status.appbase.io Premature availability doesn’t hurt. Lesson #1:
14. Lesson #2: Ensure Responsiveness (keep backups)
15. Lesson #3: Find Performance bottlenecks (before your users do)
16. Lesson #4: Stress test (know your limits)
17. Few other things: • Most performance bottlenecks are due to database operations. • Avoid the “Built-in here” syndrome. • All things equal, go for a horizontal architecture. (costs grow linearly as usage grows)
19. Introducing “RAPID” • Understand important metrics - Responsiveness, Availability, Performance • Powered by humans - Grade for Intuitive metrics, that cannot be graded automatically. • Score your app with a full Diagnostics report card.
20. Coming soon...
21. Thank you! Muito Obrigado Siddharth Kothari @sids_aquarius siddharth@appbase.io