A tale about scale. Norbert Hartl - ApptiveGrid

Talk from ESUG 2024.

"A tale about scale" by Norbert Hartl https://www.apptivegrid.de

Video: https://www.youtube.com/watch?v=UpbSktWsG7U
PDF: http://archive.esug.org/ESUG2024/day3/08-a-tale-about-scale-norbert.pdf

1. Norbert Hartl - ApptiveGrid
a tale about scale

2. Norbert Hartl - ApptiveGrid
a tale about scale
caution: only remotely technical

3. I'm talking about
pharo in practice

4. Pharo days 2015

5. What is scale?

6. technical scale

7. Stop it!

8. keeping focus - CPM model

9. CPM model
•Computer
•Pharo
•Me

11. team scale

12. two laptops
on a network

13. debugging

14. Repeat

15. DOS attacks

17. local$ scp ApptiveGrid.* backend-server:
local$ ssh backend-server
backend-server$ restart-image.sh
AGModel readStonFromDisk
deployment scale

19. deployment scale
•use database
•add migration
•do backups

21. team scale
•issue tracker
•code repository
•development process (pull requests, review,...)
•CI pipeline building artefacts
•artefact management
•deployment scripts

23. •approx. 5000 users
approx. 5000 users

24. resilience

25. resilience
•images can crash

26. resilience
•images can crash
•You need more images

27. resilience
•images can crash
•You need more images
•You need a load balancer

28. resilience
•images can crash
•You need more images
•You need a load balancer
•Machines can also fail

29. resilience
•images can crash
•You need more images
•You need a load balancer
•Machines can also fail
•You need more machines

30. resilience
•images can crash
•You need more images
•You need a load balancer
•Machines can also fail
•You need more machines
•You need a machine for the load balancer

31. resilience
•images can crash
•You need more images
•You need a load balancer
•Machines can also fail
•You need more machines
•You need a machine for the load balancer
•The load balancer machine can fail

32. resilience
•images can crash
•You need more images
•You need a load balancer
•Machines can also fail
•You need more machines
•You need a machine for the load balancer
•The load balancer machine can fail
•You need a backup of the load balancer

33. resilience
•images can crash
•You need more images
•You need a load balancer
•Machines can also fail
•You need more machines
•You need a machine for the load balancer
•The load balancer machine can fail
•You need a backup of the load balancer
•...

34. resilience
... and so far I didn't even care about
logging, monitoring, metrics, ...

35. What is scale?

36. When to scale? (Internet version)
technical scale -> success

37. When to scale? (real version)
success -> technical scale

38. Conclusion
•scaling is distraction

39. Conclusion
•scaling is distraction
•no gain without losing something

40. Conclusion
•scaling is distraction
•no gain without losing something
•when benefits outweigh the drawbacks significantly... scale

41. Conclusion
•scaling is distraction
•no gain without losing something
•when benefits outweigh the drawbacks significantly... scale
•If not then not!

42. thank you!
Norbert Hartl
[email protected]