[2016.06 Meetup #3][TALK #1] João Miranda - Zero-Friction Performance Instrumentation And Monitoring

OutSystems Platform offers performance instrumentation, monitoring and troubleshooting capabilities for the applications it creates. The goal is to offer a zero-friction experience without requiring any special instrumentation effort when building the application. We have to decide which metrics to gather, how to gather them and mash it all up in out-of-the-box dashboards.

Transcript

1. Zero-friction Performance Instrumentation And Monitoring From A To Z
2. Agenda •  The Context •  The Problem •  The Use Cases •  The Solution •  The Operation •  The Results
3. The Context
4. OutSystems Platform allows you to create, deploy and change custom mobile and web applications
5. DB Query REST Call Variable Assignment
6. Code A Visual Model Generate Application Code Deploy To Web Server And DB Operate (Monitoring, Security, …) Lifecycle
7. WHAT PROBLEM DO WE WANT TO SOLVE? The Problem
8. (Geeky) Developers Architects Listen To Your Stakeholders CTOs CEOs Ops InfoSec (Casual) Developers
9. Solve The Problems Where You Can Make a Difference
10. After Doing The Field Research…
11. Project’s Ultimate Goal “Reduce to zero the time taken to identify (mobile) performance issues and pinpoint their source”
12. Always Define Your Own Ultimate Goal!
13. WHAT ARE THE COMMON PERFORMANCE BOTTLENECKS? The Key Use Cases
14. Performance Is A Huge Topic
15. We Need To Focus
16. Module (web application) Desktop / Smartphone / Tablet Remote My App DB REST, … Custom C# or Java
17. Module (web application) Desktop / Smartphone / Tablet Remote My App DB REST, … Custom C# or Java Common Performance Bottlenecks
18. Performance Bottlenecks •  Slow (Mobile) Networks •  Slow Queries •  Slow Integration Calls (e.g.: SOAP, REST) •  Slow Extension Calls (custom C#/Java code) •  (Select) N+1 – Fast call, but executed many times
19. Raw Performance Is Meaningless
20. Module (web application) Desktop / Smartphone / Tablet Remote Expenses Report DB REST, … Custom C# or Java Login
21. Module (web application) Desktop / Smartphone / Tablet Remote Weekly Report DB REST, … Custom C# or Java Process
22. User Perceived Performance, In Context, Is What Matters
23. You Know Your Product Best. Find Your Own Common Performance Bottlenecks. Present Them In A User-Centered Focus.
24. INSTRUMENTATION WITHOUT EFFORT The Solution
25. Code A Visual Model Generate And Instrument Application Code In OutSystems Platform Case
28. There Are Many Ways To Do It You don’t necessarily need code generation...
29. Proxy Key Objects Boomerang is a JavaScript library that does real user monitoring (RUM). Boomerang provides the auto_xhr plugin. It proxies the XHR object to monitor server requests. Measuring the performance of single page applications
30. Define And Enforce Coding Guidelines •  Define the standard data access libraries •  Define the standard REST client libraries •  Do code reviews •  Leverage sta9c analysis tools
31. The Right Thing To Do Must Be Simple. The Wrong Thing To Do Must Be Hard.
32. FIGHTING THE HEISENBERG PRINCIPLE The Operation
33. Instrumentation Can Generate A Lot Of Data
34. Constraints Heisenberg principle Memory consumption Storage needs
35. Before Going Live Performance Baselines Load Testing Performance Testing
36. After Going Live Throttling Capabilities Time To Live Policies
37. Don’t Let Your Instrumentation Bring Your Application To A Halt.
38. CONSUMING INSTRUMENTATION DATA The Results
39. Instrumentation provides very rich data points
40. OutSystems Instrumentation Per Request (among many other data points) Endpoint metadata End-to-end request duration Server-side duration Slow call (queries, integrations) data Network type Mobile device type
41. We Have A Few Key Use Cases. But We Don’t Want To Keep The Data Trapped.
42. Best Of Both Worlds Public API Provides Raw Data To Cater For All The Other Known (And Unknown) Use Cases
43. Best Of Both Worlds Web Dashboard, Combined With Automatic Instrumentation, Provides Zero-Friction Monitoring For The Key Use Cases Identified During Research Phase
44. End-to-end (client, network, server) Performance
45. How much time is spent in remote calls?
46. How much time is spent in slow calls?
47. Focus On Your Key Use Cases. Keep Your Analytics Options Open. Always Provide APIs.
48. Feel free to reach out! João Miranda joao.hugo.miranda@gmail.com @jhosm https://www.linkedin.com/in/jhosm
49. Thanks For Your Time!