Is there anybody out there?

on YouTube ↗52m

" summary

Covers concrete choices: use Akka Cluster with phi accrual detector tuned to thresholds ~8–12 (recommend ~12 on AWS; detection can be ~0.5s at 6), or newer adaptive detectors tested in EC2; use SWIM with Lifeguard for scalable failure detection plus anti-entropy TCP syncs; gossip via randomized or piggybacked messages and member state vectors stored in memberlist; prefer CRDTs for low-latency replicated state, Paxos/Raft for strong logs, and deploy a split-brain resolver to decide membership and avoid divergent clusters (tested at ~2.4–2.6k nodes; Lifeguard cuts false positives to <2%).

AI-written summary · about

" chapters (48)

1. 0:00Introduction
2. 0:35Disclaimer
3. 1:08Manuel Bernhardt
4. 2:13Quote
5. 2:54Clusters
6. 4:00How do we build clusters
7. 5:04How do we implement group membership
8. 5:34Who can tell me
9. 5:44Wish you were here
10. 6:03What is completeness
11. 6:47What is accuracy
12. 7:58Many things are impossible
13. 8:49We have to make tradeoffs
14. 9:17In practice most applications
15. 10:05Reallife failure detectors
16. 13:11Threshold
17. 14:29Adaptive failure detector
18. 15:35Swim
19. 17:07Garbage Collection
20. 17:41Fanout
21. 19:05Lifeguard
22. 21:12Pulse
23. 21:28Alan Paul
24. 22:07multicast
25. 22:21multicast support
26. 22:48UDP multicast
27. 23:59Napster
28. 24:44Random Gossip
29. 25:20Binary Roundrobin
30. 25:38Sequence Check
31. 25:55Infection Style
32. 27:15Gossip Style
33. 28:47Optimizations
34. 30:57I hurt you
35. 31:15Nietzsche
36. 38:01Consensus protocols
37. 38:57Conflictfree replicated data types
38. 40:22Membership consensus
39. 40:53In practice
40. 42:18Split brain resolvers
41. 42:44Deploying technologies
42. 43:13Who is there
43. 43:24Temperature anomalies
44. 44:13Why are there so many arrows
45. 44:26Im running out of time
46. 44:40Thank you very much
47. 44:53There are a few I reviewed
48. 48:17There is a delay

" related talks

• 13mBuilding Financial Fortresses: Real-Time Fraud DetectionPrakash Vanga · Site Reliability Engineering (SRE) 2025
• 18mBuilding a Scalable, Resilient System for Real-Time Fraud DetectionSubhankar Bhattacharya · Distributed SQL Summit Asia 2025
• 29mBuilding Resilient AI Platforms: Financial Fraud DetectionRahul Ganti · Conf42 Platform Engineering 2025
• 22mReliability Patterns in Permissioned Blockchain SystemsUmegbewe Nwebedu · Site Reliability Engineering (SRE) 2025
• 46mData Analytics-Based Detection At ScalePedro Chaves, Jan Pospisil · BSides Lisbon 2018
• 25mReal-Time Fraud Detection at ScaleSatyanandam Kotha · Conf42 Machine Learning 2025
• 40mSh... Fail Happens: Fail-aware Events Processing on AWSMarcin Sodkiewicz · GOTO EDA Day London 2024
• 16mDispersedLedger: High-Throughput Byzantine Consensus on Variable Bandwidth NetworksLei Yang, David Tse, Sreeram Kannan, Seo Jin Park, Mohammad Alizadeh · 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI '22)