Using Serverless & ARM64 for Real-Time Observability

on YouTube ↗41m

" summary

Built a query engine (Retriever) that pulls recent NVMe segments and months of archive from S3, shards the work into tens of thousands of AWS Lambda workers, and merges results in a three‑tier map‑reduce so arbitrary queries return in seconds rather than minutes. Observed P99 local SSD queries drop ~2s to ~1s, scaled storage 20× without proportional cost increase, managed 100k+ concurrent vCPUs via negotiated limits, cut instance count ~20% and achieved ~30% better price‑performance after a Graviton/ARM migration while exploiting sub-second latency and the ability to buy compute in 1ms units.

AI-written summary · about

" chapters (16)

1. 0:00Intro
2. 0:27Agenda
3. 1:07What is lambda for?
4. 1:37What is retriever for?
5. 2:31What is Honeycomb for?
6. 3:09Demo
7. 4:08Interactive investigation of production behavior
8. 4:44Emphasis: interactive
9. 7:48Dynamic aggregation of any fields across any time range
10. 10:31Buy compute in 1ms units
11. 11:08Considerations
12. 18:55It's not all about the money
13. 19:59Software contributes significantly
14. 21:02Wasting less power
15. 37:23Yes*, do this at home
16. 41:06Outro

" related talks

• 49mUsing serverless functions for real-time observabilityLiz Fong Jones · CPH DevFest 2024
• 32mUsing Serverless Functions for Real-time ObservabilityLiz Fong Jones, Jessica Kerr · SREcon22 Americas
• 39mServerless in production, an experience reportYan Cui · Codemotion Milan 2017
• 36mHow we used serverless to speed up our serversJessica Kerr, Ian Wilkes · Strange Loop 2021
• 47mAdvanced Serverless Architectural Patterns on AWSAlex Casalboni · Heapcon 2019
• 54mServerless beyond FunctionsDanilo Poccia · Voxxed Days Bucharest 2018
• 01h 02mServerless in production, an experience reportYan Cui, Domas Lasauskas · NDC London 2018
• 47mServerless in production, an experience reportYan Cui · Build Stuff 2017