Jonathan Perry

CEO Unvariance (Kubernetes resource scheduling tool to improve efficiency and latency), 2xFounder, MIT CS PhD, cyclist.

[Jonathan Perry @yonchco]

@confusedqubit @lennypruss This is great, thanks for sharing! I’d love to hear more about how this works!
How this S3-based filesystem is storing metadata to start..
btw the microsecond numbers are average for cache hits, right? If two servers mount the filesystem, a cache miss would have to go to S3..?

[P99CONF @P99CONF]

When containers compete for memory bandwidth, even well-tuned applications suffer slowdowns. @yonchco's #P99CONF session will show how to identify issues, pinpoint their cause, & implement targeted fixes that improve performance & reduce cost. https://t.co/7ZKgLahO4d

#ScyllaDB

[Jonathan Perry @yonchco]

Join me at #OSSummit Europe!

Tue Aug26 14:10: why competition for memory bandwidth can severely hurt server performance and how we could mitigate such degradation. We'll cover available kernel mechanisms, their gaps, and how we're filling in those gaps.

https://t.co/65bTke5wnf

[Jonathan Perry @yonchco]

Looking forward to meeting folks at KCD Czech & Slovak in Bratislava!

[KCD Czech & Slovak @KCDCzechSlovak]

🌤️ In cloud-native setups, resource contention hurts app performance. Learn how hyperscalers like Google and Alibaba tackle it and explore ideas for a Kubernetes-native performance isolator to boost utilization without sacrificing speed. #kcdczsk2025 #cloudnative #cncf

[KCD Czech & Slovak @KCDCzechSlovak]

🌤️ In cloud-native setups, resource contention hurts app performance. Learn how hyperscalers like Google and Alibaba tackle it and explore ideas for a Kubernetes-native performance isolator to boost utilization without sacrificing speed. #kcdczsk2025 #cloudnative #cncf

[Jonathan Perry @yonchco]

Kubecon is coming up!

Please join my talk about how to mitigate increased service times (found to increase >25%), and tail latency (4-13x), due to uncontrolled use of memory bandwidth and cache resources.

On the schedule: https://t.co/xNU7SPktmK

#KubeCon + #CloudNativeCon

[Tim Urban @waitbutwhy]

Refreshingly nuanced

[Jonathan Perry @yonchco]

@richardartoul Interesting! And there’s 2x throughout “amplification” for compaction? Does that factor change as a function of retention time?

[Jonathan Perry @yonchco]

@fredbrancz @RocksetCloud talked about a cache which their rocksdb instances use to fetch blocks from S3. Not sure if it’s open source.

Seems S3 caching should be more prevalent.. Hope a strong open implementation will emerge.

https://t.co/OgZgXgknKO

[Jonathan Perry @yonchco]

@fredbrancz @PolarSignalsIO does that function start with “Robert’); DROP TABLE”?

https://t.co/0ixt2aV7Lk

[Jonathan Perry @yonchco]

@gwenshap Also apparently Dgraph uses ALICE from OSDI’14

https://t.co/5JJzTmx6sS

[Jonathan Perry @yonchco]

@gwenshap not exactly chaos engineering, FoundationDB added deterministic simulation and stress-tested the code using random execution traces overnight.

Similar ideas are Rust madsim and I saw S3 did something similar in their formal verification paper

FDB: https://t.co/gdz3UcEcp3

[Jonathan Perry @yonchco]

@YuvalAriav בהצלחה!

[Jonathan Perry @yonchco]

@rakyll I’m guessing Paxos.

[Jonathan Perry @yonchco]

@YuvalAriav הציעו לבטל את הקוד ריוויו כדי לשחרר גרסאות מהר

[Jonathan Perry @yonchco]

@ya_ronen 100%. מנהל צריך לספק הקשר על בקשות לפגישות. ויזמים צריכים להקפיד עוד יותר.

אחרי שהודענו לעובדים שעומדים למכור את הסטארט-אפ שלחתי "אפשר לדבר?" לקוני לאחד העובדים.. אני מקפיד מאז לתת סיבה בהודעה הראשונה, גם כשאין אירוע חריג.

[Jonathan Perry @yonchco]

@nadavwiz You know today is Tuesday right? 🤗

[Jonathan Perry @yonchco]

@nadavwiz Good luck wiz!

[Jonathan Perry @yonchco]

Q: How do you call developers constantly changing Story Points mid-sprint?

A: Chaos Engineering!