x10 Wins: Ugly Scaling + Bulletproof Memory Cuts

We've got 2 brilliant lectures on real production case studies on reducing memory footprint by order of magnitudes. One from DoubleVerify, using hackie but efficient data structures, and the other from Next Insurance with 4 real world methodologies shown on stage.

Thanks to DoubleVerify for sponsoring and hosting the event!

AGENDA

18:00–18:30 Mingling, food, and drinks
18:30–18:40 Opening words
18:40–19:25 Beautiful Code Is Overrated: How "Ugly" Engineering Saved Geo at Scale / Alex Cherny - DoubleVerify
19:30–20:15 Bulletproof Cookbook to Reduce Your Service's Memory / Haim Yadid - Next Insurance

Beautiful Code Is Overrated: How "Ugly" Engineering Saved Geo at Scale
We often celebrate clean, elegant architecture, but when data grows 30× overnight, beauty doesn't keep the lights on. This talk walks through a real-world case study of how the Geo traffic service at DoubleVerify broke under 500M IP ranges and how a brutal, byte-level approach outperformed every elegant solution. I'll share the challenges, trade-offs, and the unconventional design choices that helped us slash ramp-up time from 70 minutes to under 3 and memory from 193GB to 13GB. Sometimes, "ugly" is what scales.

Alex Cherny / DoubleVerify
Alex Cherny is a Director of Software Engineering at DoubleVerify, where he leads large-scale Traffic systems operating at millions of requests per second. He has extensive experience building and optimizing JVM-based distributed systems under strict latency, memory, and reliability constraints. Alex focuses on pragmatic performance engineering and believes that clean abstractions are valuable—until scale forces you to break them.

Bulletproof Cookbook to Reduce Your Service's Memory
In recent months, I’ve been working on a Java service that constructs and processes large in-memory object graphs. The memory consumption has become substantial. As a result, through a systematic analysis and targeted optimizations, I have managed to cut the memory footprint by more than 60%. Leading to cost reduction. In this session, I’ll guide you through the journey - from identifying the memory issues using Eclipse MAT to implementing practical optimizations that led to these savings. You’ll learn how to apply similar techniques to your own services to reduce resource usage and improve performance. At the end of the talk, I will also mention several optimizations that improved runtime performance by a factor of 4.

Haim Yadid / Next Insurance
A software engineer with over 30 years of experience in various technology fields: software development, team management, software architecture, HPC and algorithmic research. In my day job I work in Next Insurance as an engineering fellow where I am involved in numerous cross R&D strategic projects. Before that has been focusing for 8 years on Java platform performance optimization as an independent consultant and was involved in more than a dozen optimization projects. I am a Java Champion and one of the founders of Java.IL, the Israeli Java user group.

FOOD
Light food, light drinks, and beer will be served courtesy of DoubleVerify.

RECORDING
We are recording, as always, this meetup and uploading it to our YouTube channel (be sure to subscribe), but please keep in mind asking questions about stuff you don't understand is only possible when you attend the live event.

ADDRESS
Alon Tower 2, Yigal Alon 94, Tel Aviv
Floor 27