gRPC over HTTP/3 in Production: QUIC Handshakes, 0‑RTT Risks, and a Safe Migration Path
You’ve got a fleet of gRPC services humming along on HTTP/2. Mobile clients suffer from shaky networks, tail latencies sting, and someone on the team keeps a...
Recent posts
Comparing Workflow Architectures: Prefect vs. Dagster vs. Airflow
What really happens when you “schedule a job”? If you’ve ever pushed a button in a nice UI and watched a pipeline spring to life, you’ve probably felt the or...
Demystifying CPU Requests and Limits in Kubernetes: How the Scheduler Really Works
Most of us learn “set CPU requests and limits” as a cargo-cult rule. Then, sooner or later, a service starts stuttering under load and someone declares: “Nev...
Should You Upgrade to Python 3.13? Here’s What You’ll Gain (and Lose)
Python 3.13 shipped on October 7, 2024, and it’s not just another point release. It brings a brand‑new interactive interpreter, experimental free‑threading (...
ClickHouse: Using Indexes, Projections, and Data Skipping for Speed
TL;DR ClickHouse doesn’t use B‑trees the way OLTP databases do. Its speed comes from: a sparse primary index on the ORDER BY key to jump to releva...
Introducing LiveKit Dashboard: Simplifying Self-Hosted Real-Time Management
Running LiveKit at scale is powerful — but it can also feel like you’re managing a live concert with the lights off. You’ve got rooms, participants, servers,...
Breaking Down Python 3.13’s Experimental Free‑Threading Mode
The year the GIL blinked. With Python 3.13, CPython gained an experimental “free‑threading” build that can run threads truly in parallel on multiple cores. I...
Parts, Not Pages: A Mental Model for How ClickHouse Stores Data
If you’ve used ClickHouse for more than ten minutes, you’ve probably seen mysterious words in system tables and error messages: parts, granules, marks, mutat...
Adding a Column in ClickHouse, Deeply: What Actually Happens and What to Consider
“It’s just one more column.” Famous last words.
Achieving Sub‑Second Data Syncs: The Performance Engineering Behind PeerDB
How do you move fresh data from PostgreSQL into analytics systems, queues, and storage engines in hundreds of milliseconds, not minutes? This post walks thro...
How We Handle Schema Evolution in Real‑Time Data Pipelines with PeerDB
When your product team renames a column at 2:07 a.m., will your analytics still be green at 2:08? This post is about making sure the answer is “yes.”
Quantifying Quality: A Deep Dive into the rtcscore Library for WebRTC Mean Opinion Score (MOS) Estimation
In the domain of real-time communication, the concept of “quality” is both paramount and profoundly elusive. For developers and architects of WebRTC applicat...