Tagged articles

Low Overhead

4 articles · Page 1 of 1

Jul 6, 2026 · Cloud Native

Observe Every AI Agent Call Without Changing a Single Line of Code

OBI uses Linux kernel eBPF instrumentation to automatically capture and parse all AI‑related HTTP traffic—covering LLM, embedding, vector search, rerank and MCP tool calls—producing OpenTelemetry‑compatible traces and metrics without any code changes, enabling full‑stack observability of multi‑provider AI agents across languages with only ~1% CPU overhead.

AI ObservabilityGenAILinux kernel

0 likes · 21 min read

Observe Every AI Agent Call Without Changing a Single Line of Code

Linux Kernel Journey

Jan 21, 2025 · Fundamentals

How Kernel‑Space Memory Profiling Works with Code Tagging (memprofiling)

The article explains a kernel‑space memory profiler that uses a compile‑time code‑tagging library and special linker sections to record allocation callers and byte counts with minimal CPU overhead, allowing the tool to be deployed in production.

Low Overheadalloc_tagcode tagging

0 likes · 4 min read

How Kernel‑Space Memory Profiling Works with Code Tagging (memprofiling)

Cognitive Technology Team

Apr 17, 2024 · Operations

Using eBPF to Capture Complete Java Call Stacks from OpenJDK 8 JVM without Agent

The team successfully employed eBPF dynamic tracing to obtain full Java call‑stack traces from any point in an OpenJDK 8 JVM process with microsecond‑level overhead, without requiring any JVM agents, bytecode injection, or code modifications, making it suitable for production environments.

JVM tracingJavaLow Overhead

0 likes · 2 min read

Using eBPF to Capture Complete Java Call Stacks from OpenJDK 8 JVM without Agent

Beike Product & Technology

Jul 16, 2017 · Industry Insights

How Lianjia Built LTrace: A Low‑Overhead, Scalable Distributed Tracing Platform

This article explains how Lianjia designed and implemented LTrace, a zero‑intrusion, high‑performance distributed tracing system that captures full request chains across heterogeneous services, supports multi‑language environments, offers flexible sampling, and enables rapid fault isolation and performance optimization.

ArchitectureLow OverheadObservability

0 likes · 12 min read

How Lianjia Built LTrace: A Low‑Overhead, Scalable Distributed Tracing Platform