When a Linux server shows high iowait and becomes unresponsive, this guide walks you through using iostat, iotop, blktrace, fio and bpftrace to pinpoint the offending process, understand the full I/O stack, and apply kernel‑level and filesystem tuning to resolve the bottleneck.
This guide explains Linux disk performance indicators—utilization, saturation, IOPS, throughput, and latency—their interrelationships, and how to monitor both overall and per‑process I/O using tools like iotop, pidstat, and the /proc filesystem.
This article explains Linux disk hardware basics, the block device workflow, I/O scheduler options, how to view and change the scheduler, and key performance metrics such as utilization, saturation, IOPS, throughput, and response time using tools like iostat.
This guide explains how to install fio on Ubuntu 20.04, configure test environments, run IOPS and latency benchmarks with specific parameters, and interpret key metrics such as bandwidth, IOPS, slat, and clat to evaluate storage performance under high‑load and single‑request scenarios.
This article explains the various classifications of Linux I/O—including buffered, direct, synchronous, and asynchronous types—describes the kernel’s block layer and I/O scheduler algorithms, and provides practical guidance for measuring disk performance using the FIO tool with example commands and scripts.
This article explains why mechanical hard drives are slower and less reliable than electronic components, introduces RAID 0, RAID 1, and RAID 5 as solutions to improve speed and fault tolerance, discusses RAID card caching, and shows practical Linux commands for examining RAID hardware and configuration.
A detailed case study examines an Exadata X3‑2 deployment where CALIBRATE tests revealed intermittent disk performance drops linked to low temperatures on Seagate SAS drives, explains why moving disks between nodes can restore performance, and clarifies the normalcy of slight MB/s variations in slots 0 and 1.
This article explains how different dd command variations affect disk write‑cache handling and why using the conv=fdatasync option yields the most reliable measurement of actual disk write performance.
