Master Linux I/O Schedulers, fio Workloads, and RAID Performance

Introduction

The article discusses Linux I/O schedulers, the fio workload generator, various RAID levels, and tools for monitoring and analyzing storage performance.

Choosing an I/O Scheduler

RHEL8 uses a multi‑queue I/O scheduler ( mq-deadline) to replace the old single‑queue scheduler. It describes the Noop, Deadline, CFQ, BFQ, and Kyber algorithms, their characteristics, and typical use cases such as SSDs, databases, or workloads that should not be reordered.

Using fio to Simulate Workloads

Installation and basic usage examples are provided, e.g.:

# dnf provides fio
# dnf -y install fio
# fio --name=randwrite --ioengine=libaio --iodepth=1 \
    --rw=randwrite --bs=4k --direct=1 --size=512M \
    --numjobs=2 --group_reporting --filename=/tmp/testfile

The output shows IOPS, bandwidth, latency percentiles, and CPU usage.

RAID and Disk Arrays

The article explains RAID concepts (striping, mirroring, parity) and details RAID 0, RAID 1, RAID 4, RAID 5, RAID 6, and RAID 10. It shows how to create software RAID with mdadm:

# mdadm -C /dev/md0 -l raid0 -n 2 /dev/vdb /dev/vdc
# mdadm --stop /dev/md0
# mdadm --remove /dev/md0
# mdadm --zero-superblock /dev/vdb
# mdadm --zero-superblock /dev/vdc

and how to format the array with matching chunk size for XFS or ext4:

# mkfs -t xfs -d su=64k,sw=4 /dev/san/lun1
# mkfs -t ext4 -E stride=16,stripe-width=64 /dev/san/lun1

It also covers creating RAID logical volumes with LVM:

# lvcreate --type raid0 -L 3G --stripes 3 --stripesize 4K -n raidlv radivg

Monitoring Tools

Commands for observing I/O performance are demonstrated: top – watch the wa (I/O wait) column. iostat -Np – display per‑device I/O statistics. iotop -a -P -o – show processes actively performing I/O. blktrace -d /dev/mapper/rl-root -o disk.log – capture block‑level traces, then analyze with btt. atop (part of pcp‑system‑tools) – provide a comprehensive view of CPU, memory, disk, and network usage.

Practical Commands Summary

Key commands include installing monitoring utilities, creating and managing RAID, formatting filesystems with appropriate stride settings, and running fio tests to evaluate storage performance.