Does enabling asynchronous parallel metadata downloads in DNF alleviate I/O bottlenecks on Fedora Workstation?

Answer

Enabling DNF's asynchronous, parallel metadata downloads usually shortens wall‑clock upgrade time on fast storage (NVMe/SATA SSD) but can increase I/O queue depth and await times on HDDs or low‑end SSDs, so the net effect depends on your disk subsystem.

Confirmed facts

• DNF provides an asynchronous option in /etc/dnf/dnf.conf that launches multiple concurrent metadata fetches.
• On modern SSDs with high IOPS, parallel downloads reduce total transaction latency without raising average await time significantly.
• On spinning disks or low‑performance storage, concurrent streams can raise queue depth, increasing await measured by iostat -x and may offset latency gains.

Steps to verify on your workstation

1. Check current setting:

   grep asynchronous /etc/dnf/dnf.conf

2. Temporarily disable it for a baseline:

   sudo dnf config-manager --setopt asynchronous=False --save

   (or edit the file directly).
3. Run a typical upgrade while monitoring I/O:

   sudo dnf upgrade -y & iostat -xz 1 30

   Note the average await and total time reported by DNF.
4. Re‑enable parallel downloads:

   sudo dnf config-manager --setopt asynchronous=True --save

5. Repeat the upgrade with the same iostat capture and compare await, util%, and DNF‑reported transaction duration.
6. If the await stays low and total time drops, the feature helps; if await rises sharply, it may be hurting I/O.

Missing diagnostic detail

To give a definitive recommendation, I need to know the primary storage device in your Fedora Workstation (e.g., NVMe SSD, SATA SSD, or HDD). Please run

lsblk -d -o NAME,ROTA,MODEL

and share whether ROTA is 0 (solid‑state) or 1 (rotational).