DEV Community

Cover image for Docker Image Compression: gzip vs zstd
Daniel Pepuho
Daniel Pepuho

Posted on • Edited on • Originally published at danielcristho.site

Docker Image Compression: gzip vs zstd

#docker #containers #linux #discuss

Docker images are already compressed when you push them to registries like Docker Hub, GHCR, AWS ECR, etc.

So why would anyone compress an image again by using gzip or zstd?

why this happen meme

Because in the real world, engineers often need to:

- move images between servers (air-gapped networks),

- migrates registries,

- backup image to object,

- load or save images repeatedly in Continuous Integration (CI) pipelines.

In these cases, docker save produces a raw .tar file often huge.
Compressing that tarball can cut transfer and storage time by 50-80%.

But what’s the best compression tool? So we will test gzip vs zstd.

🤔 When Do We Need Image Compression?

Like I said before, you need to compress your image to:

- transfer image between SSH or local network (LAN),

- work with offline or air-gapped servers,

- backup images to object storage,

- migrate to new registry.

Skip compression when images are pushed directly to registry like DockerHub, registries are handled that.

Test Up

To get a realistic number, we will test three images:

- alpine:latest -> ~8MB

- maridb:10.6.18 -> ~396MB

- Custom Jupyterlab image -> ~5.4GB

Environment

1. Local Computer

- CPU: 4 cores / 8 threads

- Storage: NVMe SSD

- OS: Ubuntu 22.04

- Docker Version: 27x

- Tools: gzip, zstd, time, scp

2. VPS

- CPU: 4 cores / 8 threads

- Storage: VirtIO-backed SSD

- OS: Ubuntu 22.04

- Docker Version: 27x

- Tools: gzip, zstd, time

Command used

Below are the commands used to save, compress, transfer, decompress, and load the Docker images during testing.

1. Save the image (local machine) 

docker pull <image_name>
docker save <image_name> -o <image_name>.tar
Enter fullscreen mode Exit fullscreen mode 
$ docker pull alpine:latest

latest: Pulling from library/alpine
Digest: sha256:4b7ce07002c69e8f3d704a9c5d6fd3053be500b7f1c69fc0d80990c2ad8dd412
...
Enter fullscreen mode Exit fullscreen mode 
$ docker save alpine:latest -o alpine_latest.tar

$ docker save mariadb:10.6.18 -o mariadb_10.tar

$ docker save danielcristh0/datascience-notebook:python-3.10.11 -o jupyter_notebook.tar
Enter fullscreen mode Exit fullscreen mode 

$ ls -lh

total 5,6G
-rw------- 1 user group 8,3M Nov 28 19:51 alpine_latest.tar
-rw------- 1 user group 5,2G Nov 28 20:18 jupyter_notebook.tar
-rw------- 1 user group 384M Nov 28 20:14 mariadb_10.tar
Enter fullscreen mode Exit fullscreen mode

2. Compress the image (local machine)

- gzip

time gzip -k <image>.tar
Enter fullscreen mode Exit fullscreen mode 
$ time gzip -k alpine_latest.tar

gzip -k alpine_latest.tar  0,44s user 0,01s system 99% cpu 0,452 total
Enter fullscreen mode Exit fullscreen mode 
$ time gzip -k mariadb_10.tar

gzip -k mariadb_10.tar  17,21s user 0,62s system 99% cpu 17,979 total
Enter fullscreen mode Exit fullscreen mode 
$ time gzip -k jupyter_notebook.tar

gzip -k jupyter_notebook.tar  238,83s user 3,56s system 99% cpu 4:03,16 total
Enter fullscreen mode Exit fullscreen mode

-k → keep original file

gzip uses a single CPU thread at its default level (≈ level 6)

- zstd

time zstd -T0 -19 <image>.tar
Enter fullscreen mode Exit fullscreen mode 
$  time zstd -T0 -19 alpine_latest.tar

alpine_latest.tar    : 37.01%   (8617984 => 3189867 bytes, alpine_latest.tar.zst)
zstd -T0 -19 alpine_latest.tar  3,64s user 0,10s system 100% cpu 3,734 total
Enter fullscreen mode Exit fullscreen mode 
$  time zstd -T0 -19 mariadb_10.tar

mariadb_10.tar       : 16.95%   (402636288 => 68258055 bytes, mariadb_10.tar.zst)
zstd -T0 -19 mariadb_10.tar  172,89s user 0,66s system 191% cpu 1:30,81 total
Enter fullscreen mode Exit fullscreen mode 
$  time zstd -T0 -22 jupyter_notebook.tar
Warning : compression level higher than max, reduced to 19
zstd: jupyter_notebook.tar.zst already exists; overwrite (y/n) ? y

jupyter_notebook.tar : 24.79%   (5560227328 => 1378450873 bytes, jupyter_notebook.tar.zst)
zstd -T0 -22 jupyter_notebook.tar  4759,54s user 19,32s system 188% cpu 42:11,68 total
Enter fullscreen mode Exit fullscreen mode

-T0 → use all CPU threads

-22 → request maximum compression (automatically reduced to -19 by zstd)

3. Transfer to VPS

- gzip

time scp <image_name>.tar.gz user@vps:/tmp/
Enter fullscreen mode Exit fullscreen mode 
$ time scp alpine_latest.tar.gz onomi@myserver:/tmp
alpine_latest.tar.gz    100% 3588KB 174.8KB/s   00:20

scp alpine_latest.tar.gz onomi@myserver:/tmp  0,11s user 0,29s system 1% cpu 23,208 total


$ time scp mariadb_10.tar.gz onomi@myserver:/tmp/
mariadb_10.tar.gz   100%  114MB   2.2MB/s   00:50

scp mariadb_10.tar.gz onomi@myserver:/tmp/  0,46s user 0,84s system 2% cpu 52,457 total


$ time scp jupyter_notebook.tar.gz onomi@myserver:/tmp/
jupyter_notebook.tar.gz  100% 1765MB   3.4MB/s   08:35

scp jupyter_notebook.tar.gz onomi@myserver:/tmp/  5,03s user 10,42s system 2% cpu 8:38,50 total
Enter fullscreen mode Exit fullscreen mode

- zstd

time scp <image_name>.tar.zst user@vps:/tmp/
Enter fullscreen mode Exit fullscreen mode 
$ time scp alpine_latest.tar.zst onomi@myserver:/tmp
alpine_latest.tar.zst   100% 3115KB 343.4KB/s   00:09

scp alpine_latest.tar.zst onomi@myserver:/tmp  0,10s user 0,18s system 1% cpu 22,728 total


$ time scp mariadb_10.tar.zst onomi@myserver:/tmp/
mariadb_10.tar.zst  100%   65MB   3.0MB/s   00:21

scp mariadb_10.tar.zst onomi@myserver:/tmp/  0,29s user 0,59s system 3% cpu 23,285 total


$ time scp jupyter_notebook.tar.zst onomi@myserver:/tmp/
jupyter_notebook.tar.zst    100% 1315MB   2.3MB/s   09:44

scp jupyter_notebook.tar.zst onomi@myserver:/tmp/  3,94s user 7,64s system 1% cpu 9:46,33 total
Enter fullscreen mode Exit fullscreen mode

4. Load the Image on the Server (VPS)

Now the compressed images are transferred to the VPS, the next step is to decompress them and load the Docker image into the remote server.

- gzip

time gzip -dk <image>.tar.gz
Enter fullscreen mode Exit fullscreen mode 
$ time gzip -dk alpine_latest.tar.gz

real    0m0.189s
user    0m0.116s
sys     0m0.039s


$ time gzip -dk mariadb_10.tar.gz


real    0m5.108s
user    0m3.813s
sys     0m1.129s

$ time gzip -dk jupyter_notebook.tar.gz

real    1m8.344s
user    0m48.466s
sys     0m13.408s
Enter fullscreen mode Exit fullscreen mode

- zstd

time zstd -d <image>.tar.zst
Enter fullscreen mode Exit fullscreen mode 
$ time zstd -d alpine_latest.tar.zst
alpine_latest.tar.zst: 8617984 bytes

real    0m4.121s
user    0m0.041s
sys     0m0.043s

$ time zstd -d mariadb_10.tar.zst
mariadb_10.tar.zst  : 402636288 bytes

real    0m3.455s
user    0m0.983s
sys     0m0.927s

$ time zstd -d jupyter_notebook.tar.zst

jupyter_notebook.tar.zst: 5560227328 bytes

real    0m31.810s
user    0m14.599s
sys     0m13.600s
Enter fullscreen mode Exit fullscreen mode

Decompression in zstdis extremely fast, 5-10x faster than compression, even for large files.

5. Loading the Image Into Docker

Once decompressed, load the .tar file:

docker load -i <image>.tar
Enter fullscreen mode Exit fullscreen mode 
$ docker load -i jupyter_notebook.tar

Loaded image: danielcristh0/datascience-notebook:python-3.10.11
Enter fullscreen mode Exit fullscreen mode 
$ docker images

IMAGE                                               ID             DISK USAGE   CONTENT SIZE   EXTRA
danielcristh0/datascience-notebook:python-3.10.11   9b38bf7c570f       11.4GB         5.56GB
Enter fullscreen mode Exit fullscreen mode

6. Analysis: gzip vs zstd

After running all compression, transfer, decompression, and loading tests across three different Docker images, let's compare gzip and zstd.

Size Comparison

zstd consistently produces much smaller output files than gzip, especially on medium and large images.

Image Actual Size gzip Size zstd Size Reduction (gzip) Reduction (zstd)
alpine:latest 8.3 MB 3.5 MB 3.1 MB ~57% ~62%
mariadb:10.6.18 384 MB 114 MB 65 MB ~70% ~83%
jupyter-notebook 5.2 GB 1.7 GB 1.3 GB ~67% ~75%

zstd gives around 20-50% better compression than gzip.

Speed

gzipis faster than zstd at compression.

Image gzip (time) zstd (time) Notes
alpine 0.45 s 3.7 s 8x slower
mariadb 17.9 s 90.8 s 5x slower
jupyter-notebook 243 s 42 minutes 10x slower

zstd gives better compression but requires significantly more CPU.

Transfer Speed (via SCP)

Because zstd produces smaller files, transfer times are 2x faster. But on larger files, zstd can still lose to gzip depending on CPU and disk performance.

Image gzip Transfer zstd Transfer
alpine 20 s 9 s
mariadb 50 s 21 s
jupyter-notebook 8m 35s 9m 44s

When Should You Use gzip or zstd?

Use zstd when you want

- The smallest compressed Docker images

- Fast decompression

- Faster transfers across networks

- Long-term backups

Use gzip when you want:

- Fast compression

- Low CPU usage

- Simple, predictable behavior

- Occasional small image transfers

Conclusion

If you need to compress Docker images, here’s the quick answer:

Use zstd when you want:

- Smaller archive sizes (around 20-50% smaller than gzip)

- Faster decompression

- Faster network transfers

Use gzip when you want:

- Fast compression

- Low CPU usage

- Simplicity

Aight, that's all. Thank you for taking your time to read this post.

Feel free to give me feedback, tips, or different benchmarks. I’d love to hear your feedback and continue the discussion.

Happy containerization! 🐳

Top comments (0)