Marek Goldmann

Here is some report that shows iowait too: ``` $ sar 1 Linux 3.10.0-327.10.1.el7.x86_64 (performance) 03/29/2016 _x86_64_ (4 CPU) 10:35:38 AM all 0.00 0.00 0.25 0.00 0.00 99.75 10:35:39 AM...

And here is a simple graph:  It seems that iowait is not the issue here.

A few words on the environment I test on: 1. This is a VM (KVM) with 20GB of RAM, 4vCPU (Intel(R) Xeon(R) CPU E5-2698 v3 @ 2.30GHz) 2. One virtio...

Now looking from the Docker storage driver used: ### xfs ``` [root@performance docker-scripts]# time python docker_scripts/cli.py squash -f 7427c9af1454 -t image:squashed image:latest 2> /dev/null real 0m33.035s user 0m15.463s sys 0m4.497s...

It seems that PR #68 can increase performance by up to 40%, depending on the case. This is huge and already available in master.

Comparison of squashing the same image before and after merging #68 and #69. Before:  After:  Please note that the 2nd run took only 22 seconds compared...

Here is an annotated version of the CPU usage graph that shows what's going on. Orange blocks are CPU cycles used by Docker daemon which is very hard to improve...

## Saving image This focuses on the "saving image" process highlighted in above's graph (first orange field). ### Logs This is log from Docker daemon. It shows what happens when...

## Unpacking tar archive Investigation of the dark blue part of the graph. ### Logs ``` 2016-03-30 13:34:54,399 root INFO Unpacking /tmp/docker-squash-E0Up8C/old/image.tar tar file to /tmp/docker-squash-E0Up8C/old directory 2016-03-30 13:34:54,786 root...

@rindeal I'll take a closer look on this at some point. I cannot really promise any dates though. Performance is an important thing for this tool. I think it does...