Tweet
So I just got around to installing/testing linux-mainline 4.14rc8. The experience is stable, but I usually don't go to the RC channel. I did so this time to test out ZSTD on my filesystems. So I switched my compress=lzo mount option for btrfs to compress=zstd, and began to defrag the drives to compress all of the existing files. Here are the results:
Samsung 950 pro 512 GB: OS Drive
LZO: 132.2 GB Free
ZSTD: 160.0 GB Free
Seagate 6TB: Videos (mostly HEVC)
LZO: 4.1 TB Free
ZSTD: 4.1 TB Free
Seagate 6TB: Storage; Mixed video, music, and text files
LZO: 2.3 TB Used, 3.1 TB Free
ZSTD: 3.4 TB Free
Seagate 4TB Raid 1: Backup of OS drive, Code Repository, Old Packages
LZO: 2.1 TB Free
ZSTD: 2.7 TB Free
Overall, I think these are excellent results. The only thing that didn't seem to compress well was the HEVC video files, but the other savings are massive. While having 6 defrags running simultaneously, my 1700 was at approximately 25% CPU usage. Additionally, according to benchmarks, the decompression speeds are even faster than LZO. As such, once it is tested for long term stability, I recommend people switching their BTRFS transparent compression algorithm.
Samsung 950 pro 512 GB: OS Drive
LZO: 132.2 GB Free
ZSTD: 160.0 GB Free
Seagate 6TB: Videos (mostly HEVC)
LZO: 4.1 TB Free
ZSTD: 4.1 TB Free
Seagate 6TB: Storage; Mixed video, music, and text files
LZO: 2.3 TB Used, 3.1 TB Free
ZSTD: 3.4 TB Free
Seagate 4TB Raid 1: Backup of OS drive, Code Repository, Old Packages
LZO: 2.1 TB Free
ZSTD: 2.7 TB Free
Overall, I think these are excellent results. The only thing that didn't seem to compress well was the HEVC video files, but the other savings are massive. While having 6 defrags running simultaneously, my 1700 was at approximately 25% CPU usage. Additionally, according to benchmarks, the decompression speeds are even faster than LZO. As such, once it is tested for long term stability, I recommend people switching their BTRFS transparent compression algorithm.