It’s PCIe gen 2, so it won’t get full performance out of the NVMe drives currently on the market (gen 4 and 5) but it should still be a nice improvement on a mechanical HDD.
1 Like
In fact, Btrfs handles SMR/Zoned devices quite well, better than ext4, as far as I know.
There are reports of users seeing improved performance with SMR drives compared to ext4, so in this case I wouldn’t blame the SMR + Btrfs combination.
Fedora enables compression by default, and with a decent CPU you should see some benefits from it, since in this case the bottleneck is the hard drive, at the cost of slightly higher CPU usage.
As for wear, compression also helps here, as it reduces both read and write operations (writing fewer data means fewer head movements and fewer sectors being allocated).
1 Like
I would! For any device that is more than about 30% used.
I personally have seen the degradation of performance of the HDD that uses SMR tech (with ext4). After reaching about 40% - 50% usage the speed drastically dropped and became abysmal.
Yes, the drive is the bottleneck, and SMR simply makes that bottleneck slower as the drive gets more data written. All the reads & writes for the shingled layers are handled within the drive itself and not by the OS.
The extra writes done by BTRFS when using CoW can only add to the degradation in spite of compression.
While BTRFS may be a good file system, the CoW and SMR can only aggravate the already abysmal performance of SMR devices once they reach about 1/3 of capacity. Using SMR devices in a “write once – read mostly” config is acceptable to most users. Using the same device for constant writes becomes totally unacceptable for performance.