What do you agree about large clusters? That they run a stripped down and tailor made Linux kernel, which is not a normal Linux anymore?
Originally Posted by Lennie
Does it all matter ?
I was just showing that btrfs isn't (yet?) faster in every environment, like the author of the article almost suggested.
I doubt I'll be using (Open)Solaris anytime soon, it's interresting but Oracle pretty much closed the door. I doubt I'll be buying from them either.
So I'm just waiting for btrfs and ceph (distributed filesystem) to improve enough so I can start to use them in a production environment.
Hopefully I can use FS-cache/cache-FS as well (second level cache for Linux filesystems).
Many companies are already involved with btrfs development and I think this will only increase.
If they improve enough, which their is atleast a lot of potential because the Linux-kernel community is so vast, I doubt their will be anything/much which is faster on the same hardware.
For example Google currently has their own (Google) 'filesystem', which just distributes and stores blocks, ceph already has an API for doing blocks. What if Google thought they possible want to use that on the long run and started testing it and adding patches.
Then speed (of development) could increase quiet a bit.
Maybe soon people will start deploying many-core lowpower ARM-based storage servers running Linux with btrfs and ceph. With large HDD's and fast SSD second level cache. Who knows.
Tags for this Thread