SSD’s, Journaling, and noatime/relatime | Thoughts by Ted
The chief mechanism for benchmarking Linux filesystems seems to be operations on the Linux kernel's source tree.
This seems to me a dangerous over-optimization for kernel developer-centric usage. What if people want to put files other than the Linux kernel source on their disk? Other use cases are completely untested!
OK I'm being facetious.
Dumb questions: Are snapshotted filesystems like btrfs, and version control systems like git two things that could converge, or do they just look like they could converge and really never will?
Comments (4)
A lot of people thought filesystems and databases would converge (WinFS,Reiser4) too. I don't think we will see anything mainstream here for at least another 5-10 years though. Oddly, IBM has pulled this off pretty well since the late '80s with what is currently called 'i OS'.
Mainly I see a COW FS like Btrfs being a compliment to userspace datastores like Git or a DB. Fast, almost free snapshots can be taken at will. Same idea as people using LVM to do hot DB backups.
I guess it all boils down to user space moves faster and is safer than kernel. The usual benefit of convergence is speed and access to raw hardware, but we always have to question the cost.
Posted by Kevin Bowling | March 4, 2009 9:26 PM
Posted on March 4, 2009 21:26
A snapshotting filesystem could be used as a store for a VCS, but wouldn't be equivalent to one
Posted by Matthew W. S. Bell | March 5, 2009 5:53 AM
Posted on March 5, 2009 05:53
If you want to give a filesystem a workout, aggregate lidar point data files and create raster data from the vector files. Also try reprojecting multigigbyte Erdas Imagine format data.
Posted by Doug Newcomb | March 5, 2009 7:31 AM
Posted on March 5, 2009 07:31
1. Not while POSIX rules. If you're storing your data with full versioning, including edit-merges, conflict resolution, and patch generation, then IMHO you've gone far beyond anything that can reasonably be shoehorned into the POSIX model. Maybe the Linux VFS will grow some very impressive new features, or maybe we'll have to wait for microkernels that never arrive.
2. The Sugar user interface (sugarlabs.org), originally developed for OLPC, is designed to do exactly this, by replacing the "file manager" in the UI with a version-aware object storage system that supports this sort of revision control. Currently this sort of versioning exists only in mockups and pre-alphas, so don't hold your breath.
Posted by Ben Schwartz | March 5, 2009 6:29 PM
Posted on March 5, 2009 18:29