Separate names with a comma.
Discussion in 'Memory and Storage' started by HFS+, Jan 10, 2012.
I go with HFS+ because HFS+ is basically B-trees in every sturcture, while the NTFS MFT is a table.
Umm... it doesn't use B-trees in every structure. The file allocation is still a bitmap. NTFS uses B+-trees for directory content, while HFS+ uses B-trees. HFS+ also has a year 2040 bug which isn't too far away.
If I could do away with NTFS on my Windows systems?.. I'd do it in a heartbeat. NTFS has so many fundamental flaws it's not even funny. The nature of the file system guarantess massive fragmentation of an OS volume by its very design. Unless it's simply used for storage duty and data rarely changes of course.
I'd opt for HFS+ or EXT3 as they can optimize file system layout in much more efficient ways. Now some might argue that SSD's use of randomized data layout will not show the gains to be had by those other file systems?.. I have tested many different structural layouts for NTFS and have even added slack space in between blocks to hopefully bypass/slow the inevitable without much success.
Fact is that contiguous file system layout can and does help SSD performance from the logical side of things as it increases efficiency down the entire stack. Its also more common than some/most will know that SSD controllers/newest firmware take every opportunity possible to write data contiguously at the physical level as well. Within the wear leveling requirements of the drive of course. A little bit here.. and a little bit there.. can add up cumulatively to show measurable and perceivable results in some configs.
Not much time to get thoughts out right now and that's the simplistic view.. but hopefully the points were articulated well enough to get them across.
^Why not EXT4?
I've been debating this recently as well. My storage drives are NTFS because it works fine in Linux and OSX (with Paragon NTFS). But if there's a better file system that works cross platform reasonably well (i.e. can be accessed by Explorer/Finder; Linux can already do it) I'd switch.
Also, my year and a half old Windows 7 install was as fast as day one without me ever touching any defrag tools. I know because I upgraded to Sandy Bridge, reinstalling Windows in the process and it feels exactly the same. Maybe Windows 7 deals with it better or something. Using a 3-4 year old 640GB WD Blue drive.
yeah.. that too. lol
and as for efficiency in the long term?.. well of course many systems can maintain decent performance over time as defragging is all built in and does a fair job at repairing the ever constant damage that we call fragmentation.
Point was that internal and external fragmentation is literally built into the NTFS file system.
As for effects of file system speeds/efficiency and their perceptible differences? Windows caching does such an excellent job.. its almost hard to even feel those losses for some users. However.. benchmarks and file transfers can and do tell another story when we run them on highly fragmented drives(HDD in particular).
Not to start another giant debate and start speaking in tongue here.. but I even see minor losses/gains on SSD's. Splitting pretty small hairs at that point.. but still shows the logical side losses associated with fragmentation. NTFS almost guarantees those losses by design and we've been forcefed that lifespan and gains are non-existent with SSD. I've found otherwise.. or at least the small amount of writes to move files around are worth the lifespan loss and gains to be had, IMO.
HFS+ is more faster than NTFS from a technical stand point.
On a filesystem with millions of files HFS+ is guarenteed to find a file faster than NTFS. HFS+ uses a B-tree (catalog) to sort the file inodes, while in NTFS is must search for the inode linearly.
HFS+ also has a better layout mechanism than NTFS. HFS+ is setup so that all the filesystem strucutres are at the beginning of the disk and HFS+ has a hot file zone which mac os x will move into the metadata zone (beginning of disk) for even faster file loading.
I dont know NTFSs layout Microsoft hasnt released the specs to it.
Of Course HFS+ is built off of HFS which is decades old, while NTFS is only 18 years old?
NTFS. It's been the most reliable FS I've used (though EXT4 has been solid, so far!), and HFS+ sucks in Windows far more than NTFS does in OS X.
NTFS may be slower than HFS+, but usually, performance is lost at a higher level or lower level than the FS, except in heavy server loads. FI, turning indexing off in Win7 with a platter drive will nearly get rid of pauses when multitasking, or trying to do work while copying a bunch of files.
As of Windows 7 and mature 2.6 kernels (including 2.6.40 and up, aka 3.x ), FS choice really only matters for the reasons it aught to: how safe your data will be, when commodity hardware and software fail on you (an area where EXT3 sucked monkey balls, despite great reliability in your typical RHEL clone server with RAID, backup, and a UPS), an area where NTFS has excellent since back in the NT4 days.
Linus says HFS is complete and utter crap so I'm going with NTFS.
NTFS has its warts, but overall it's a damned good filesystem. But now with ZFS (and maybe BTRFS to an extent) going the extra mile with checksumming and such it makes pretty much all other filesystems seem flimsy. If they only would have left out the LVM stuff from ZFS and just let it be a filesystem...