>Oh god, the NTFS code is a purple opium-fueled Victorian horror novel that uses global recursive locks and SEH for flow control. Let’s write ReFs instead. (And hey, let’s start by copying and pasting the NTFS source code and removing half the features! Then let’s add checksums, because checksums are cool, right, and now with checksums we’re just as good as ZFS? Right? And who needs quotas anyway?)
Daily reminder: Do not put your shit on NTFS
Other urls found in this thread:
xfs.org
twitter.com
what then
No problemo familia, FAT16 all the way.
>filesystems
ZFS, BTRFS, XFS, EXT*, F2FS...
BTRFS is dead, it will be replaced with BCACHEFS in time
Pro tip: In practice, NTFS is one of the most reliable file system with possibly only ZFS or XFS being more objectively reliable in most usage scenarios.
Microsoft employs many level 99 wizards whose sole job is to maintain NTFS reliability while slowly adding on features onto a file system that's now 24 years old (they are close to the breaking point and made a new file system called ReFS).
The fact that NTFS is able to limp along on piece of shit $200 laptops with bugged chink SATA chipsets randomly dropping SATA commands is a proof it's an software engineering marvel.
Why is it dead?
WinFS. Get on my level, plebians.
>ReFS
>Not CSVFS_ReFS
who gives a shit unless it makes my games faster
Faster file system can make your games load faster.
where do I sign up
>bad internal issues = bad product
it's one of the best general use case filesystems as it stands right now. The code being bad means nothing about the actual product.
>needs to be defraged
>A good thing
Does it even have copy on write?
I guess you could call NTFS "reliable" if you're limiting yourself to Windows.
It's a joke compared to ext4, though. But I guess it's not all that strange since it's newer. A comparison between NTFS and ext2 would probably me more "fair".
>Microsoft employs many level 99 wizards
they need to employ more of them before the pajeet horde takes over the company
ext4 is not nearly as reliable or fast as you think it is.
Because it's the "gotta go fast" filesystem, XFS and ZFS are industrial filesystems for reliability.
>"You've heard if serverless technology"
>"Now introducing diskless technology"
>"Never wrestle with filesystems again"
>2017
>mechanical hard drives
It's not, some random company just stopped using it and haters use that as evidence that it's dying.
ZFS is difficult to set up.
BTFRS is new and maybe unreliable, idk, it seems good on paper at least.
REFS is actually old as fuck and Microsoft is slowly removing it.
I'm stuck with Windows because of work and it is really painful to hoard things on it. I can't afford a server and use Linux on it which sucks. Why is it so hard for Microsoft to release a good filesystem?
Not an argument
That's what I figured
Have any of you had to maintain legacy software? I'm curious.
What does this have to do with anything?
>"not an argument"
>this filesystem performs bad on data tapes but GREAT on modern systems with SSD, must mean it's SHIT :^)
fuck you
>as long as we have devices that ignore the shortcomings of the filesystem it's a good system :^)
Not all systems are modern. Also you didn't answer my CoW question.
your cow question?
I wasn't him and I just noticed some retardation in passing without really paying attention that I wanted to correct
Copy on write
>REFS is actually old as fuck and Microsoft is slowly removing it.
ReFS first appeared in 2012 and is Flash-optimized from ground-up.
>not using the god of filesystems: EXT4
>not using RedSea
>REFS is actually old as fuck and Microsoft is slowly removing it.
they're just moving it to another SKU, and if you're using windows for work you should have that SKU come next week
The graybeard wizards are still the ones that do all the deep under the hood shit, the retards are still in charge for the UI and other modern bullshit. Ever since windows 8 they have lighted up the kernel without breaking everything, atleast in my experience.
ReFS will still work on any Win10 version, there just moving the creation of ReFS drives to another sku (Enterprise and by extension Education will still have it). Existing drives will still work.
They've been making that shit modular since what, Vista? Those level 99 wizards have been chipping away at this absolute fucking beast for longer than most posters on Sup Forums have been alive.
Also the additional SKU will be called Pro for Workstations.
Vista actually had alot of issues with system requirements, stability aside it was heavily bloated and needed a good system to run smooth. OEMs sold underpowered systems at the time which didn't help.
Windows 7 came and had the same system requirement issues but fixed the stability and by then OEMs caught up and started selling better systems.
Windows 8 came during the tablet era which again where under powered that is when they started to shave down the system so they could sell cheap atom shit with 1gb of ram and still be somewhat usable
What's a SKU and why isn't REFS replacing NTFS? Why did Microsoft remove it from common operative systems?
Probably because a number of third party crap would break without NTFS for whatever reason.
SKU is just an ID to differentiate between products. (think Windows 10 Home and Windows 10 Pro)
ReFS is being removed from Windows 10 Pro and Home and being put into a new SKU called Windows 10 Pro for Workstations.
>why isn't REFS replacing NTFS?
No idea, I can only guess they haven't made ReFS bootable as a system drive that and it isn't ready yet
sku is mainly a term utilized in distribution which essentially can be used to refer to different product lines
like the iphone x is a different sku than the iphone 8
because you can't boot from refs, it's not meant as a 1:1 ntfs replacement, will that change in the distant future? perhaps, but for the time being it will predominantly be used for excessively large scale data storage
>most reliable
>Metadata gets damaged on the folder structure
>Destroys the entire winsxs directory
Yeah, real reliable.
You have to understand that Wintoddlers hold a very different standard on reliability.
>t. doesn't even understand what a file system is other than files being involved
KEK
>MS comes up with good shit
>decides to stop working on it
Every time for every product of theirs.
>ZFS is difficult to set up
zpool create youareretarded disk1 disk2
>ZFS is hard
not if you use BSD (or some Linux that makes it the default file system)
>pls wait (eta 4 hours for 3TB pool)
See: OG WHS and one of the few good ideas they ever had (disk-pooling)
WHS disk pooling kept losing data. They realized they can't do reliable disk pooling through volume manager layer without NTFS internals natively supporting disk pooling-related metadata/checksum to tracking.
I understand that, but why not take the plunge then and ditch NTFS?
Bcache and btrfs are nothing alike. And, if anything, btrfs is better now than it has ever been.
That's what they did with ReFS.
Isn't bcache another new attempt to use SSDs as a cache for spinning rust?
Yes, and they dropped the indiscriminate disk pooling that WHS fans loved. I know it's becoming a moot point with SSDs taking over, but still...
Yes, and it works really well, HDDs get quite a big speedup especially in IOPS
Though if you're shuffling important data I wouldn't add in another point of failure to my pool, bcache is pretty new.
It takes a few seconds bro.
Who's dropped olschool partitions from their Linux installs? LVM kicks ass.
Its good when you have nothing but compared to ZFS which doesn't even need partitions at all it seems just as rustic as MBR partitions.
I'm not planning on running BSD on my everyday laptop, firewall, seedbox, mail server and NAS, yes, but not my daily laptop.
Start small and increase partitions as needed. I doubt you are making use of that 40GB filesystem.
How is Opensuse? Coming from Fedora, would prefer some more stability without going for the apt/deb/canonical cancer
I have room to shrink or grow, on demand I guess.
Snapshots can get pretty hungry if forgotten.
Don't know about btrfs but just remember you can't shrink XFS.
I normally use this setup:
/ - 8-16GB FS: noatime,nodiratime
/usr - 16GB: nodev,noatime,nodiratime
/var - 8GB: nodev,nosuid,noexec,noatime,nodiratime
/var/log - 8GB: nodev,nosuid,noexec,noatime,nodiratime
/opt - 512M: nodev,nosuid,noatime,nodiratime
/home - 8GB+: nodev,nosuid,noexec,noatime,nodiratime
/tmp (just mount from tmpfs): nodev,nosuid,noexec,noatime,nodiratime
>turtles go fast when thrown from a building
>that must mean turtles run fast
>you can't shrink XFS.
Is that feature ever gonna be implemented?
pour one out for head nigga
I haven't touched fstab options ever since I got SSDs, any point nowadays?
Its kinda there but I wouldn't use it. I don't even trust ext3-4 shrink capabilities.
xfs.org
Security.
>/var - 8GB: nodev,nosuid,noexec,noatime,nodiratime
>/var/log - 8GB: nodev,nosuid,noexec,noatime,nodiratime
Why both?
A 100% /var/log because of a rogue application or DoS won't invalidate your /var.
Also some distros may want to exec or place devices on your /var for whatever reason. I think Debian does. In that case you can remove the noexec flag from /var and still keep execs forbidden on /var/log.
Isn't /var/log where the journal is? How can you even read a binary without +x
> Isn't /var/log where the journal is?
You mean the filesystem journal? AFAIK its stored internally within each filesystem.
> How can you even read a binary without +x
You don't need to to the executable bit to read data even binaries. The executable bit is meant to run programs or scripts. Thats why I don't set it on / and /usr.
I also recommend a 200-500MB separate /boot partition with nodev,nosuid,noexec,noatime,nodiratime which can also be read only or no auto. Just remember to mount it rw when updating kernels (apt can do it automatically for you if you set the right parameters in /etc/apt.conf.d/).
So Microshaft is pulling an apple?
>And who needs quotas anyway?
Uh you don't need quotas when you can create datasets and specify a limit to them. ZFS doesn't have a quota in its classic sense either (set quotas for groups and users). The quota properties are there to limit the size of the dataset.
>bugged chink SATA chipsets randomly dropping SATA commands
The Intel ICH does that too back on my core 2. I had to use generic Microsoft drivers instead of the Intel ones for it to not decide to stall on a read command randomly for 5 minutes whenever it felt like doing so.
How do fucking thin pools and thin provisioning work? What's the point of making a partition that's physically smaller or bigger than what's indicated..?
Some file systems come to a crawl when low on space. I least I think I remember reading that.
>customer:
GIBBE SOME 600GB I NEED TO BACKUP MY WORD FILES
>you:
(why does this retard need so much space for text files, I'll give him half and use the other half on some other moron, I can always enlarge it, not like he'd even use half of that)
Sure dear go- I mean customer.
I think it had something to do with that, it's strictly a enterprise feature.
That sounds mighty jewish.
>This is a feature that is common in most specialist storage hardware and is also part of LVM2, Creating LVM thinly provisioned volumes allow more space to be provisioned than may actually exist on the system. Although this may sound mad it can be beneficial. It is a human trait that we will ask for more than we need. This is true for storage space requirements. So, we may ask for 15TB of data but only actually every use 3TB. Using Thin Provisioning in LVM2 will allow for use to deploy volumes with more space than actually exists on the system. This keeps are users happy. We must make sure, though, that we adequately monitor the system to stop that undersupply becoming an issue.
Jews are crafty beings.
This is like paying for 4GB of VRAM but getting 3.5GB
>One fuckhueg bad sector scrapes your MFT
Have fun~
>bad sectors
Why aren't you using an SSD, user?
Bad sectors's got nothing to do with physical media you fucking idiot, it's all on the block level
A sector is a rotational media specific thing, SSDs don't spin so they don't have sectors to go bad. They just fail all at once.
....
Don't fall into that false sense of security user.
SSDs fail. I've used a shitton of drives in my life and I've killed 4 of the things.
Each one died fairly brutally compared to most HDDs.
HDDs are quite recoverable as you can retry mechanical parts with ddrescue.
SSDs are fucked from every vector once you drop sectors.
One even stopped reading/ writing at all past a certain size.
It would just hang until it crashed and forcibly ejected from the list of available disks.
>SSDs don't have sectors
Most semi-modern SSD firmware automatically detects bad sectors and reallocates them so it isn't suddenly in the middle of your fucking libc
I still having a big fat red "FAILING" in SMART data on my 7 year old HDD.
It's been there for the last 2 years.
As far as I'm concerned SMART data is just a load of horseshit.
Let me give it to you straight.
The feature you're referencing is exactly what my screenshot portrays.
>Sector goes bad
>Data is irrevokeably lost
>Absolutely gone and you're going to love how the system deals with it...
>SSD does its damndest to pretend that sector doesn't exist and stalls the OS when it requests it, sweating bullets the entire time as it fumbles to retrieve it
>On a rare stroke of luck the OS decides to write something to disk without first reading what is there
>The SSD promptly marks the sector as dead and puts the new data somewhere else, assuming what was there doesn't matter because it's been overwritten.
>All other cases result in data destruction
Modern SSDs are TLC garbage, born out of the necessity of MLC shrinks, and before SLC when flash was actually good.
Don't even get me started on how firmware can shit the bed.
I had checksum errors on my initramfs.
You have to actually read SMART, not look at the pretty colors.
It lost a certain number of sectors but luckily I had it in a ZFS mirror and just had to rebuild my damaged /boot and put it on a USB.
BTRFS is alive, it's just that features Facebook doesn't use like RAID56 or encryption have been mostly ignored.
RAID56 is officially "mostly works" now and Oracle Linux has started to support it.
>purple opium fueled victorian horror novel
sounds pretty cool desu, any reccs?