Originally posted by cen1
View Post
Announcement
Collapse
No announcement yet.
ZFS On Linux Landing Workaround For Linux 5.0 Kernel Support
Collapse
X
-
- Likes 4
-
Originally posted by oiaohm View Posthttps://github.com/zfsonlinux/zfs/wi...e-requirements Having a starting requirement of 8G of memory is not good.
- 8GB+ of memory for the best performance. It's perfectly possible to run with 2GB or less (and people do), but you'll need more if using deduplication.
- Likes 3
Comment
-
Originally posted by Niarbeht View PostSo, you can run ZFS on a system with 2GB or less of RAM, but if you want the best performance, get more, especially if you want to do deduplication. That's a far cry from a "starting requirement" there.
Problem is 2G or less you can do deduplication with btrfs or xfs/vdo solutions.
2GB is xfs with vdo(that is with deduplication) for up to 50TB of storage. This is a best performance number. Very much more detailed best performance numbers. ZFS numbers are not great.
Comment
-
Originally posted by Weasel View PostGee, maybe because they could have used the normal (non-SIMD) path without having the functions removed, and yet didn't do it for obvious reasons?
I didn't do benchmarks, but I'm pretty sure the ZFS guys who wrote the SIMD checksumming (or ported it, w/e) did when implementing it. It's there for a reason.
November 13-15 2018, Vancouver, BC The Linux Plumbers Conference is the premier event for developers working at all levels of the plumbing layer and beyond. LPC 2018 will be held November 13-15 in Vancouver, BC, Canada. We are looking forward to seeing you there!
What to do after PREEMPT_RT is accepted into mainline
This means your messing around with the FPU now need to take out a protective lock. The performance profile since the merge of PREEMPT_RT is very different. Yes OpenSolaris/Illumos kernel does not have PREEMPT_RT and this is where ZoL accelerated SIMD check-summing comes from.
Sorry what was obvious reason to-do something under a Solaris/illumos kernel could be totally stupidity on the newer Linux kernel. Yes using SIMD by the FPU instruction is triggering locking that need to be PREEMPT_RT compatible.
Gee they have ported something from a different OS to Linux that may not in fact be compatible any more and it not only because the functions were removed the scheduler inside the Linux kernel now is now nothing like a OpenSolaris/Illumos kernel scheduler.
Basically we need the benchmarks. Linux 5.0 and newer kernels have the PREEMPT_RT stuff as default feature.
- Likes 2
Comment
-
Originally posted by starshipeleven View PostIf this is true then please explain why NO OTHER module inside the linux kernel was using that. The only user for a long while was some UEFI-related infrastructure thing.
Are you saying Torvalds used his thought-police to "convince" them not to? Is everyone but ZOL developers (and you) a moron?
Comment
-
Originally posted by hreindl View PostZOL has nothing in common with Oracle long ago, but yes that all is ridiculous, look at trools like "Weasel" which argues completly of assumptions and "i want", "i believe", "i do not know but they for sure have a reason" and so on
Your bullshit diarrhea on the other hand, which is provably wrong (refer to prior pages about the "implement the checksum themselves" you said) truly qualifies for trolling.
Comment
-
Originally posted by hreindl View Postnope, you pretend they do - prove it or shut up
Originally posted by hreindl View Postyou are a poor enduser claming things backed by shit freshly pulled from his arse
Comment
-
Originally posted by hreindl View Postcome on and show me the massive pacthes of Fedora
the don't exist because they are way to expensive to maintain given the rebase every few weeks and other than you i run it in production from telephony over database and whatever network usecase you can imagine for long enough
Originally posted by hreindl View PostBRUHAHAHAHA "something similar" - just inform youself what clean linux is and how it gains the performacne, mostly by compiler flags and automatic vecorization
You can't use the fucking vectorization instructions without those functions. That's the entire point. YOU CANNOT USE THE INSTRUCTIONS AND SIMD REGISTERS. Compiler generated or not.
Use them and it corrupts usermode or crashes the entire system, that's why those functions were there, to allow you to use them and "auto vectorize" code if you are too retarded to write inline asm.
You're a fucking joke. I'm done with you.
Comment
-
Originally posted by hreindl View Postit's amazing what a braindead idiot you are
https://www.gnu.org/software/gcc/pro...orization.html
i wonder how all the userland software is using vectorization all the time
I know it may be too much for you, but here's a hint: the kernel's job is to task switch and save register state between processes. Usermode code doesn't have to care about it.
The kernel can compromise any usermode code as well or state as it sees fit. As you know, usermode processes are completely separated in address space, so they don't have to worry about this crap. You CANNOT USE vectorization in kernel without proper state saving and restoring.
Like I said, you have ZERO technical knowledge. Retard.
I'm an actual developer not a piece of shit "IT guy" go take a piss with your shit "admin" job you clueless clown.
Comment
Comment