Slashdot Mirror
The Linux Foundation Forms Open Source Effort To Advance IO Services (linuxfoundation.org)
The Linux Foundation is announcing FD.io ("Fido"), a Linux Foundation Project. FD.io is an open source project to provide an IO services framework for the next wave of network and storage software. Early support for FD.io comes from founding members 6WIND, Brocade, Cavium, Cisco, Comcast, Ericsson, Huawei, Inocybe Technologies, Intel Corporation, Mesophere, Metaswitch Networks (Project Calico), PLUMgrid and Red Hat.
Architected as a collection of sub-projects, FD.io provides a modular, extensible user space IO services framework that supports rapid development of high-throughput, low-latency and resource-efficient IO services. The design of FD.io is hardware, kernel, and deployment (bare metal, VM, container) agnostic.
Architected as a collection of sub-projects, FD.io provides a modular, extensible user space IO services framework that supports rapid development of high-throughput, low-latency and resource-efficient IO services. The design of FD.io is hardware, kernel, and deployment (bare metal, VM, container) agnostic.
GREAT... Just great... This will surely not cause the least bit of confusion at all with the existing FIDO Alliance. https://fidoalliance.org/
BUGS
POSIX requires that opening a file with the O_APPEND flag should have no affect on the location at which pwrite() writes data. However, on Linux, if a file is opened with O_APPEND, pwrite() appends data to the end of the file, regardless of the value of offset.
Or is that too high a standard?
You don't have any Juniper or Cisco gear at any of your "mission-critical" sites?
If you do, there's plenty of Linux running underneath.
Pain is merely failure leaving the body
How about a systemd alliance, might help to unite systemd supporters into a single powerful group intent on ruining the founding foundations of Unix as well as a vast chunk of the Linux user community. They can do more damage to the reputation of Linux if they work together.
Microsoft has a strong reason to ensure performance and stability in their products: recurring revenue. Otherwise common sense dictates they would have gone out of business a long time ago.
Hobby software like Linux doesn't have that priority so it has no place in my mission-critical installations. But it's certainly fine for browsing, email and playing games.
No shit, Sherlock.
Linux file system corruption questions on Stack Exchange.
FD.io ("Fido")
That's not how that works, in any language. Just call it FDIO, or go register fi.do which is available if someone is willing to pay up.
Hobby software like Linux doesn't have that priority so it has no place in my mission-critical installations. But it's certainly fine for browsing, email and playing games.
Exactly! That's why the Microsoft's Hotmail servers ran BSD for so long. Because Windows wasn't fine for email.
See subject: REMOVE ANTIQUATED FILESYSTEMS THAT ADD UNCESSARY OVERHEADS DESIGNED FOR HDD SECTORING!
Khan: "Oh, there has been technical advancement, but how LITTLE MAN HIMSELF has changed - Improve Man, you gain a 1,000-fold..."
* Same goes with filesystems, MAN (pun-intended)...
APK
P.S.=> Test that one out in practice & read up on IRON FILESYSTEMS if you get a chance too... apk
Ummm, if you don't like a supported file system, you don't have to use it.
And I'm not exactly a Linux fan. Standard support is little different than Microsoft, with GNU extensions replacing MS's "embrace, extend, extinguish"; the VM system gets permanently brain-damaged if you drive it into hard swapping; the "try to make it work instead of failing in strict accordance with standards" makes fielding reliable Linux systems of systems harder than it has to be.
But damn, that's about the worst critique of Linux that I've ever seen then feigned to be serious.
FD.io ummm wouldnt that sound more like 'Video' than 'Fido'?
... ducks
http://saveie6.com/
Architected as a collection of sub-projects, FD.io provides a modular, extensible user space IO services framework that supports rapid development of high-throughput, low-latency and resource-efficient IO services. The design of FD.io is hardware, kernel, and deployment (bare metal, VM, container) agnostic.
How about learning to do IO without C++ streams to hold your pee-pee? Or anything like that?
Doing high-speed IO is no different than figuring out how to get the best performance from ANY hardware. You can't do it while ignoring the hardware. And abstracting away everything hides that.
It's really simple: you can't run any hardware anywhere near its design limits if you ignore the design.
See subject, my last post & spellcheck UNCESSARY = unecessary
APK
P.S.=> Haven't seen one of those around here in awhile BUT before it can go down? There's the interceptor missile vs. that puny brand of TROLL, lol... apk
Last time I looked closely -- which, admittedly, was late 2.6 era -- Linux's VM would get pretty brain-damaged just by having swap enabled in the kernel. I had a system that wanted hard real-time responses, but within 60 seconds of starting, my SCHED_RR app would end up with an 80- or 100-ms spike of latency somewhere in the kernel. There was no swap in the system, but after a week or two of trying everything else I could think of, recompiling the kernel with swap disabled fixed it.
You miss my point: Structures of filesystems are encumbered by things that are legacy to HDD's vs. SSD that are unncessary.
APK
P.S.=> See my subject - it will do that for it... apk
The virtual file system layer in Linux imposes no limits on the underlying hardware - it's implemented rather high up in the kernel. And any implementation is free do what it wants under the hood.
And given how deeply the virtual file system is embedded into accessing data, good luck with any replacement.
NOT about swap/page etc. -> http://linux.slashdot.org/comm...
APK
P.S.=> It's about latency reduction beyond what hardware alone has managed to achieve in the SLOWEST portion of a computer (which I've been doing since 1991 on a personal computer using both ramdrive software, even writing up my own, & what I call TRUE Solid-State ramdrive cards (Gigabyte IRAM & CENATEK RocketDrive) - it's about LOGICAL & BLOCK DEVICE accessing software alteration (I used the generic term filesystem in that capacity - does this help you understand it better now? It should...)
... apk
Dude, were you dropped on your head immediately at birth? He was commenting about the Linux VM system getting brain-damaged if you dared to actually use swap.
The IRON Filesystem research project from ms research was implemented on linux :
Finally, we design, implement, and evaluate a prototype IRON file system, Linux ixt3, showing that techniques such as in-disk checksumming, replication, and parity greatly enhance file system robustness while incurring minimal time and space overheads.
this is the result of the failure of the Open Group to provide a POSIX standard to do fast file descriptor checking. poll() and select() are absurdly inefficient and just about everyone with a kernel has invented their own faster alternative. your move Open Group!
Anons need not reply. Questions end with a question mark.
Don't forget that DOS started out as a Unix flavor...
Wow. You know, I had completely forgotten that. Please post more of this interesting history.
MS-DOS was more or less a CP/M clone. Unless I misremember, the first version didn't even have folders.
I go to the website fd.io and there is no intro or anything just some cheesy graphs and blurbs about processing arrays of packets with "extreme performance". The announcement itself is equally full of gibberish. I assume this is related to ongoing SDN craziness. What does it actually do? How would it be used? Why would anyone use it?
SDN just gives me a headache.. The complexity of open stack is frankly breathtakingly insane. It is worse than trying to make sense of telephone networks with stacks of two letter words and assorted insanity to make up for architectural mistake of having intelligence in the core with ultimately very little to show for it in return.
I simply don't understood the point of all this virtual networking craziness or what it gets you over flat networks which simply reflect physical reality associated with physical layout and limitations of communications network. I kind of understand why people want to virtualize shit and networks of shit for logical reasons because it is easier to deal with I guess....but ultimately architecturally like virtual machines ... seems like the same old story playing out again and again... don't fix anything just add a layer of indirection and don't actually address problems.
What a Retard Would Say for $500 Alex. And the answer is RabidReindeer
See subject: Your BULLSHIT has nothing to do w/ it & neither did his crap in reply to it -> http://linux.slashdot.org/comm...
* So, "SHOO", immature, illiterate & IMBECILIC little troll(s), ok?
APK
P.S.=> Moron(s) - obviously SAME troll both times - Didn't even GET what I was speaking of, that's how FUCKING DUMB they/he are, lol... apk
Current filesystems on OS account for HDD legacy structures that SSD don't need: CAN YOU READ? No! Read again -> http://linux.slashdot.org/comm... & when you PULL out that legacy stuff for HDD's & design a filesystem for SSD minus that crap, it WILL go faster... improving I/O!
APK
P.S.=> Honestly - how FUCKING DUMB are you imbeciles that you don't get my 1st sentence? apk
See subject: REMOVE ANTIQUATED FILESYSTEMS ADDING UNECESSARY OVERHEADS DESIGNED FOR HDD SECTORS/CLUSTERS IN THE LOGICAL & BLOCK DEVICE FILESYSTEM SOFTWARE'S ABSTRACTIONS FOR OPERATING THEM.
I.E.-> Structures of filesystems are encumbered by things that are legacy to HDD's vs. SSD that are unncessary. Remove them? The software itself involved will be faster...
Khan: "Oh, there has been technical advancement, but how LITTLE MAN HIMSELF has changed - Improve Man, you gain a 1,000-fold..."
* Same goes with filesystems, MAN (pun-intended)...
(Test that one out in practice & read up on IRON FILESYSTEMS if you get a chance too!)
APK
P.S.=> REPOSTING TO AVOID BOGUS DOWNMODS BY IMBECILIC TROLLS EARLIER THAT COULDN'T EVEN UNDERSTAND WHAT I MEANT (it was hilarious here on downward though-> http://linux.slashdot.org/comm... )... apk
You don't have any Juniper or Cisco gear at any of your "mission-critical" sites?
Juniper uses FreeBSD & Cisco use their own proprietary operating system called ios (yup, same name as apple's mobile O/S)
"Juniper uses FreeBSD & Cisco use their own proprietary operating system called ios (yup, same name as apple's mobile O/S)"
Cisco has a number of "operating systems":
- IOS, used on older router platforms and Catalyst switches (which are now limited mostly to use as access switches)
- IOS-XR which runs on high-end routers (CRS, ASR9K, C12K), which is based on QNX
-IOS-XE which runs on current entry-level to mid-range routers (ASR1K), which is Linux-based
-NX-OS which powers most current Cisco data-centre offerings (Nexus), which is Linux-based
If you have Cisco equipment and no Linux, your equipment is most likely all EOL or very close to it.
Amazing, not a single post gets displayed, because they are all below Score 1.
'BeauHD 'sounds like a fucking genius; so amazing s/he doesn't even reveal her/his account.
The new Slashdot pwners must be so proud.
What things would those be? SSDs have logical blocks, which are analogous to, and often larger than, hard disk sectors. If you don't allocate on logical block boundaries you'll wear through the SSD's write endurance far faster. The filesystem itself still needs to manage allocations and free space in some reasonable-sized unit. You're not going to be able to remove the need for "sectoring" from either direction.
http://www.top500.org/statisti...
That's some hobby.
98.8% of the top 500 supercomputers run linux. The other six run unix. Not on the list: Windows.
Have you guys considered this could be timothy saying: "Discuss."?
http://cs.rochester.edu/~sandh... PAY ATTENTION to caching, clustering, & optimized for SSD hardware
APK
P.S.=> That's just the TIP of the iceberg too from the University of Rochester - searching "SSD future filesystems" gets you more to think about... apk
Per my subject: NTFS can be sped up by reducing writes it performs by default (not best most exacting solid example but point's here) via:
Windows Registry Editor Version 5.00
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\FileSystem]
"NtfsDisable8dot3NameCreation"=dword:00000000
"NtfsEncryptionService"="Efs"
"Win31FileSystem"=dword:00000000
"Win95TruncatedExtensions"=dword:00000001
(DOING THAT STOPS A LOT OF WORK PER FILE UNIT BEING DONE, ESSENTIALLY SPEEDING UP PROCESSING BY OMITTING THOSE TASKS...)
* Now, that alongside the article for FUTURE SSD FILESYSTEMS from the University of Rochester kept in mind, the hardware abstraction level layers is where a LOT of this can be done (as well as block & logical filesystem drivers level) DETERMINING WHAT THE DIFFERENCES ARE IN THE HARDWARE (clusters/sectors geometry is a BIG one HDD's need, SSD really doesn't for the most part other than emulation of HDD for current antiquated filesystems) - REMOVING, as the saying goes, "LEGACY CRUFT"...
APK
P.S.=> Generating Performance Counters by default for NO NEED (analysis) is another such example of useless overheads that unless you NEED them going on? Are "dragging performance DOWN"... apk
Wow, down-voted twice.
Looks like MIcrosoft just got off work...
~ People that think they are better than anyone else for any reason are the cause of all the strife in the world.
What's IO?