What is Digg?82 Comments
- aflury, on 10/11/2007, -2/+17"Linux records the times when files were created, last modified, and last accessed. The latter usually implies a penalty on file access, since even if you only read a file, the system will update the directory entry for the file to record the latest timestamp."
Ok so I'm pedantic... but this is false. The access time is stored in the inode, not the directory entry. - nyx210, on 10/11/2007, -1/+14There are other Linux sites that can give you better help than the answers that you'll probably get on Digg.
- stormgren, on 10/11/2007, -0/+13"Dont get me wrong, I am an advocate of Linux and Open Sauce"
Mmmmm, open sauce. - forceflow2, on 10/11/2007, -5/+17How did a story about Linux get to the front page without having Ubuntu in the title? Someone needs to record this moment in time.
- leszek, on 10/11/2007, -0/+11All recent distributions automatically enable dma ...
This tip is unnecessary. - 4DFX, on 10/11/2007, -2/+11That was like going to Baghdad and yelling: "ALLAH SUX ASS!"
- Megatog615, on 10/11/2007, -2/+9Dugg for not being called "A Three Pronged Attack to Increase Ubuntu Performance".
- jacekpoplawski, on 10/11/2007, -1/+7Nothing really new.
DMA is a classic example of "optimization", but truth is it is always on.
It's off only if you don't have chipset support in your kernel, in that case you should try to upgrade/recompile your kernel, hdparm won't help you. - Chandon, on 10/11/2007, -1/+7That's a great idea. He should definitely go do that.
- netkid91, on 10/11/2007, -0/+5Ummm, you can't pipe output from CP.... How about: cat /home/user/reallybigfile > /dev/null
- Chandon, on 10/11/2007, -1/+6skidme -
That's great. But... I'm not sure why anyone who would be in this thread would care. - malkir, on 10/11/2007, -0/+5Different != less usable. At this point any usability issues arise from a lack of familiarity. I personally can't think of specific tasks that an average user would need to do on a computer that can't be done just as easily on linux as under windows. This is, of course, assuming someone set it up properly in the first place.
Also about the performance issues. The hard-drive is the biggest bottleneck in performance for PCs if you can get an 11x performance increase out of it you're doing really well. - Timmmm, on 10/11/2007, -0/+5DMA is on by default in most modern distros. 'noatime' doesn't make that much difference. Didn't read the rest - what a waffly article!
- columbusgeek, on 10/11/2007, -2/+6deleted for me being dumb
- clickwir, on 10/11/2007, -1/+5uhhh no. These software settings don't make the hardware work harder. It simply changes the methods of communication to and from the hard drive. You can't type some commands and make your platters fly away.
Though most modern distros already set these automatically and/or have no need for them because they use different/better methods anyway. This guy is talking about getting 33MB/sec from his drive. I have an old Seagate 120gb SATA and it does 50MB/sec on Kubuntu Feisty with no tweaks. What the hell distro was he really running here? Slackware 4 on a Pentium II 333 with 64MB ram? - OneAndOnlySnob, on 10/11/2007, -1/+5Skidme! Will you switch!!! Please, everyone is dying to know!
- ShadwDrgn, on 10/11/2007, -1/+5uh... you can.
- muszek, on 10/11/2007, -1/+4Just as I wanted to write that I wish there were tips like this for Ubuntu :)
- malkir, on 10/11/2007, -2/+5"As with all optimizations, you won't be able to tell whether you are really getting better results without doing some simple benchmarking." This is a statement of fact no matter what system you are running. Just because you _think_ you are getting better/worse performance doesn't mean you actually are. The only way to know is through benchmarks.
- complich8, on 10/11/2007, -0/+3SATA drives are in DMA by default. You can do other small behavior tweaks with smartctl on some disks, but not on others. Since SATA disks are fast becoming the norm, this article's about 3 years late to the party.
- malkir, on 10/11/2007, -0/+3All of those issues are social issues; people weren't willing to learn, it's not that the software on Linux couldn't handle their business processes. The specialized applications will most likely end up bogging down the organization sooner or later, they typically do. When you need to many macros in excel it's usually time to move to a real database, same goes for access, most access databases usually outgrow themselves. I'm not here dissing Windows, I use it for most of my work and like it just fine, I'm just arguing against what I perceive as flawed arguments.
- idonthack, on 10/11/2007, -1/+4More like Linux apples run faster than Microsoft apples.
Also, oranges rule. - malkir, on 10/11/2007, -0/+3By "set it up properly in the first place". I was talking about usability, not stability. An average user needs their printers, their email, and their companies file share, along with their own files. IT staff have to set this up for Windows users anyway so setting it up for a Linux user would only impact them. What you are basically asking for is for some Linux distribution to copy windows and put all the menus in the same spot. IANAL but I'm betting you could get in some serious trouble for doing that.
The only issues with moving a normal company from Windows to Linux would be educating users about file types. And the occasional program that has no decent replacement in Linux (i.e. Quickbooks). - renegadeafk, on 10/11/2007, -1/+3Is Ubuntu faster than linux?
- complich8, on 10/11/2007, -0/+2DMA is NOT bad for your disk. reading faster means reading in fewer passes, allowing IO operations to be more streamlined. PIO is more likely to pile up conflicting IO operations and keep the disk active, seeking and generally working for substantially longer. It's not like changing how the controller talks to the disk itself is going to change how fast the disk is rotating or how fast the read heads move.
The only reason PIO is the default is that linux supports ancient, ancient hardware that doesn't support DMA. DMA has been normal since like ATA/33 for IDE devices. - Megatog615, on 10/11/2007, -0/+2They don't control the advertising. The ad company just pays them to host the ads.
- ShadwDrgn, on 10/11/2007, -3/+5this is lame. the same hard drive optimizations can be made in XP/Vista, and getting rid of atime info is an idiotic move as security goes. not worth it.
- complich8, on 10/11/2007, -0/+1There are two useful patterns of studies: "within-subjects" and "between-subjects". They're useful for different things.
You're effectively saying "within-subjects designs are useless, only between-subjects matters". But if you've ever sat down at a system with DMA disabled and tried to do anything substantial, you'd be quite well aware of the effects of doing so.
Seriously, PIO mode is horrible. But every distro around turns on DMA by default for IDE disks, as has been said. This has been the norm since the 2.6 kernel came out, and was not uncommon before that.
This article isn't saying "linux is better than windows if you do these things and not otherwise". It's saying "make the most out of linux". But because it's listing what on most systems is a default setting, followed by something that improves access performance by a fraction of a percent in the average case, followed by prelinking (which only makes much difference with huge apps and only at load-time), the article is fluff. - clickwir, on 10/11/2007, -0/+1wow. that has nothing to do with anything in this article
- redxii, on 10/11/2007, -2/+3rm -Rf /
- bjnord, on 10/11/2007, -0/+1aflury is correct (inode, not directory entry), but hagnar: Setting noatime when you don't need last access time does indeed give you a significant performance boost. I remember the FreeBSD folks recommending this to speed up "make buildworld" back in the day, and they were right.
- forceflow2, on 10/11/2007, -1/+2Can't decide if this is sarcasm or not...
- inactive, on 10/11/2007, -0/+1I'm sure you'll steal the article, change the title to "A Three Pronged Attack to Increase ubuntu Performance, and post it in your blog.
- osirisx01, on 10/11/2007, -3/+4To me, tip one implies that linux cripples the hard drive, and pretty badly considering that the read speed gains a 10x boost just from changing a few settings.
- inactive, on 10/11/2007, -0/+1Also SATA drives have the improvements on (hdparm is IDE only).
Is IDE still sold much? - PacketBoy, on 10/11/2007, -1/+2Three Pronged Attack - I was thinking:
SubLiminal, Liminal, and SuperLiminal - Megatog615, on 10/11/2007, -0/+1The readahead tweak gave my hard drive about .5MB/sec increase. The multcount tweak gave me nothing, and it's also potentially dangerous to use. Other than that, all the documented tweaks were already enabled for me. Maybe someone else will have better luck with the readahead tweak.
- Sparkster185, on 10/11/2007, -0/+1I work for a company that receives lots of grants from the government. All the software we create as a result of these grants is all free/open. So I guess, in a way, I work for an open-source company.
- drmsux, on 10/11/2007, -0/+1noatime:
fsutil behavior set disablelastaccess 1
"getting rid of atime info is an idiotic move as security goes."
lastaccesstime does not have ANY impact on security - (at least on NT, where only the timestamp is stored, and not user SID, and there is a SetFileTime API to set all file timestamps, including last access), you'd want to use audit (that Linux does not have by default?) for that - TruckStuff, on 10/11/2007, -0/+1Desktop vs Server is irrelevant. If you have an intrusion, every piece of information you can gather is helpful in finding the source of the intrustion. Mtimes are only helpful if a file has changed, but any blackhat worth his salt is going to cover that up. While he might be smart enough to adjust the clock to account for the mtime of /etc/passwd, he is going to leave atime tracks on /etc or /root as he is leaving after he has put the clock back.
You +1 clue. - Sdiggmatism, on 10/11/2007, -0/+0Kubuntu 7.04 i815 P3-933 all drives are /dev/sd*
WD80 PATA 53MB/sec
Hitachi250 PATA 43MB/sec
Seagate300-USB2 20MB/sec
What Linux does he have that he doesn't have wicked sick performance by default? Showing how to use 'init 1' and 'hdparm -t' are highly useful but the rest of the article seems more appropriate for a Linux long ago before the aggressive performance options were well tested enough to enable them automatically. - credence, on 10/11/2007, -0/+0rm -vrf / is more fun, then you can watch the files on your harddisk as they vanish forever.
- Stoical, on 10/11/2007, -11/+11Yep, some useful stuff, although i always wince a little when i read: "Linux can outperform the same computer running Windows XP or Vista" While i agree to a certain extent i still feel that XP and Vista offer big advantages over Linux. The main one being useability. If your a switched on Power user then getting the most out of linux is great, but for your average Joe XP and Vista are superior because they are simply easier to use and more user friendly.
Comments like this also make me wonder: "As with all optimizations, you won't be able to tell whether you are really getting better results without doing some simple benchmarking." So is this suggesting that under normal use i wont notice the optimizations? I mean if your goal is to beat last weeks benchmark results then cool, but surely you would be using your OS for more productive tasks, and if an optimization doesnt provide a noticeably better performance increase is it worth the effort?? Dont get me wrong, I am an advocate of Linux and Open Sauce, but i prefer to use Linux for the tasks that it excels, my back end infrastructure i.e. my servers.
Until Linux can match Windows for pure useability from an average Joe's perpective i will continue to use Windows boxes for my users. The latest generation of Linux Desktop distributions have made a giant leap forward, but have some way to go. I look forward to the day Linux can match MS for ease of use, it wont be long i'm sure. Good tips for the power user though. - Chandon, on 10/11/2007, -1/+1> Yes Vista and XP have many more apps and more collaboration out there but Linux out there like Ubuntu is pretty easy to use.
There are some specific proprietary apps that Linux doesn't have, but I wouldn't go so far as to say that Windows has more apps - at least not in any method of counting that will produce useful results. - complich8, on 10/11/2007, -1/+1I don't get why you'd think noatime is a security risk. From an intrusion detection perspective, you only really care when a file was created and when it's modified. CTIME and MTIME give you that. Just about nothing uses ATIME for anything...
but you're right. In fact, XP and Vista both default to DMA enabled, with XP only going out of DMA if you've got a controller that has bad support in the default windows driver and no better driver installed. The other tweaks are for the most part pointless. - complich8, on 10/11/2007, -0/+0sure you can pipe output from cp...
like, if you did "cp -v *.txt textfiles/. > list.txt"
it's sort of pointless without a -v though :p.
I like how that command overwrites whatever file is lexically last in the directory though. Makes you wonder if he even actually tried the command himself... - inactive, on 10/11/2007, -2/+2Open source can only sustain itself so much. I always find it odd when people say they "work" for open source companies... Im thinking "as a volunteer?"
- johnnyrotten, on 10/11/2007, -0/+0Interesting how I got dug down and "corrected" when all I did was to quote the article. I think the author meant to say "cat" rather than "cp", but none of you caught it.
- decoherence, on 10/11/2007, -0/+0the script to run the hdparm test should be
for ((i=0;i - jbhannah, on 10/11/2007, -1/+1Triple-boot for teh win.
-
Show 51 - 78 of 78 discussions
Digg is coming to a city (and computer) near you! Check out all the details on our 
© Digg Inc. 2009
— Content created and posted by Digg users is