{UPDATED} HOWTO : Ubuntu 10.04 runs more faster (with demo)

本帖最後由 samiux 於 2010-5-9 21:21 編輯

Hi all,

Ubuntu 10.04 is running quiet fast.  However, you want more speed ....

Here you are.

YouTube demo

Enjoy!

Samiux

you should explain what it does instead of tell other just to follow instructions.

and you may mention "noatime" as well and discuss that "noatime" give you the best performance but it may break some programs that relies on file access time, for instance, mutt

TOP

我覺得唔用 noatime 好啲,唔只mutt,仲有其他問題;如果想加速,但又唔想太麻煩,用 relatime 好少少。
正如 icarus-c 講,好多唔知頭唔知路嘅人,如果只係跟住啲 step一步步做,又唔知其實做緊咩,會好危險...

TOP

本帖最後由 samiux 於 2010-5-8 14:08 編輯
you should explain what it does instead of tell other just to follow instructions.

and you may ment ...
icarus-c 發表於 2010-5-8 10:16


and

我覺得唔用 noatime 好啲,唔只mutt,仲有其他問題;如果想加速,但又唔想太麻煩,用 relatime 好少少。
正 ...
lazyfai 發表於 2010-5-8 13:50


Please read my tutorial carefully.  :funk:

The tutorial is not using noatime, it uses norelatime instead.

My tutorial is not writing for newbies who do not Google at all.

TOP

The problem is not "noatime" or "norelatime", both of them ask kernel don't update access time, to get the balance, use relatime is better AND safer.

TOP

The problem is not "noatime" or "norelatime", both of them ask kernel don't update access time, to g ...
lazyfai 發表於 2010-5-8 14:10


Quote from http://kerneltrap.org/node/14148

About noatime that will affect to mutt like software :
  1. From: Ingo Molnar
  2. To:        Jörn Engel [email blocked]
  3. Subject: Re: per device dirty throttling -v8
  4. Date:        Sat, 4 Aug 2007 21:21:30 +0200


  5. * Jörn Engel [email blocked] wrote:

  6. > > I actually vote for that.  IMO, distros should turn -on- atime
  7. > > updates when they know its needed.
  8. >
  9. > If you mean "relatime" I concur.  "noatime" hurts mutt and others
  10. > while "relatime" has no known problems, afaics.

  11. so ... one app can keep 30,000+ apps hostage?

  12. i use Mutt myself, on such a filesystem:

  13.    /dev/md0 on / type ext3 (rw,noatime,nodiratime,user_xattr)

  14. and i can see no problems, it notices new mails just fine.

  15.         Ingo
複製代碼
Ubuntu packed with noatime and nodiratime :
  1. From: Ingo Molnar
  2. To:        Alan Cox
  3. Subject: Re: per device dirty throttling -v8
  4. Date:        Sun, 5 Aug 2007 09:13:20 +0200


  5. * Alan Cox wrote:

  6. > > > People just need to know about the performance differences - very
  7. > > > few realise its more than a fraction of a percent. I'm sure Gentoo
  8. > > > will use relatime the moment anyone knows its > 5% 8)
  9. > >
  10. > > noatime,nodiratime gave 50% of wall-clock kernel rpm build
  11. > > performance improvement for Dave Jones, on a beefy box. Unless i
  12. > > misunderstood what you meant under 'fraction of a percent' your
  13. > > numbers are _WAY_ off.
  14. >
  15. > What numbers - I didn't quote any performance numbers ?

  16. ok, i misunderstood your "very few realise its more than a fraction of a
  17. percent" sentence, i thought you were saying it's a fraction of a
  18. percent.

  19. Measurements show that noatime helps 20-30% on regular desktop
  20. workloads, easily 50% for kernel builds and much more than that (in
  21. excess of 100%) for file-read-intense workloads. We cannot just walk
  22. past such a _huge_ performance impact so easily without even reacting to
  23. the performance arguments, and i'm happy Ubuntu picked up
  24. noatime,nodiratime and is whipping up the floor with Fedora on the
  25. desktop.

  26. just look at the spontaneous feedback this thread prompted:

  27. | ...For me, I would say 50% is not enough to describe the _visible_
  28. | benefits... Not talking any specific number but past 10sec-1min+
  29. | lagging in X is history, it's gone and I really don't miss it that
  30. | much... :-) Cannot reproduce even a second long delay anymore in
  31. | window focusing under considerable load as it's basically
  32. | instantaneous (I can see that it's loaded but doesn't affect the
  33. | feeling of responsiveness I'm now getting), even on some loads that I
  34. | couldn't previously even dream of... I still can get drawing lag a bit
  35. | by pushing enough stuff to swap but still it's definately quite well
  36. | under control, though rare 1-2 sec spikes in drawing appear due to
  37. | swap loads I think. ...And this is 2.6.21.5 so no fancies ala Ingo's
  38. | CFS or so yet...
  39. |
  40. | ...Thanks about this hint. :-)

  41. much of the hard performance work we put into the kernel and into
  42. userspace is basically masked by the atime stupidity. How many man-years
  43. did it take to implement prelink? It has less of an impact than noatime!
  44. How much effort did we put into smart readahead and bootup
  45. optimizations? It has less of an impact than noatime.

  46.         Ingo
複製代碼
Just for reference.

Samiux

TOP

本帖最後由 icarus-c 於 2010-5-8 17:16 編輯

let me clarify my comment again, the main concern isn't about relatime/norelatime/noatime, etc

1. You should tell people what it does when you tell someone to do something.
# in this case, tell people what is norelatime, atime, etc and the reason behind the tweak

2.  you may (optional) tell people there is  also  "noatime" & "nodiratime" option, explain its advantages and drawbacks

I agree that norelatime is a safer and better option for general users. but people have the rights to know : )

*note that when you set noatime,  it will enforce nodiratime and ignore further nodiratime options.  Refer to  kernel source  touch_atime()  @ fs/inode.c

TOP

let me clarify my comment again, the main concern isn't about relatime/norelatime/noatime, etc

1. Y ...
icarus-c 發表於 2010-5-8 17:14

Your concern is absolutely right! The creator has said "My tutorial is not writing for newbies who do not Google at all." We don't have to argue anymore.

TOP

1. You should tell people what it does when you tell someone to do something.
# in this case, tell people what is norelatime, atime, etc and the reason behind the tweak
icarus-c 發表於 2010-5-8 17:14


First of all, why "should" I tell everybody the technology and/or theory of the tweak?  You said that they have rights to know, but in my opinion, I have rights NOT to tell.  It is freedom.  You have no rights to force me to do so.  Who you are?  My tutor?  My mentor?

The optional mount parameters can be found in the internet very easily by using Google.  Why bother me to tell you?  I am not your tutor or mentor.  If you want to learn something new, you should study yourself or pay someone to teach you.  Not just asking for but not contribution.

Your concern is absolutely right! The creator has said "My tutorial is not writing for newbies who  ...
小小小俗辣 發表於 2010-5-9 05:06


Sure, all my contributions are not aiming for newbies.  In my dictionary, the word "newbies" is targetted for the one who doing nothing (e.g. search by Google) but just asking for everything.

If you have doubt, you have rights not to implement the tweaks.  If you dare to do so, you can do it on an experiment machine to test the performance or the result.  It is your turn.  "LEARNING BY TRYING SOMETHING NEW!" always in the first page of my dictionary.

TOP

其實我覺得大家都只係畀少少意見去令你的教學更加有用,聽唔聽都好,反應不用太大。

TOP