Stories
Slash Boxes
Comments

SoylentNews is people

posted by martyb on Thursday November 03 2016, @09:29AM   Printer-friendly
from the those-who-do-not-learn-from-the-past dept.

Submitted via IRC for TheMightyBuzzard

Veteran dev says timed sampling's arrival in Berkeley Packet Filter makes Linux 4.9 a match for Solaris' DTrace

In 2004 former Reg hack Ashlee Vance brought us news of DTrace, a handy addition to Solaris 10 that "gives administrators thousands upon thousands of ways to check on a system's performance and then tweak ....production boxes with minimal system impact". Vance was excited about the code because "it can help fix problems from the kernel level on up to the user level."

Vance's story quoted a chap called Brendan Gregg who enthused about [the] tool after using it and finding "... DTrace has given me a graph of a hundred points that leaves nothing to the imagination. It did more than just help my program, it helped me understand memory allocation so that I can become a better programmer."

[...] As Gregg explains on his blog, Linux has had plenty of tracing tools for a long time, but they were miscellaneous kernel capabilities rather than dedicated tools and didn't match DTrace's full list of functions. But over time developers have worked on further tracing tools and Facebook developer Alexei Starovoitov recently offered up some enhancements to the Linux kernel that Gregg feels mean it now matches DTrace.

Gregg reckons Starovoitov's contribution, plus efforts like the bcc project he's worked on will offer Linux users their best ever chance to conduct really detailed tracing of Linux.

[...] Gregg's post has oodles more detail about DTrace's long history, plus links a-plenty to tools you can use to employ the tool.

Source: http://www.theregister.co.uk/2016/11/01/linux_in_2016_catches_up_to_solaris_from_2004/


Original Submission

This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 1, Insightful) by Anonymous Coward on Thursday November 03 2016, @09:44AM

    by Anonymous Coward on Thursday November 03 2016, @09:44AM (#421984)

    BPFCC is called 'BCC' completely ignoring that there is already a bcc on linux called 'Bruce's C Compiler' which is used for legacy 16 bit executable support.

    Can't people fucking google before arbitrarily fucking things up with overloaded naming for the rest of us?

    • (Score: 0) by Anonymous Coward on Thursday November 03 2016, @12:40PM

      by Anonymous Coward on Thursday November 03 2016, @12:40PM (#422008)

      Neither are tools regularly found on Linux boxes at all, let alone at the same time. You're safe.

      • (Score: 2) by letssee on Thursday November 03 2016, @03:30PM

        by letssee (2537) on Thursday November 03 2016, @03:30PM (#422076)

        You can safely assume the older BCC *is* found on GP's computer :-)

      • (Score: 2) by linuxrocks123 on Thursday November 03 2016, @05:35PM

        by linuxrocks123 (2557) on Thursday November 03 2016, @05:35PM (#422140) Journal

        Umm, yes, bcc is regularly found on Linux boxes. I just checked, and it's on mine. I didn't do anything special to put it there, and in fact have never used it, so it must have been installed by default.

        It's in the dev86 package on Slackware.

  • (Score: 4, Insightful) by FatPhil on Thursday November 03 2016, @10:14AM

    by FatPhil (863) <{pc-soylent} {at} {asdf.fi}> on Thursday November 03 2016, @10:14AM (#421991) Homepage
    I hear that finally Daft Punk used a sample of a banjo.
    --
    Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
  • (Score: 0) by Anonymous Coward on Thursday November 03 2016, @10:38AM

    by Anonymous Coward on Thursday November 03 2016, @10:38AM (#421997)

    Great news! Larry is hiring, I hear...

    • (Score: 0) by Anonymous Coward on Thursday November 03 2016, @02:03PM

      by Anonymous Coward on Thursday November 03 2016, @02:03PM (#422035)

      Bend over and spread your cheeks

  • (Score: 3, Insightful) by Runaway1956 on Thursday November 03 2016, @11:02AM

    by Runaway1956 (2926) Subscriber Badge on Thursday November 03 2016, @11:02AM (#421999) Journal

    TFS and TFA imply that Linux is finally catching up to Solaris, twelve years later. Now, I'm not a Solaris guy, and I've never used dtrace, but the facts as presented just don't add up for me.

    In 2004, dtrace was touted as an "advanced tool". I won't argue with that description of dtrace. But, I can't see where any claims are made that dtrace was then a standard part of the Solaris kernel. It looks like a bolt-on aftermarket sort of thing.

    Now, 12 years later, Linux has incorporated all of that funcitonality into the kernel. That's cool and all, but I haven't seen any press releases announcing the importance, or the indispenability of these functions.

    In short, the author, Gregg, is expressing his own personal feelings about how great dtrace is/was. Which is cool, but I wonder how that becomes a real story.

    • (Score: 4, Informative) by tbuskey on Thursday November 03 2016, @12:11PM

      by tbuskey (6127) on Thursday November 03 2016, @12:11PM (#422002)

      Dtrace *is* part of Solaris 10 and up, not a bolt on. Dtrace is in MacOSX (it has fewer trace points). I think it's in Oracle's Linux as well.

      When you're looking for bottlenecks in your code, you typically attach a debugger to your app. That changes it a bit and slows it down alot. Plus you need to restart the app.

      Dtrace is lightweight and you don't need to restart your app.

      If you're doing performance tuning, it's a great tool. Gregg does performance tuning in depth on Linux, Solaris, Macintosh and other OSen. He'll use any tool that's available. It's hard to find network flow issues with top for example. dtrace is another tool in the box.

      • (Score: 0) by Anonymous Coward on Thursday November 03 2016, @12:35PM

        by Anonymous Coward on Thursday November 03 2016, @12:35PM (#422007)

        Dtrace *is* part of Solaris 10 and up, not a bolt on. Dtrace is in MacOSX (it has fewer trace points). I think it's in Oracle's Linux as well.

        So any predictions when it will be available on Windows? <gd&r>

        • (Score: 0) by Anonymous Coward on Thursday November 03 2016, @01:19PM

          by Anonymous Coward on Thursday November 03 2016, @01:19PM (#422018)

          As soon as Microsoft invents it.

          • (Score: 2) by Gaaark on Thursday November 03 2016, @01:36PM

            by Gaaark (41) on Thursday November 03 2016, @01:36PM (#422027) Journal

            And then patents it..... "hey, take us to court.... how much money YOU got?"

            --
            --- Please remind me if I haven't been civil to you: I'm channeling MDC. ---Gaaark 2.0 ---
        • (Score: 1, Informative) by Anonymous Coward on Thursday November 03 2016, @01:51PM

          by Anonymous Coward on Thursday November 03 2016, @01:51PM (#422031)

          >So any predictions when it will be available on Windows?

          No but it's been included in FBSD's default GENERIC kernal since 2012 :-p

          https://wiki.freebsd.org/DTrace/KernelSupport [freebsd.org]

        • (Score: 2) by TheRaven on Thursday November 03 2016, @01:57PM

          by TheRaven (270) on Thursday November 03 2016, @01:57PM (#422034) Journal
          Someone (at MSR, I think) did port it to Windows a couple of years ago, but I don't think it ever made it into deployment anywhere.
          --
          sudo mod me up
      • (Score: 2) by Runaway1956 on Thursday November 03 2016, @03:01PM

        by Runaway1956 (2926) Subscriber Badge on Thursday November 03 2016, @03:01PM (#422060) Journal

        That is informative, thank you. And, helps to put the whole story into perspective.

      • (Score: 1, Interesting) by Anonymous Coward on Thursday November 03 2016, @08:21PM

        by Anonymous Coward on Thursday November 03 2016, @08:21PM (#422223)

        Why restart? Did I miss something? gdb -p 1234. Or are we talking about recompiling to support prof/gprof (-p/-pg), etc?

  • (Score: 2) by requerdanos on Thursday November 03 2016, @01:10PM

    by requerdanos (5997) Subscriber Badge on Thursday November 03 2016, @01:10PM (#422013) Journal

    Linux in 2016 Catches up to Solaris... From 2004

    Okay. Let's try to improve this headline a little...

    "In Obscurity News, A Feature On Linux In 2016 Is Now Done Like A Similar Feature That Solaris Provided Many Years Ago."

    Sigh. The influences of accuracy and perspective make the headline all but unusable. The headline implying that Linux was somehow inferior to Solaris for twelve years was catchier, although more full of baloney.

    • (Score: 2) by letssee on Thursday November 03 2016, @03:35PM

      by letssee (2537) on Thursday November 03 2016, @03:35PM (#422080)

      Well, it *was* inferior, in some ways. And in some ways probably still is. I know DTrace was missed by a lot of people (myself included).

      And in other ways it is/was superior.

      Come on, it's a a catchy headline that well describes the subject. It *feels* like catching up for people who have used DTrace on solaris.

      • (Score: 2) by requerdanos on Thursday November 03 2016, @03:44PM

        by requerdanos (5997) Subscriber Badge on Thursday November 03 2016, @03:44PM (#422082) Journal

        Well, it *was* inferior, in some ways. And in some ways probably still is.

        Well, sure; but it's the important qualifier "in some ways" that was left out of the "catchy" headline.

  • (Score: 1) by zugedneb on Thursday November 03 2016, @03:25PM

    by zugedneb (4556) on Thursday November 03 2016, @03:25PM (#422073)

    give me a way to limit the IO cache to a physical adderss range, so that the random files in transfer do not mess up the memory?

    as it is, even if you limit with various methods the cache to 1 MB, it will spit that 1 MB all over the ram, so every other thing gets messed up.

    anyways, linux IO cachhing sux big time.

    --
    old saying: "a troll is a window into the soul of humanity" + also: https://en.wikipedia.org/wiki/Operation_Ajax
  • (Score: 3, Interesting) by Entropy on Thursday November 03 2016, @06:04PM

    by Entropy (4228) on Thursday November 03 2016, @06:04PM (#422160)

    If you've never used ZFS(from Solaris, now on Linux & others) it's worth a try. Snapshots that maintain performance, amazing interface, and volumes to physically back virtual machine storage(that also support snapshots). BTRFS has always been touted as a better ZFS, but it lags behind in many ways: No volumes for VMs, and snapshots for some reason always stay mounted(why would you want this?!) and read-write(again--why???). Not to mention stability.

    Linux has been light years ahead of Solaris in quite a few ways, but some of the fundamentals suchs as DTrace and ZFS are amazing things that still haven't been adopted today.

    • (Score: 1) by tbuskey on Sunday November 06 2016, @12:14AM

      by tbuskey (6127) on Sunday November 06 2016, @12:14AM (#422955)

      Linux has been light years ahead of Solaris in quite a few ways, but some of the fundamentals suchs as DTrace and ZFS are amazing things that still haven't been adopted today.

      Partly because they're reimplementing it a SystemTap and BTFS

      I think the reimplementation of Zones as containers has gone much better. I wasn't doing enough with Crossbow to see how the networking compares.

      Of course any new cool stuff is not going into Solaris. No one is paying a mass of developers to advance it and the ones that are developing it at Oracle or the companies around what was left in OpenSolaris won't share.

      ZFS advancement is similar but OpenZFS is getting implemented on Linux and the BSDs too. I've happily replaces Solaris with ZFS on Linux since shortly after it came out.

      • (Score: 2) by Entropy on Sunday November 06 2016, @02:31PM

        by Entropy (4228) on Sunday November 06 2016, @02:31PM (#423114)

        Yes, they are in a way trying to reimplement it. It's unfortunate that the focus seems against the features I find most important. Physical volumes(zvols) are making zero progress, and the rather intuitive interface is being left behind:

        zfs snapshot myfiles/whatever@today
        vs
        btrfs subvolume create myfiles/whatever myfiles/whatever/whydoiwantthismounted

        Then you have to deal with BTRFS having the snapshot mounted somewhere for tools like rsync to accidentally pick up, or deal with it being writeable by who knows what application or user accidentally. Suddenly your snapshot isn't a snapshot anymore. I kind of just wish they reimplemented the rather intuitive interface, and not leave behind enterprise important things like zvols, or pretend snapshots are a file copy operation.

        I think btrfs can replicate changes nowdays, so that's good.