Alexander Kolbasov
2005-Aug-10 00:41 UTC
[dtrace-discuss] A bit of IBM anti-DTrace propaganda
IBM published a 14 pages document explaining why Linux on POWER is better than Solaris 10: http://www-03.ibm.com/servers/enable/linux/power/pdfs/lop_vs_solaris10.pdf Here is the bits relevant for DTrace. Hope you enjoy it as much as I did: ------------------------------------------------------------------------------- Although SystemTap is still a work-in-progress, its current design features are comparable to those offered in DTrace. SystemTap will be safe and lightweight enough to use with live production systems. It will be able to instrument both kernel and user space programs even in the absence of source code. SystemTap''s probe language will be easy to use and users will be able to reuse general scripts written by others. One of the differences between SystemTap and Dtrace is that Dtrace uses an in-kernel interpreter whereas SystemTap uses compiled native code. Compiled native code is faster than interpreted code. Therefore, using SystemTap will not affect the performance of the system while performing performance measurements. The in-kernel interpreter has to be completely bug free, otherwise problems in the interpreter itself can cause the system to crash. Moreover, the interpreter is newly developed and not as mature as the compiler, hence there is a higher possibility of encountering bugs. There are a few kernel debugging features that will be supported by SystemTap but not by Dtrace. These features include the ability to write arbitrary locations in kernel memory and the ability to invoke arbitrary kernel subroutines. ------------------------------------------------------------------------------- The last sentence is particularly entertaining. Indeed SystemTap will have more "features" designed in!
Devon H. O''Dell
2005-Aug-10 00:55 UTC
[dtrace-discuss] A bit of IBM anti-DTrace propaganda
On Tue, Aug 09, 2005 at 05:41:01PM -0700, Alexander Kolbasov wrote:> IBM published a 14 pages document explaining why Linux on POWER is better > than Solaris 10: > > http://www-03.ibm.com/servers/enable/linux/power/pdfs/lop_vs_solaris10.pdf > > Here is the bits relevant for DTrace. Hope you enjoy it as much as I did: > > ------------------------------------------------------------------------------- > Although SystemTap is still a work-in-progress, its current design features are > comparable to those offered in DTrace. SystemTap will be safe and lightweight > enough to use with live production systems. It will be able to instrument both > kernel and user space programs even in the absence of source code. SystemTap''s > probe language will be easy to use and users will be able to reuse general > scripts written by others.I like how they start talking about what SystemTap WILL BE IN THE FUTURE here...> One of the differences between SystemTap and Dtrace is that Dtrace uses an > in-kernel interpreter whereas SystemTap uses compiled native code. > Compiled native code is faster than interpreted code. Therefore, using > SystemTap will not affect the performance of the system while performing > performance measurements. The in-kernel interpreter has to be completely > bug free, otherwise problems in the interpreter itself can cause the > system to crash. Moreover, the interpreter is newly developed and not as > mature as the compiler, hence there is a higher possibility of encountering > bugs. There are a few kernel debugging features that will be supported by > SystemTap but not by Dtrace. These features include the ability to write > arbitrary locations in kernel memory and the ability to invoke arbitrary > kernel subroutines. > -------------------------------------------------------------------------------...and how it''s already better than DTrace here. I can almost see Captain Contradiction rounding the corner... Of course, our interpreter has to be completely bug-free, but their C code doesn''t. Give me a break. It''s a million times easier to verify the security of an interpreter than it is the verify the security of tens of thousands of hand-coded probes contributed by a mass of random developers.> The last sentence is particularly entertaining. Indeed SystemTap will have > more "features" designed in!I particularly like that it''s left as an excercise for the reader to write the probes themselves. And that it is fundamentally insecure by design: allowing full read, write and execute access of kernel memory is ridiculous. I''ll laugh when the first SystemTap/KProbe based exploits come around and nobody can find out what the hell is going on because they''ve patched SystemTap and KProbe to ``ignore'''' the ``functionality.'''' That their experimental code is more trustworthy than production-ready and deployed systems code also gave me a good laugh. Unfortunately, there are many users who will be propogating this all over the place because they have No Clue(tm). The Joys of the Internet. IBM, come back when you have a product that you can compare. This is apples to cheeseburgers. --Devon
some day i will learn to hit reply to all on mailing lists. On 8/9/05, James Dickens <jamesd.wi at gmail.com> wrote:> On 8/9/05, Alexander Kolbasov <akolb at eng.sun.com> wrote: > > IBM published a 14 pages document explaining why Linux on POWER is better > > than Solaris 10: > > > > http://www-03.ibm.com/servers/enable/linux/power/pdfs/lop_vs_solaris10.pdf > > > > Here is the bits relevant for DTrace. Hope you enjoy it as much as I did: > > i responded to this in my blog at > http://uadmin.blogspot.com/2005/08/dtrace-equivalent-for-linux-only.html > > i also used there propaganda against them as i used there own > demonstration of Kprobes and compared them to dtrace. > > James Dickens > uadmin.blogspot.com > > > > > > > ------------------------------------------------------------------------------- > > Although SystemTap is still a work-in-progress, its current design features are > > comparable to those offered in DTrace. SystemTap will be safe and lightweight > > enough to use with live production systems. It will be able to instrument both > > kernel and user space programs even in the absence of source code. SystemTap''s > > probe language will be easy to use and users will be able to reuse general > > scripts written by others. > > > > One of the differences between SystemTap and Dtrace is that Dtrace uses an > > in-kernel interpreter whereas SystemTap uses compiled native code. > > Compiled native code is faster than interpreted code. Therefore, using > > SystemTap will not affect the performance of the system while performing > > performance measurements. The in-kernel interpreter has to be completely > > bug free, otherwise problems in the interpreter itself can cause the > > system to crash. Moreover, the interpreter is newly developed and not as > > mature as the compiler, hence there is a higher possibility of encountering > > bugs. There are a few kernel debugging features that will be supported by > > SystemTap but not by Dtrace. These features include the ability to write > > arbitrary locations in kernel memory and the ability to invoke arbitrary > > kernel subroutines. > > ------------------------------------------------------------------------------- > > > > The last sentence is particularly entertaining. Indeed SystemTap will have > > more "features" designed in! > > > > > > _______________________________________________ > > dtrace-discuss mailing list > > dtrace-discuss at opensolaris.org > > >
David Jafferian - Sun Microsystems
2005-Aug-10 03:39 UTC
[dtrace-discuss] A bit of IBM anti-DTrace propaganda
This seems like IBM is comparing apples to oranges. SystemTap seems to want to be more like mdb than DTrace. Yes, "Compiled native code is faster than interpreted code." Until you want to change it. That''s why system administrators like to use "scripting languages". The writer also seems to be trying to make the argument that compiled code has no probe effect when used to measure performance. As if Heisenberg''s Uncertainty Principle was null and void. Alexander Kolbasov writes: > IBM published a 14 pages document explaining why Linux on POWER is better > than Solaris 10: > > http://www-03.ibm.com/servers/enable/linux/power/pdfs/lop_vs_solaris10.pdf > > Here is the bits relevant for DTrace. Hope you enjoy it as much as I did: > > ------------------------------------------------------------------------------- > Although SystemTap is still a work-in-progress, its current design features are > comparable to those offered in DTrace. SystemTap will be safe and lightweight > enough to use with live production systems. It will be able to instrument both > kernel and user space programs even in the absence of source code. SystemTap''s > probe language will be easy to use and users will be able to reuse general > scripts written by others. > > One of the differences between SystemTap and Dtrace is that Dtrace uses an > in-kernel interpreter whereas SystemTap uses compiled native code. > Compiled native code is faster than interpreted code. Therefore, using > SystemTap will not affect the performance of the system while performing > performance measurements. The in-kernel interpreter has to be completely > bug free, otherwise problems in the interpreter itself can cause the > system to crash. Moreover, the interpreter is newly developed and not as > mature as the compiler, hence there is a higher possibility of encountering > bugs. There are a few kernel debugging features that will be supported by > SystemTap but not by Dtrace. These features include the ability to write > arbitrary locations in kernel memory and the ability to invoke arbitrary > kernel subroutines. > ------------------------------------------------------------------------------- > > The last sentence is particularly entertaining. Indeed SystemTap will have > more "features" designed in! > > > _______________________________________________ > dtrace-discuss mailing list > dtrace-discuss at opensolaris.org
Alexander Kolbasov wrote:>IBM published a 14 pages document explaining why Linux on POWER is better >than Solaris 10: > >http://www-03.ibm.com/servers/enable/linux/power/pdfs/lop_vs_solaris10.pdf > >Wow. They are comparing our 64-bit specjbb number to their 32-bit number. (See pages 8 and 9.) I suppose that since we don''t publish 32 bit numbers and they don''t publish 64 bit numbers there is no way to go apples-to-apples. -- Rich
Isaac Rozenfeld
2005-Aug-10 13:10 UTC
[dtrace-discuss] A bit of IBM anti-DTrace propaganda
Interesting how the author fails to reference key Dtrace features like being able to aggregate and quantize....and, in the summary, he claims to have introduced Dtrace ? The Usenix Dtrace paper is referenced, but obviously not all unique features of Dtrace are mentioned. How, then, can a heavily worded statement such as Linux on Power having "some, if not all, of the features provided by DTrace " on page 4 be considered truthful ? Isaac ----- Original Message ----- From: Richard Berlin <Richard.Berlin at Sun.COM> Date: Wednesday, August 10, 2005 0:53 am Subject: Re: [dtrace-discuss] A bit of IBM anti-DTrace propaganda> Alexander Kolbasov wrote: > > >IBM published a 14 pages document explaining why Linux on POWER > is better > >than Solaris 10: > > > >http://www- > 03.ibm.com/servers/enable/linux/power/pdfs/lop_vs_solaris10.pdf> > > > Wow. They are comparing our 64-bit specjbb number to their 32-bit > number.(See pages 8 and 9.) I suppose that since we don''t publish > 32 bit > numbers and > they don''t publish 64 bit numbers there is no way to go apples-to- > apples. > -- Rich >-------------- next part -------------- A non-text attachment was scrubbed... Name: iz.vcf Type: text/x-vcard Size: 263 bytes Desc: Card for Isaac Rozenfeld <Isaac.Rozenfeld at Sun.COM> URL: <http://mail.opensolaris.org/pipermail/dtrace-discuss/attachments/20050810/12885cce/attachment.vcf>
Joerg Schilling
2005-Aug-10 15:05 UTC
[dtrace-discuss] A bit of IBM anti-DTrace propaganda
"Devon H. O''Dell" <dodell at offmyserver.com> wrote:> ...and how it''s already better than DTrace here. I can almost see Captain > Contradiction rounding the corner... > > Of course, our interpreter has to be completely bug-free, but their C > code doesn''t. Give me a break. It''s a million times easier to verify the > security of an interpreter than it is the verify the security of tens of > thousands of hand-coded probes contributed by a mass of random > developers.If you run machine code in something like dtrace, anything could happen because there is no chance to verify the machine code. Verifying an interpreter is easier. J?rg -- EMail:joerg at schily.isdn.cs.tu-berlin.de (home) J?rg Schilling D-13353 Berlin js at cs.tu-berlin.de (uni) schilling at fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
On 8/9/05, James Dickens <jamesd.wi at gmail.com> wrote:> some day i will learn to hit reply to all on mailing lists. > > > > On 8/9/05, James Dickens <jamesd.wi at gmail.com> wrote: > > On 8/9/05, Alexander Kolbasov <akolb at eng.sun.com> wrote: > > > IBM published a 14 pages document explaining why Linux on POWER is better > > > than Solaris 10: > > > > > > http://www-03.ibm.com/servers/enable/linux/power/pdfs/lop_vs_solaris10.pdf > > > > > > Here is the bits relevant for DTrace. Hope you enjoy it as much as I did: > > > > i responded to this in my blog at > > http://uadmin.blogspot.com/2005/08/dtrace-equivalent-for-linux-only.html > > > > i also used there propaganda against them as i used there own > > demonstration of Kprobes and compared them to dtrace. > > > > James Dickens > > uadmin.blogspot.com > > > >well my blog has recived a few comments, that i should look into Systemtap, so i have looked into systemtap and my breif evaluation is in my blog if any of you want to follow this debate and or comment feel free. http://uadmin.blogspot.com/2005/08/systemtap-alpha.html James Dickens uadmin.blogspot.com> > > > > > > > ------------------------------------------------------------------------------- > > > Although SystemTap is still a work-in-progress, its current design features are > > > comparable to those offered in DTrace. SystemTap will be safe and lightweight > > > enough to use with live production systems. It will be able to instrument both > > > kernel and user space programs even in the absence of source code. SystemTap''s > > > probe language will be easy to use and users will be able to reuse general > > > scripts written by others. > > > > > > One of the differences between SystemTap and Dtrace is that Dtrace uses an > > > in-kernel interpreter whereas SystemTap uses compiled native code. > > > Compiled native code is faster than interpreted code. Therefore, using > > > SystemTap will not affect the performance of the system while performing > > > performance measurements. The in-kernel interpreter has to be completely > > > bug free, otherwise problems in the interpreter itself can cause the > > > system to crash. Moreover, the interpreter is newly developed and not as > > > mature as the compiler, hence there is a higher possibility of encountering > > > bugs. There are a few kernel debugging features that will be supported by > > > SystemTap but not by Dtrace. These features include the ability to write > > > arbitrary locations in kernel memory and the ability to invoke arbitrary > > > kernel subroutines. > > > ------------------------------------------------------------------------------- > > > > > > The last sentence is particularly entertaining. Indeed SystemTap will have > > > more "features" designed in! > > > > > > > > > _______________________________________________ > > > dtrace-discuss mailing list > > > dtrace-discuss at opensolaris.org > > > > > >
Merle Ilgenfritz
2005-Aug-12 22:00 UTC
[dtrace-discuss] A bit of IBM anti-DTrace propaganda
Hi all; First posting here on the list. As a Sun certified instructor, I''m enjoying all of the "insight" you guys are posting on this list. I plan on incorporating it into my Solaris 10 classes. I looked at this IBM document and after reading all of your rebuttal''s I am wondering if Sun has done an "official" response to this propaganda? I agree with the statement:> Unfortunately, there are many users who will be propogating this all > over the place because they have No Clue(tm). The Joys of the Internet.Thanks for the "enlightenment"! Merle Ilgenfritz Sun Certified Instructor/Course Developer/Consultant Ilgenfritz Consulting, LLC Exeter, RI -------Original Message-------> From: Devon H. O''Dell <dodell at offmyserver.com> > Subject: Re: [dtrace-discuss] A bit of IBM anti-DTrace propaganda > Sent: Aug 09 ''05 19:55 > > On Tue, Aug 09, 2005 at 05:41:01PM -0700, Alexander Kolbasov wrote: > > IBM published a 14 pages document explaining why Linux on POWER is better > > than Solaris 10: > > > > http://www-03.ibm.com/servers/enable/linux/power/pdfs/lop_vs_solaris10.pdf > > > > Here is the bits relevant for DTrace. Hope you enjoy it as much as I did: > > > > ------------------------------------------------------------------------------- > > Although SystemTap is still a work-in-progress, its current design features are > > comparable to those offered in DTrace. SystemTap will be safe and lightweight > > enough to use with live production systems. It will be able to instrument both > > kernel and user space programs even in the absence of source code. SystemTap''s > > probe language will be easy to use and users will be able to reuse general > > scripts written by others. > > I like how they start talking about what SystemTap WILL BE IN THE FUTURE > here... > > > One of the differences between SystemTap and Dtrace is that Dtrace uses an > > in-kernel interpreter whereas SystemTap uses compiled native code. > > Compiled native code is faster than interpreted code. Therefore, using > > SystemTap will not affect the performance of the system while performing > > performance measurements. The in-kernel interpreter has to be completely > > bug free, otherwise problems in the interpreter itself can cause the > > system to crash. Moreover, the interpreter is newly developed and not as > > mature as the compiler, hence there is a higher possibility of encountering > > bugs. There are a few kernel debugging features that will be supported by > > SystemTap but not by Dtrace. These features include the ability to write > > arbitrary locations in kernel memory and the ability to invoke arbitrary > > kernel subroutines. > > ------------------------------------------------------------------------------- > > ...and how it''s already better than DTrace here. I can almost see Captain > Contradiction rounding the corner... > > Of course, our interpreter has to be completely bug-free, but their C > code doesn''t. Give me a break. It''s a million times easier to verify the > security of an interpreter than it is the verify the security of tens of > thousands of hand-coded probes contributed by a mass of random > developers. > > > The last sentence is particularly entertaining. Indeed SystemTap will have > > more "features" designed in! > > I particularly like that it''s left as an excercise for the reader to write > the probes themselves. > > And that it is fundamentally insecure by design: allowing full read, write > and execute access of kernel memory is ridiculous. I''ll laugh when the > first SystemTap/KProbe based exploits come around and nobody can find out > what the hell is going on because they''ve patched SystemTap and KProbe to > ``ignore'''' the ``functionality.'''' > > That their experimental code is more trustworthy than production-ready > and deployed systems code also gave me a good laugh. > > Unfortunately, there are many users who will be propogating this all > over the place because they have No Clue(tm). The Joys of the Internet. > > IBM, come back when you have a product that you can compare. This is > apples to cheeseburgers. > > --Devon > _______________________________________________ > dtrace-discuss mailing list > dtrace-discuss at opensolaris.org >-------Original Message-------