Re: Clustering FreeBSD
I'm going to try these ideas out, thanks for the pointers. I'm highly motivated to stop waiting so long :-). And a nice use for the systems that have been piling up, if this works out. I'll be reporting back... Cheers, Russell % %On 20-Jan-01 Wes Peters wrote: % "Russell L. Carter" wrote: % %details snipped % % %See the paper "Recursive Make Considered Harmful." Make is an amazing % %tool when used correctly. % % That's not the problem, unfortunately. I've never had a problem % rebuilding dependencies unnecessarily, or any of those % other problems described. Well precompiled headers would be % really really cool. The problem, again, is that parallelism % is limited by the directory structure, and the directory structure % is entirely rational. % % The directory structure has nothing to do with the Makefiles. To % obtain the goal the paper suggests, you replace the recursive % Makefiles with a single top-level Makefile that describes ALL of the % targets and ALL of the dependencies. Note that this does not require % a single mono- lithic Makefile; the top level Makefile can be a shell % that includes per-directory Makefiles. The important part is to get a % single dependency tree with no cycles in the graph. % %I was so impressed by the clarity in the paper and dicussions with %friends that use Plan 9's "mk", that I put together "remake". This is a %Makefile framework that implements the per-directory Makefiles to build %the dependency tree. If anyone one wants to take a look it's at %http://www.ragnet.demon.co.uk/RM/remake.html %I haven't used it for a year or two and can only point to %http://www.ragnet.demon.co.uk/mynews %as an example of its use. % %If anyone gets interested drop me a line and I will try and remember how %it works. % %Duncan % To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: Clustering FreeBSD
% No it would not! Back in '94 I ported dmake to FreeBSD % and built just about every numerics package out there % on a 4 CPU cluster. Worked fine, but not much in overall % speedup, because... tadum! Where do you get the source % files, and how do you get the objs back :-) Not low % latency, eh? F-Enet then, G-Enet now :) % %You need a better file server. My previous employer, where the software %staff recompiles 3 million lines of code 20 or 30 times a day, employs %pmake and a farm of Sun Ultra-5 workstations to parallelize their makes. %It allows them to complete a build in an hour that would take a single %Ultra-5 almost 20 hours to complete, even with 3 or 4 builds running in %parallel. The network is 100BaseTX to the workstations and 1000BaseSX %to the (NFS) fileserver. Cool! I'd like to learn more. Then... can you elaborate on the build structure a bit? Is it a single large dir (surely not), or how do the dependencies work? For instance, with ACE/TAO (many hours to build when including orbsvcs) there's only a few large directories that can be parallelized over say 10 cpus by gmake, at least. The rest have ten files or less where each file takes maybe 45s to compile on a 1GHz processor. There are quite a few of these. And directories are compiled sequentially. % Nowadays, you'd want to "globus ify" things, rather than % use use PVM. % % But critically, speedup would only happen if jobs were % allocated at a higher level than they are now. % % Now for building something like a full version of TAO, % why that might work. But even then, a factor of 2x is % unlikely until the dependencies are factored out at % the directory level. % %See the paper "Recursive Make Considered Harmful." Make is an amazing %tool when used correctly. That's not the problem, unfortunately. I've never had a problem rebuilding dependencies unnecessarily, or any of those other problems described. Well precompiled headers would be really really cool. The problem, again, is that parallelism is limited by the directory structure, and the directory structure is entirely rational. Thanks! Russell To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: Clustering FreeBSD
%Uwe Pierau wrote: % % Jamie Heckford wrote: % # Hi, % # Does anyone have any details of Open Source, or software included % # with FreeBSD that allows the clustering of FreeBSD? % % Maybe you mean something like this... % http://acme.ecn.purdue.edu/index.html % ?! % %Yes! % %When is somebody going to get around to making a PVM version of make? %Wouldn't that help those "build world" times a bit? No it would not! Back in '94 I ported dmake to FreeBSD and built just about every numerics package out there on a 4 CPU cluster. Worked fine, but not much in overall speedup, because... tadum! Where do you get the source files, and how do you get the objs back :-) Not low latency, eh? F-Enet then, G-Enet now :) Nowadays, you'd want to "globus ify" things, rather than use use PVM. But critically, speedup would only happen if jobs were allocated at a higher level than they are now. Now for building something like a full version of TAO, why that might work. But even then, a factor of 2x is unlikely until the dependencies are factored out at the directory level. Russell %-- %"Where am I, and what am I doing in this handbasket?" % %Wes Peters Softweyr LLC [EMAIL PROTECTED] http://softweyr.com/ % % %To Unsubscribe: send mail to [EMAIL PROTECTED] %with "unsubscribe freebsd-hackers" in the body of the message % To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: Clustering FreeBSD
%In all honesty, I am just looking for something to play %with and see how fast FreeBSD can go. % %Sort of thing where those two guys clustered about 200 486's %or something stupid like that.. Go to google and search for Beowulf. Or Mosix. Or Ron Minnich :-) Or "smart networks", if all you want to do is serve up web pages. Russell %:) % %Jamie % %On 2001.01.16 18:31:43 + [EMAIL PROTECTED] wrote: % % % On 16 Jan, Jamie Heckford wrote: % Hi, % % Does anyone have any details of Open Source, or software included % with FreeBSD that allows the clustering of FreeBSD? % % I have 55 racks sitting here to play with, and want to start doing % some serious work (for me anyway!) with fBSD % % Plz. let me know! :) % % I've been working on some stuff for over a year, but % it nowhere near anything. What was it you were planning on doing? % % Jessem. % % % % % % % % %To Unsubscribe: send mail to [EMAIL PROTECTED] %with "unsubscribe freebsd-hackers" in the body of the message % To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: ld-elf.so.1: assert failed: /usr/src/libexec/rtld-elf/rtld.c:2033
%In article [EMAIL PROTECTED],
%Russell L. Carter [EMAIL PROTECTED] wrote:
%
% Bingo!
%
% Thanks guys!
%
%Not so fast there, fella. You're not getting off that easily. ;-)
%Could you please try the patch below? It is like the patch that Paul
%sent, except it should handle error conditions better.
%
%This patch is against -current, but I think it will apply cleanly to
%-stable too.
My pleasure. This patch applies cleanly against a two day old
-stable, and works just as well as the first patch, i.e., my
program works as expected.
Thanks!
Russell
%Thanks,
%John
%
%Index: rtld.c
%===
%RCS file: /home/ncvs/src/libexec/rtld-elf/rtld.c,v
%retrieving revision 1.50
%diff -u -r1.50 rtld.c
%--- rtld.c 2000/11/07 22:41:53 1.50
%+++ rtld.c 2001/01/05 00:13:18
%@@ -77,6 +77,8 @@
% static Obj_Entry *digest_phdr(const Elf_Phdr *, int, caddr_t, const char *);
% static Obj_Entry *dlcheck(void *);
% static bool donelist_check(DoneList *, const Obj_Entry *);
%+static void errmsg_restore(char *);
%+static char *errmsg_save(void);
% static char *find_library(const char *, const Obj_Entry *);
% static const char *gethints(void);
% static void init_dag(Obj_Entry *);
%@@ -457,6 +459,30 @@
% va_end(ap);
% }
%
%+/*
%+ * Return a dynamically-allocated copy of the current error message, if any.
%+ */
%+static char *
%+errmsg_save(void)
%+{
%+return error_message == NULL ? NULL : xstrdup(error_message);
%+}
%+
%+/*
%+ * Restore the current error message from a copy which was previously saved
%+ * by errmsg_save(). The copy is freed.
%+ */
%+static void
%+errmsg_restore(char *saved_msg)
%+{
%+if (saved_msg == NULL)
%+ error_message = NULL;
%+else {
%+ _rtld_error("%s", saved_msg);
%+ free(saved_msg);
%+}
%+}
%+
% static const char *
% basename(const char *name)
% {
%@@ -696,7 +722,7 @@
% if (obj == (Obj_Entry *) handle)
% break;
%
%-if (obj == NULL || obj-dl_refcount == 0) {
%+if (obj == NULL || obj-refcount == 0 || obj-dl_refcount == 0) {
% _rtld_error("Invalid shared object handle %p", handle);
% return NULL;
% }
%@@ -1184,13 +1210,20 @@
% objlist_call_fini(Objlist *list)
% {
% Objlist_Entry *elm;
%+char *saved_msg;
%
%+/*
%+ * Preserve the current error message since a fini function might
%+ * call into the dynamic linker and overwrite it.
%+ */
%+saved_msg = errmsg_save();
% STAILQ_FOREACH(elm, list, link) {
% if (elm-obj-refcount == 0) {
% dbg("calling fini function for %s", elm-obj-path);
% (*elm-obj-fini)();
% }
% }
%+errmsg_restore(saved_msg);
% }
%
% /*
%@@ -1202,11 +1235,18 @@
% objlist_call_init(Objlist *list)
% {
% Objlist_Entry *elm;
%+char *saved_msg;
%
%+/*
%+ * Preserve the current error message since an init function might
%+ * call into the dynamic linker and overwrite it.
%+ */
%+saved_msg = errmsg_save();
% STAILQ_FOREACH(elm, list, link) {
% dbg("calling init function for %s", elm-obj-path);
% (*elm-obj-init)();
% }
%+errmsg_restore(saved_msg);
% }
%
% static void
%@@ -2030,7 +2070,8 @@
% {
% const Needed_Entry *needed;
%
%-assert(root-refcount != 0);
%+if (root-refcount == 0)
%+ return;
% root-refcount--;
% if (root-refcount == 0)
% for (needed = root-needed; needed != NULL; needed = needed-next)
%
%
%To Unsubscribe: send mail to [EMAIL PROTECTED]
%with "unsubscribe freebsd-hackers" in the body of the message
%
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
Re: ld-elf.so.1: assert failed: /usr/src/libexec/rtld-elf/rtld.c:2033
Bingo!
Thanks guys!
Russell
%John Polstra ([EMAIL PROTECTED]) wrote:
% In article [EMAIL PROTECTED],
% Russell L. Carter [EMAIL PROTECTED] wrote:
%
% On a fairly recent -STABLE I am getting this failure:
%
% ld-elf.so.1: assert failed: /usr/src/libexec/rtld-elf/rtld.c:2033
%
% I assume I'm doing something stupid, however the same code
% works on Linux gcc-2.95.2, so I'm looking for what the
% difference might be.
%
% The program is an ACE/TAO C++ program that dlsym()s an object,
% uses it happily, and then gets the assertion when dlclose()ing
% from the containing object's dtor.
%
% The assertion is that the refcount != 0. What should I
% do to fix that?
%
% There is a bug in the dynamic linker in connection with calling dlopen
% or dlclose from a static constructor or destructor, and this sounds
% like it's related to that. I came up with a fix for it which Peter
% Wemm was testing at Yahoo. That was a few months ago, and I'll have
% to dig it out again after the holiday madness has subsided. If you
% haven't heard from me by Saturday Jan. 6, I'd appreciate a gentle
% reminder to [EMAIL PROTECTED].
%
%Do you mean this patch?
%--- rtld.c 2000/08/01 07:17:54 1.1.1.3
%+++ rtld.c 2000/09/05 22:16:49 1.2
%@@ -694,7 +694,7 @@
% if (obj == (Obj_Entry *) handle)
% break;
%
%-if (obj == NULL || obj-dl_refcount == 0) {
%+if (obj == NULL || obj-refcount == 0 || obj-dl_refcount == 0) {
% _rtld_error("Invalid shared object handle %p", handle);
% return NULL;
% }
%@@ -1994,6 +1994,8 @@
% {
% const Needed_Entry *needed;
%
%+if (root-refcount == 0)
%+ return;
% assert(root-refcount != 0);
% root-refcount--;
% if (root-refcount == 0)
%
%--
%Paul Saab
%Technical Yahoo
[EMAIL PROTECTED] - [EMAIL PROTECTED] - [EMAIL PROTECTED]
%Do You .. uhh .. Yahoo!?
%
%
%To Unsubscribe: send mail to [EMAIL PROTECTED]
%with "unsubscribe freebsd-hackers" in the body of the message
%
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
ld-elf.so.1: assert failed: /usr/src/libexec/rtld-elf/rtld.c:2033
Greetings, On a fairly recent -STABLE I am getting this failure: ld-elf.so.1: assert failed: /usr/src/libexec/rtld-elf/rtld.c:2033 I assume I'm doing something stupid, however the same code works on Linux gcc-2.95.2, so I'm looking for what the difference might be. The program is an ACE/TAO C++ program that dlsym()s an object, uses it happily, and then gets the assertion when dlclose()ing from the containing object's dtor. The assertion is that the refcount != 0. What should I do to fix that? Thanks! Russell To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: ACE wrappers woes on 4.x-stable (pthreads)
| |forgot to add: I need pointers to start digging around It might be interesting to trace through libgcc_r. Russell | |/fjoe | | | |To Unsubscribe: send mail to [EMAIL PROTECTED] |with "unsubscribe freebsd-hackers" in the body of the message | To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: FreeBSD Posix threads
%
%I have searched the archive for a while and still have some confusions
%about this subjects:
%
%(1) Some people say "For I/O bound activity, kernel threads are a really
%bad idea". But I read the following passage from else where:
%
%Kernel threads perform better in I/O-intensive applications because system
%call can be blocking for kernel threads (so there is only one
%user-to-kernel and kernel-to-user transition per I/O operation rather than
%two).
%
%So which one is the correct answer? I know there is only userland thread
%in FreeBSD, but I would like to know the answer.
%
%(2) User threads are supposed to be faster than kernel threads in context
%switch. But since we poll devices during each context switch, it is
%actually slower (poll() is the extra system call). Is this correct?
%
%(3) Can I do cooperative thread scheduling on FreeBSD? I guess the
%constant SCHED_RR (round robin) means preemptive. How about SCHED_FIFO
%and SCHED_OTHER defined in posix4/sched.h? Can I choose from them? In
%the case of SCHED_RR, I still do part of scheduling by routines like
%yield(), cond_signal(). Am I right?
%
%Any help is appreciated.
Excellent questions.
However, what is not stated is perhaps the most important (to me,
and other people burning up and showering down $$$). C++ programmers
use threads to share class members. In that context, the POSIX
realtime scheduling classes (SCHED_*) are quite useful, albeit abstractly
for now.
IMNSHO, focusing on fine-grained benchmarks ("perform better") for
an execution model based on C processes may be misleading in the
realistic broader picture.
Russell
%-Zhihui
%
%
%
%
%To Unsubscribe: send mail to [EMAIL PROTECTED]
%with "unsubscribe freebsd-hackers" in the body of the message
%
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
Re: FreeBSD Threads (was Re: Practical limit for number of TCP connections?)
%On Sun, Dec 19, 1999 at 02:49:26PM -0800, Ronald F. Guilmette wrote: % % I dunno if it goes without saying or not, but this certainly makes % the current FreeBSD threads implementation highly unpalatable, except % to support ported code which has been developed elsewhere and which % is already written to assume the presence/availability of thread % library primitives. % %I wouldn't go that far. The current implementation is not nearly %as good as it could be, but it works nicely for some reasonable %range of applications. Adding to Richard's comments: libc_r threads work damn fine for wide variety of applications. There are specific user-space issues which preclude using a lot of POSIX RT stuff, and libc_r threads naturally don't scale to multiple processors. But they're relatively lightweight :-). And I believe most of the essential interfaces are there, and not harmful if used. % Certainly, if I were starting out to write a brand new new multi-threaded % application program specifically for execution on FreeBSD _today_, I'd % either (a) wait for the new thread implementation to be ready or else % (b) not use threads at all. % %If user threads don't work for you, then the linuxthreads port %(uses kernel threads) might. Its a POSIX style pthreads implementation, %so when FreeBSD does "kernel threads", or the equivalent, if should %be as easy (cross your fingers) as recompiling and relinking %to a different library. Yep. You might try the information on www.pinyon.org/ace to see any differences in compilation requirements. % (Actually, I guess there is a third option, % i.e. writing my code to use the existing threads API, and then just hoping % that the new threads implementation shows up before I was done coding my % application.) I build against both existing interfaces, and adhere to the several POSIX specs. It seems to work fine. Russell To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: Basic question about threads and SMP
%Nick Hibma wrote: % % Being multi-threaded has almost nothing to do with being % multi-processor. Multi-threading means that your application has % multiple threads of execution that are able to run simultaneously. % % The multi-processing capability of your box means that 2 threads of % execution, be it a process or a thread within a process, are executed % _literally_ at the same time, and not in simulated concurrency like it % happens on a UP box. % %Note that this happens ONLY if both threads of execution are processor %mobile. If your system supports user-space threads as part of a %process and the process can't be split across CPUs, you might as well %have a UP system. (Except everything else can run on the other %processor, so SMP is still a small win.) % %This is the situation with threads and SMP in -current. % The LinuxThreads port is currently busted for SMP but when it is fixed it will indeed use multiple processors: last pid: 395; load averages: 0.50, 0.11, 0.04 up 0+00:13:00 20:00:40 37 processes: 6 running, 31 sleeping CPU states: 65.8% user, 0.0% nice, 0.0% system, 0.0% interrupt, 34.2% idle Mem: 8780K Active, 9116K Inact, 19M Wired, 68K Cache, 7000K Buf, 466M Free Swap: 1024M Total, 1024M Free PID USERNAME PRI NICE SIZERES STATE C TIME WCPUCPU COMMAND 393 rcarter 31 0 876K 152K CPU1 1 0:01 17.16% 2.39% ex3 394 rcarter 29 0 876K 152K RUN0 0:00 15.76% 2.20% ex3 391 rcarter 30 0 876K 152K RUN1 0:01 14.71% 2.05% ex3 395 rcarter 30 0 876K 152K RUN0 0:01 14.01% 1.95% ex3 392 rcarter 30 0 876K 152K RUN0 0:00 13.66% 1.90% ex3 Russell %-- %"Where am I, and what am I doing in this handbasket?" % %Wes Peters Softweyr LLC [EMAIL PROTECTED] http://softweyr.com/ % % %To Unsubscribe: send mail to [EMAIL PROTECTED] %with "unsubscribe freebsd-hackers" in the body of the message % To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
/usr/include/posix4/semaphore.h
Hi All, The stuff defined in /usr/include/posix4/semaphore.h, is not implemented in -current, right? If I missed it, I'd appreciate a pointer... Russell To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: TCP stack hackers take a bow
%Unmodified FreeBSD TCP at 1Gb/s. % %http://www.sciencedaily.com/releases/1999/08/990802072727.htm That is so very cool. There is a separate war going on optimizing bandwidth, latency, and QoS for IIOP, i.e. CORBA's usual protocol. Against all of the heavyweights, RTOS's etc. etc., linux is looking very good. Surprisingly good. Thus... Cheers, Russell To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: TCP stack hackers take a bow
%Unmodified FreeBSD TCP at 1Gb/s. % %http://www.sciencedaily.com/releases/1999/08/990802072727.htm That is so very cool. There is a separate war going on optimizing bandwidth, latency, and QoS for IIOP, i.e. CORBA's usual protocol. Against all of the heavyweights, RTOS's etc. etc., linux is looking very good. Surprisingly good. Thus... Cheers, Russell To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-hackers in the body of the message
Re: What good PII/PIII Motherboards for FreeBSD and Celeron CPU's
|At 10:35 AM 7/23/99 -0400, Adrian Filipi-Martin wrote: |On Thu, 22 Jul 1999, Alex Zepeda wrote: | | On Thu, 22 Jul 1999, Adrian Filipi-Martin wrote: | |I know a lot of people like the ASUS P2B boards, but I've noticed a | tendency for the systems to reset occasionally when plugging in a |keyboard. | I've seen this with both FreeBSD and NT, so I'm considering it a property | of the board. | | If this is a PS/2 style keyboard, don't plug and unplug them when the host | is powered up. The PS/2 style stuf seems to be very sensitive to that | sort of thing. If it was a USB keyboard doing that... then that would be | odd. | | Yes, it is a PS/2 keyboard, but these are the only MB's that I have |run into this problem with other than some AIX boxes years ago that would |blow a fuse when disconnecting the keyboard from a running system. | | This isn't that big a problem. I only have keyboard installed |while building the systems. Thereafter they are serial console only. | |I've seen it on an SBC (with a KB connector on board). it IS a problem if |they want to do maintanence without bringing down the system. blush I fried two P6 ASUS motherboards this way, sorta along these lines, "hmm, keyboard seems to be dead, maybe try it in this machine" Russell | |Dennis | | |To Unsubscribe: send mail to [EMAIL PROTECTED] |with "unsubscribe freebsd-hackers" in the body of the message | To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: What good PII/PIII Motherboards for FreeBSD and Celeron CPU's
|At 10:35 AM 7/23/99 -0400, Adrian Filipi-Martin wrote: |On Thu, 22 Jul 1999, Alex Zepeda wrote: | | On Thu, 22 Jul 1999, Adrian Filipi-Martin wrote: | |I know a lot of people like the ASUS P2B boards, but I've noticed a | tendency for the systems to reset occasionally when plugging in a |keyboard. | I've seen this with both FreeBSD and NT, so I'm considering it a property | of the board. | | If this is a PS/2 style keyboard, don't plug and unplug them when the host | is powered up. The PS/2 style stuf seems to be very sensitive to that | sort of thing. If it was a USB keyboard doing that... then that would be | odd. | | Yes, it is a PS/2 keyboard, but these are the only MB's that I have |run into this problem with other than some AIX boxes years ago that would |blow a fuse when disconnecting the keyboard from a running system. | | This isn't that big a problem. I only have keyboard installed |while building the systems. Thereafter they are serial console only. | |I've seen it on an SBC (with a KB connector on board). it IS a problem if |they want to do maintanence without bringing down the system. blush I fried two P6 ASUS motherboards this way, sorta along these lines, hmm, keyboard seems to be dead, maybe try it in this machine Russell | |Dennis | | |To Unsubscribe: send mail to majord...@freebsd.org |with unsubscribe freebsd-hackers in the body of the message | To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-hackers in the body of the message
Re: Redundant Remote Webserver clustering
% Define clustering. If you mean a bunch of boxes that serve up HTTP % requests and the lot of them continue working in the face of a % failure on one, you CAN do this with FreeBSD, and the Beowulf % software you're probably thinking of for Linux WILL NOT do this. %I have looked into the Beowulf system alot recently. It is nothing but %a glorified COW design. And it uses off the shelf software components %that run under FreeBSD as well of better than linux often. I used to %thing it was a big deal. Not any more :I This is a tangent though :) That's exactly right, and has been true for about 4 years now. % You do this on FreeBSD (or Linux or Solaris) by creating a layer 4 % router or HTTP switch that directs traffic evenly among your several % web servers, and stops sending traffic to servers that have failed. %Where could someone find information on setting this up, and what software %to use? I have someone who would be very interested in this. Isn't the %layer 4 router a SPoF though? I haven't actually installed one but my understanding, based on evaluating it as a competitor to some work that I am doing, is you just buy these off the shelf now. I don't have the vendor list handy at home. These things apparently are pretty good at looking inside of packets and making (fast) routing decisions based on the packet protocol and a set of site configured policies. The jargon thrown around my neck of the woods is smart networks. They understand HTTP pretty darn well. Luckily, not IIOP, yet, which is when my interest faded. Oh yes, about that SPoF, they happily run redundantly, too, with the obvious performance optimizations for redundant channels, degrading as needed. Apparently, these things don't need proprietary extensions for their functionality so use of FreeBSD for the backend OS is unimpaired. I suspect Yahoo has a few... Anyway, maybe followup this to freebsd-isp? Not much -hackers stuff here. And if my assessment is inaccurate, I can find out quicker :-). Russell % %-- %David Cross | email: cro...@cs.rpi.edu %Systems Administrator/Research Programmer | Web: http://www.cs.rpi.edu/~crossd %Rensselaer Polytechnic Institute, | Ph: 518.276.2860 %Department of Computer Science| Fax: 518.276.4033 %I speak only for myself. | WinNT:Linux::Linux:FreeBSD To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-hackers in the body of the message
Re: Microsoft performance (was: ...)
%Basically there are some applications and benchmarks for which FreeBSD uh, benchmarks only, until evidence is produced otherwise. Tuning for benchmarks has been around a long long time. People get worked up about this because the people who give out the money to buy the systems use benchmarks to decide whom to give the money to. It's really, really stupid to rely on generic benchmarks. But people do, anyway. So I guess whistle and some others should invest in tuning for the benchmarks. Like jupiter, eh? Or maybe Apple. But for the rest, I wouldn't panic. In fact, there's probably some interesting kernel architecture issues here. Let's hear them, now! If I wanted secrecy about architecture details there's a shitload less time consuming ways to do it then follow FreeBSD. Russell To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-hackers in the body of the message
Re: Microsoft performance (was: ...)
% % %On Wed, 23 Jun 1999, Russell L. Carter wrote: % % % %Basically there are some applications and benchmarks for which FreeBSD % % uh, benchmarks only, until evidence is produced otherwise. [...] %ok here are some of the problems.. % %Matt's changes allow dd to copy data at 2.5 times the rate it did before. %I consider dd to be an application. The problem is due to resource %handling in the kernel and results in large amounts of Idle CPU time. Ok, why doesn't this show up in any of the disk or network benchmarks? %Another primary problem with the FreeBSD kernel (being addressed by Kirk) %is that after writing a file, once the data has been queued for IO you %cannot read the data in that file (even though it is present) until the IO %is complete. With 64 tags, it is concievable that this could take a half %second on a modern disk. That's interesting. %These are problems shown up by the benchmarks but %which can be shown to affect ordinary operations. % %There are other problems related to SMP and the GKL.. %e.g.. two processes cannot access buffers at the same time, even though %they are both present , because only one of them is allowed in the kernel %at a time. Therefore One processor will spend a bunch of time at idle.. Yup. Thanks for filling us in! Russell To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-hackers in the body of the message
