Обсуждение: [PATCH] PostgreSQL 9.4 mmap(2) performance regression on FreeBSD...
One of the patches that I've been sitting on and am derelict in punting upstream is the attached mmap(2) flags patch for the BSDs. Is there any chance this can be squeezed in to the PostreSQL 9.4 release?
The patch is trivial in size and is used to add one flag to mmap(2) calls in dsm_impl.c. Alan Cox (FreeBSD alc, not Linux) and I went back and forth regarding PostgreSQL's use of mmap(2) and determined that the following is correct and will prevent a likely performance regression in PostgreSQL 9.4. In PostgreSQL 9.3, all mmap(2) calls were called with the flags MAP_ANON | MAP_SHARED, whereas in PostgreSQL 9.4 this is not the case.
Digging in to the patch, in reviewing src/backend/storage/ipc/dsm_impl.c, it's clear that rhaas@ understood the consequences of mmap(2), and the possible consequences of having dirty pages gratuitously flushed to disk:
src/backend/storage/ipc/dsm_impl.c:781
In order for the above comment to be true for FreeBSD, an extra flag needs to be passed to mmap(2). From FreeBSD 10's mmap(2) page[2]:
MAP_NOSYNC Causes data dirtied via this VM map to be flushed to
physical media only when necessary (usually by the
pager) rather than gratuitously. Typically this pre-
vents the update daemons from flushing pages dirtied
through such maps and thus allows efficient sharing of
memory across unassociated processes using a file-
backed shared memory map. Without this option any VM
pages you dirty may be flushed to disk every so often
(every 30-60 seconds usually) which can create perfor-
mance problems if you do not need that to occur (such
as when you are using shared file-backed mmap regions
for IPC purposes). Note that VM/file system coherency
is maintained whether you use MAP_NOSYNC or not. This
option is not portable across UNIX platforms (yet),
though some may implement the same behavior by
default.
WARNING! Extending a file with ftruncate(2), thus
creating a big hole, and then filling the hole by mod-
ifying a shared mmap() can lead to severe file frag-
mentation. In order to avoid such fragmentation you
should always pre-allocate the file's backing store by
write()ing zero's into the newly extended area prior
to modifying the area via your mmap(). The fragmenta-
tion problem is especially sensitive to MAP_NOSYNC
pages, because pages may be flushed to disk in a
totally random order.
The same applies when using MAP_NOSYNC to implement a
file-based shared memory store. It is recommended
that you create the backing store by write()ing zero's
to the backing file rather than ftruncate()ing it.
You can test file fragmentation by observing the KB/t
(kilobytes per transfer) results from an ``iostat 1''
while reading a large file sequentially, e.g. using
``dd if=filename of=/dev/null bs=32k''.
The fsync(2) system call will flush all dirty data and
metadata associated with a file, including dirty
NOSYNC VM data, to physical media. The sync(8) com-
mand and sync(2) system call generally do not flush
dirty NOSYNC VM data. The msync(2) system call is
usually not needed since BSD implements a coherent
file system buffer cache. However, it may be used to
associate dirty VM pages with file system buffers and
thus cause them to be flushed to physical media sooner
rather than later.
The man page for madvise(2) has more pointed advise[3]:
MADV_NOSYNC Request that the system not flush the data associated
with this map to physical backing store unless it needs
to. Typically this prevents the file system update dae-
mon from gratuitously writing pages dirtied by the VM
system to physical disk. Note that VM/file system
coherency is always maintained, this feature simply
ensures that the mapped data is only flush when it needs
to be, usually by the system pager.
This feature is typically used when you want to use a
file-backed shared memory area to communicate between
processes (IPC) and do not particularly need the data
being stored in that area to be physically written to
disk. With this feature you get the equivalent perfor-
mance with mmap that you would expect to get with SysV
shared memory calls, but in a more controllable and less
restrictive manner. However, note that this feature is
not portable across UNIX platforms (though some may do
the right thing by default). For more information see
the MAP_NOSYNC section of mmap(2)
Anyway, could you give this a quick review and apply the patch in time so the build farm can get a full build completed before the release?
Thanks in advance. -sc
[1] https://kib.kiev.ua/kib/pgsql_perf.pdf
[2] http://www.freebsd.org/cgi/man.cgi?query=mmap&apropos=0&sektion=0&manpath=FreeBSD+10.0-stable&arch=default&format=html
[3] http://www.freebsd.org/cgi/man.cgi?query=madvise&sektion=2&apropos=0&manpath=FreeBSD+10.0-stable
The patch is trivial in size and is used to add one flag to mmap(2) calls in dsm_impl.c. Alan Cox (FreeBSD alc, not Linux) and I went back and forth regarding PostgreSQL's use of mmap(2) and determined that the following is correct and will prevent a likely performance regression in PostgreSQL 9.4. In PostgreSQL 9.3, all mmap(2) calls were called with the flags MAP_ANON | MAP_SHARED, whereas in PostgreSQL 9.4 this is not the case.
Digging in to the patch, in reviewing src/backend/storage/ipc/dsm_impl.c, it's clear that rhaas@ understood the consequences of mmap(2), and the possible consequences of having dirty pages gratuitously flushed to disk:
src/backend/storage/ipc/dsm_impl.c:781
* Operating system primitives to support mmap-based shared memory.
*
* Calling this "shared memory" is somewhat of a misnomer, because what
* we're really doing is creating a bunch of files and mapping them into
* our address space. The operating system may feel obliged to
* synchronize the contents to disk even if nothing is being paged out,
* which will not serve us well. The user can relocate the pg_dynshmem
* directory to a ramdisk to avoid this problem, if available.
*
* Calling this "shared memory" is somewhat of a misnomer, because what
* we're really doing is creating a bunch of files and mapping them into
* our address space. The operating system may feel obliged to
* synchronize the contents to disk even if nothing is being paged out,
* which will not serve us well. The user can relocate the pg_dynshmem
* directory to a ramdisk to avoid this problem, if available.
In order for the above comment to be true for FreeBSD, an extra flag needs to be passed to mmap(2). From FreeBSD 10's mmap(2) page[2]:
MAP_NOSYNC Causes data dirtied via this VM map to be flushed to
physical media only when necessary (usually by the
pager) rather than gratuitously. Typically this pre-
vents the update daemons from flushing pages dirtied
through such maps and thus allows efficient sharing of
memory across unassociated processes using a file-
backed shared memory map. Without this option any VM
pages you dirty may be flushed to disk every so often
(every 30-60 seconds usually) which can create perfor-
mance problems if you do not need that to occur (such
as when you are using shared file-backed mmap regions
for IPC purposes). Note that VM/file system coherency
is maintained whether you use MAP_NOSYNC or not. This
option is not portable across UNIX platforms (yet),
though some may implement the same behavior by
default.
WARNING! Extending a file with ftruncate(2), thus
creating a big hole, and then filling the hole by mod-
ifying a shared mmap() can lead to severe file frag-
mentation. In order to avoid such fragmentation you
should always pre-allocate the file's backing store by
write()ing zero's into the newly extended area prior
to modifying the area via your mmap(). The fragmenta-
tion problem is especially sensitive to MAP_NOSYNC
pages, because pages may be flushed to disk in a
totally random order.
The same applies when using MAP_NOSYNC to implement a
file-based shared memory store. It is recommended
that you create the backing store by write()ing zero's
to the backing file rather than ftruncate()ing it.
You can test file fragmentation by observing the KB/t
(kilobytes per transfer) results from an ``iostat 1''
while reading a large file sequentially, e.g. using
``dd if=filename of=/dev/null bs=32k''.
The fsync(2) system call will flush all dirty data and
metadata associated with a file, including dirty
NOSYNC VM data, to physical media. The sync(8) com-
mand and sync(2) system call generally do not flush
dirty NOSYNC VM data. The msync(2) system call is
usually not needed since BSD implements a coherent
file system buffer cache. However, it may be used to
associate dirty VM pages with file system buffers and
thus cause them to be flushed to physical media sooner
rather than later.
The man page for madvise(2) has more pointed advise[3]:
MADV_NOSYNC Request that the system not flush the data associated
with this map to physical backing store unless it needs
to. Typically this prevents the file system update dae-
mon from gratuitously writing pages dirtied by the VM
system to physical disk. Note that VM/file system
coherency is always maintained, this feature simply
ensures that the mapped data is only flush when it needs
to be, usually by the system pager.
This feature is typically used when you want to use a
file-backed shared memory area to communicate between
processes (IPC) and do not particularly need the data
being stored in that area to be physically written to
disk. With this feature you get the equivalent perfor-
mance with mmap that you would expect to get with SysV
shared memory calls, but in a more controllable and less
restrictive manner. However, note that this feature is
not portable across UNIX platforms (though some may do
the right thing by default). For more information see
the MAP_NOSYNC section of mmap(2)
Anyway, could you give this a quick review and apply the patch in time so the build farm can get a full build completed before the release?
Thanks in advance. -sc
[1] https://kib.kiev.ua/kib/pgsql_perf.pdf
[2] http://www.freebsd.org/cgi/man.cgi?query=mmap&apropos=0&sektion=0&manpath=FreeBSD+10.0-stable&arch=default&format=html
[3] http://www.freebsd.org/cgi/man.cgi?query=madvise&sektion=2&apropos=0&manpath=FreeBSD+10.0-stable
--
Sean Chittenden
Sean Chittenden
Вложения
On 2014-08-12 09:42:30 -0700, Sean Chittenden wrote: > One of the patches that I've been sitting on and am derelict in punting > upstream is the attached mmap(2) flags patch for the BSDs. Is there any > chance this can be squeezed in to the PostreSQL 9.4 release? > > The patch is trivial in size and is used to add one flag to mmap(2) calls in > dsm_impl.c. Alan Cox (FreeBSD alc, not Linux) and I went back and forth > regarding PostgreSQL's use of mmap(2) and determined that the following is > correct and will prevent a likely performance regression in PostgreSQL 9.4. > In PostgreSQL 9.3, all mmap(2) calls were called with the flags MAP_ANON | > MAP_SHARED, whereas in PostgreSQL 9.4 this is not the case. The performancewise important call to mmap will still use that set of flags, no? That's the one backing shared_buffers. The mmap backend for *dynamic* shared memory (aka dsm) is *NOT* supposed to be used on common platforms. Both posix and sysv shared memory will be used before falling back to the mmap() backend. Greetings, Andres Freund -- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services
On Tue, Aug 12, 2014 at 12:59 PM, Andres Freund <andres@2ndquadrant.com> wrote: > On 2014-08-12 09:42:30 -0700, Sean Chittenden wrote: >> One of the patches that I've been sitting on and am derelict in punting >> upstream is the attached mmap(2) flags patch for the BSDs. Is there any >> chance this can be squeezed in to the PostreSQL 9.4 release? >> >> The patch is trivial in size and is used to add one flag to mmap(2) calls in >> dsm_impl.c. Alan Cox (FreeBSD alc, not Linux) and I went back and forth >> regarding PostgreSQL's use of mmap(2) and determined that the following is >> correct and will prevent a likely performance regression in PostgreSQL 9.4. >> In PostgreSQL 9.3, all mmap(2) calls were called with the flags MAP_ANON | >> MAP_SHARED, whereas in PostgreSQL 9.4 this is not the case. > > The performancewise important call to mmap will still use that set of > flags, no? That's the one backing shared_buffers. > > The mmap backend for *dynamic* shared memory (aka dsm) is *NOT* supposed > to be used on common platforms. Both posix and sysv shared memory will > be used before falling back to the mmap() backend. Hmm, yeah. This might still be a good thing to do (because what do we lose?) but it shouldn't really be an issue in practice. -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company
On Tue, Aug 12, 2014 at 07:08:06PM -0400, Robert Haas wrote: > On Tue, Aug 12, 2014 at 12:59 PM, Andres Freund <andres@2ndquadrant.com> wrote: > > On 2014-08-12 09:42:30 -0700, Sean Chittenden wrote: > >> One of the patches that I've been sitting on and am derelict in punting > >> upstream is the attached mmap(2) flags patch for the BSDs. Is there any > >> chance this can be squeezed in to the PostreSQL 9.4 release? > >> > >> The patch is trivial in size and is used to add one flag to mmap(2) calls in > >> dsm_impl.c. Alan Cox (FreeBSD alc, not Linux) and I went back and forth > >> regarding PostgreSQL's use of mmap(2) and determined that the following is > >> correct and will prevent a likely performance regression in PostgreSQL 9.4. > >> In PostgreSQL 9.3, all mmap(2) calls were called with the flags MAP_ANON | > >> MAP_SHARED, whereas in PostgreSQL 9.4 this is not the case. > > > > The performancewise important call to mmap will still use that set of > > flags, no? That's the one backing shared_buffers. > > > > The mmap backend for *dynamic* shared memory (aka dsm) is *NOT* supposed > > to be used on common platforms. Both posix and sysv shared memory will > > be used before falling back to the mmap() backend. > > Hmm, yeah. This might still be a good thing to do (because what do we > lose?) but it shouldn't really be an issue in practice. Is there a reason this was not applied? -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. +
On 2014-10-11 20:33:57 -0400, Bruce Momjian wrote: > On Tue, Aug 12, 2014 at 07:08:06PM -0400, Robert Haas wrote: > > On Tue, Aug 12, 2014 at 12:59 PM, Andres Freund <andres@2ndquadrant.com> wrote: > > > On 2014-08-12 09:42:30 -0700, Sean Chittenden wrote: > > >> One of the patches that I've been sitting on and am derelict in punting > > >> upstream is the attached mmap(2) flags patch for the BSDs. Is there any > > >> chance this can be squeezed in to the PostreSQL 9.4 release? > > >> > > >> The patch is trivial in size and is used to add one flag to mmap(2) calls in > > >> dsm_impl.c. Alan Cox (FreeBSD alc, not Linux) and I went back and forth > > >> regarding PostgreSQL's use of mmap(2) and determined that the following is > > >> correct and will prevent a likely performance regression in PostgreSQL 9.4. > > >> In PostgreSQL 9.3, all mmap(2) calls were called with the flags MAP_ANON | > > >> MAP_SHARED, whereas in PostgreSQL 9.4 this is not the case. > > > > > > The performancewise important call to mmap will still use that set of > > > flags, no? That's the one backing shared_buffers. > > > > > > The mmap backend for *dynamic* shared memory (aka dsm) is *NOT* supposed > > > to be used on common platforms. Both posix and sysv shared memory will > > > be used before falling back to the mmap() backend. > > > > Hmm, yeah. This might still be a good thing to do (because what do we > > lose?) but it shouldn't really be an issue in practice. > > Is there a reason this was not applied? IIRC, as pointed out above, it's primarily based on a misunderstanding about when mmap is used for in dsm. I.e. that it's essentially just a fallback/toy implementation and that posix or sysv should rather be used. Greetings, Andres Freund -- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services
On Sun, Oct 12, 2014 at 5:36 AM, Andres Freund <andres@2ndquadrant.com> wrote: > On 2014-10-11 20:33:57 -0400, Bruce Momjian wrote: >> On Tue, Aug 12, 2014 at 07:08:06PM -0400, Robert Haas wrote: >> > On Tue, Aug 12, 2014 at 12:59 PM, Andres Freund <andres@2ndquadrant.com> wrote: >> > > On 2014-08-12 09:42:30 -0700, Sean Chittenden wrote: >> > >> One of the patches that I've been sitting on and am derelict in punting >> > >> upstream is the attached mmap(2) flags patch for the BSDs. Is there any >> > >> chance this can be squeezed in to the PostreSQL 9.4 release? >> > >> >> > >> The patch is trivial in size and is used to add one flag to mmap(2) calls in >> > >> dsm_impl.c. Alan Cox (FreeBSD alc, not Linux) and I went back and forth >> > >> regarding PostgreSQL's use of mmap(2) and determined that the following is >> > >> correct and will prevent a likely performance regression in PostgreSQL 9.4. >> > >> In PostgreSQL 9.3, all mmap(2) calls were called with the flags MAP_ANON | >> > >> MAP_SHARED, whereas in PostgreSQL 9.4 this is not the case. >> > > >> > > The performancewise important call to mmap will still use that set of >> > > flags, no? That's the one backing shared_buffers. >> > > >> > > The mmap backend for *dynamic* shared memory (aka dsm) is *NOT* supposed >> > > to be used on common platforms. Both posix and sysv shared memory will >> > > be used before falling back to the mmap() backend. >> > >> > Hmm, yeah. This might still be a good thing to do (because what do we >> > lose?) but it shouldn't really be an issue in practice. >> >> Is there a reason this was not applied? > > IIRC, as pointed out above, it's primarily based on a misunderstanding > about when mmap is used for in dsm. I.e. that it's essentially just a > fallback/toy implementation and that posix or sysv should rather be > used. Perhaps, but I still see no reason not to apply it. It may not help many people, but it won't hurt anything, either. So why not? -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company
On 2014-10-13 10:15:29 -0400, Robert Haas wrote: > On Sun, Oct 12, 2014 at 5:36 AM, Andres Freund <andres@2ndquadrant.com> wrote: > > IIRC, as pointed out above, it's primarily based on a misunderstanding > > about when mmap is used for in dsm. I.e. that it's essentially just a > > fallback/toy implementation and that posix or sysv should rather be > > used. > > Perhaps, but I still see no reason not to apply it. It may not help > many people, but it won't hurt anything, either. So why not? More complicated, less tested code. For no practial benefit, it'll still be slower than posix shm if there's any memmory pressure. But if you want to apply it, go ahead, I won't cry louder than this email. I still think the mmap dsm implementation is a bad idea. We shouldn't put additional effort into it. If anything we should remove it. Greetings, Andres Freund -- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services
>> Perhaps, but I still see no reason not to apply it. It may not help >> many people, but it won't hurt anything, either. So why not? > > More complicated, less tested code. For no practial benefit, it'll still > be slower than posix shm if there's any memmory pressure. But if you > want to apply it, go ahead, I won't cry louder than this email. > > I still think the mmap dsm implementation is a bad idea. We shouldn't > put additional effort into it. If anything we should remove it. While you're not wrong in that use of mmap(2) here is potentially a bad idea, much of that is mitigated through the correctuse of flags to mmap(2) (i.e. prevent mmap(2) pages from hooking in to the syncer). In the same breath, it wouldalso be nice if the following were committed: > --- src/template/freebsd.orig 2014-05-26 23:54:53.854165855 +0300 > +++ src/template/freebsd 2014-05-26 23:55:12.307880900 +0300 > @@ -3,3 +3,4 @@ > case $host_cpu in > alpha*) CFLAGS="-O";; # alpha has problems with -O2 > esac > +USE_NAMED_POSIX_SEMAPHORES=1 -sc -- Sean Chittenden sean@chittenden.org
On 2014-10-13 07:49:44 -0700, Sean Chittenden wrote: > >> Perhaps, but I still see no reason not to apply it. It may not help > >> many people, but it won't hurt anything, either. So why not? > > > > More complicated, less tested code. For no practial benefit, it'll still > > be slower than posix shm if there's any memmory pressure. But if you > > want to apply it, go ahead, I won't cry louder than this email. > > > > I still think the mmap dsm implementation is a bad idea. We shouldn't > > put additional effort into it. If anything we should remove it. > > While you're not wrong in that use of mmap(2) here is potentially a > bad idea, much of that is mitigated through the correct use of flags > to mmap(2) (i.e. prevent mmap(2) pages from hooking in to the syncer). > In the same breath, it would also be nice if the following were > committed: Unless I'm mistaken the pages will still be written back to disk (and not just swap, the actual backing file) if there's memory pressure, no? > > --- src/template/freebsd.orig 2014-05-26 23:54:53.854165855 +0300 > > +++ src/template/freebsd 2014-05-26 23:55:12.307880900 +0300 > > @@ -3,3 +3,4 @@ > > case $host_cpu in > > alpha*) CFLAGS="-O";; # alpha has problems with -O2 > > esac > > +USE_NAMED_POSIX_SEMAPHORES=1 If so, that should be a separate change. But why? Greetings, Andres Freund -- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services
Sean Chittenden <sean@chittenden.org> writes:
> In the same breath, it would also be nice if the following were committed:
> [ use named POSIX semaphores on FreeBSD ]
Really? Why? According to the notes in our code, named POSIX semaphores
are the least attractive of the three Unixoid semaphore APIs we support,
because they require eating a file descriptor per backend per
max_connection slot. That's a lot of FDs in any large configuration.
FreeBSD's support for SysV semaphores would have to be pretty darn awful
to make me think this was a good change, and I've not heard complaints
in that direction before.
If you meant to propose using *unnamed* POSIX semaphores, that might be
a reasonable change, but it would still need some supporting evidence.
regards, tom lane
On Mon, Oct 13, 2014 at 04:19:39PM +0200, Andres Freund wrote: > On 2014-10-13 10:15:29 -0400, Robert Haas wrote: > > On Sun, Oct 12, 2014 at 5:36 AM, Andres Freund <andres@2ndquadrant.com> wrote: > > > IIRC, as pointed out above, it's primarily based on a misunderstanding > > > about when mmap is used for in dsm. I.e. that it's essentially just a > > > fallback/toy implementation and that posix or sysv should rather be > > > used. > > > > Perhaps, but I still see no reason not to apply it. It may not help > > many people, but it won't hurt anything, either. So why not? > > More complicated, less tested code. For no practical benefit, it'll still > be slower than posix shm if there's any memmory pressure. But if you > want to apply it, go ahead, I won't cry louder than this email. > > I still think the mmap dsm implementation is a bad idea. We shouldn't > put additional effort into it. If anything we should remove it. If we have it, we should improve it, or remove it. We might want to use this code for something else in the future, so it should be improved where feasible. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. +
On 2014-10-13 11:18:26 -0400, Bruce Momjian wrote: > On Mon, Oct 13, 2014 at 04:19:39PM +0200, Andres Freund wrote: > > On 2014-10-13 10:15:29 -0400, Robert Haas wrote: > > > On Sun, Oct 12, 2014 at 5:36 AM, Andres Freund <andres@2ndquadrant.com> wrote: > > > > IIRC, as pointed out above, it's primarily based on a misunderstanding > > > > about when mmap is used for in dsm. I.e. that it's essentially just a > > > > fallback/toy implementation and that posix or sysv should rather be > > > > used. > > > > > > Perhaps, but I still see no reason not to apply it. It may not help > > > many people, but it won't hurt anything, either. So why not? > > > > More complicated, less tested code. For no practical benefit, it'll still > > be slower than posix shm if there's any memmory pressure. But if you > > want to apply it, go ahead, I won't cry louder than this email. > > > > I still think the mmap dsm implementation is a bad idea. We shouldn't > > put additional effort into it. If anything we should remove it. > > If we have it, we should improve it, or remove it. We might want to use > this code for something else in the future, so it should be improved > where feasible. Meh. We don't put in effort into code that doesn't matter just because it might get used elsewhere some day. By that argument we'd need to performance optimize a lot of code. And actually, using that code somewhere else is more of a counter indication than a pro argument. MAP_NOSYNC isn't a general purpose flag. Greetings, Andres Freund -- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services
On Mon, Oct 13, 2014 at 05:21:32PM +0200, Andres Freund wrote: > > If we have it, we should improve it, or remove it. We might want to use > > this code for something else in the future, so it should be improved > > where feasible. > > Meh. We don't put in effort into code that doesn't matter just because > it might get used elsewhere some day. By that argument we'd need to > performance optimize a lot of code. And actually, using that code > somewhere else is more of a counter indication than a pro > argument. MAP_NOSYNC isn't a general purpose flag. The key is that this is platform-specific behavior, so if we should use a flag to use it right, we should. You are right that optimizing rarely used code with generic calls isn't a good use of time. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. +
> Really? Why? Because it was found to be lighter weight. See §5 (bottom of ppg 8). https://kib.kiev.ua/kib/pgsql_perf.pdf > According to the notes in our code, named POSIX semaphores > are the least attractive of the three Unixoid semaphore APIs we support, > because they require eating a file descriptor per backend per > max_connection slot. That's a lot of FDs in any large configuration. > FreeBSD's support for SysV semaphores would have to be pretty darn awful > to make me think this was a good change, and I've not heard complaints > in that direction before. > > If you meant to propose using *unnamed* POSIX semaphores, that might be > a reasonable change, but it would still need some supporting evidence. https://lists.freebsd.org/pipermail/svn-src-stable-10/2014-October/003515.html -sc -- Sean Chittenden sean@chittenden.org
On Mon, Oct 13, 2014 at 11:35:18AM -0400, Bruce Momjian wrote: > On Mon, Oct 13, 2014 at 05:21:32PM +0200, Andres Freund wrote: > > > If we have it, we should improve it, or remove it. We might want to use > > > this code for something else in the future, so it should be improved > > > where feasible. > > > > Meh. We don't put in effort into code that doesn't matter just because > > it might get used elsewhere some day. By that argument we'd need to > > performance optimize a lot of code. And actually, using that code > > somewhere else is more of a counter indication than a pro > > argument. MAP_NOSYNC isn't a general purpose flag. > > The key is that this is platform-specific behavior, so if we should use > a flag to use it right, we should. You are right that optimizing > rarely used code with generic calls isn't a good use of time. I have adjusted Sean's mmap() options patch to match our C layout and plan to apply this to head, as it is from August. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. +
Вложения
On Thu, Mar 19, 2015 at 12:16 PM, Bruce Momjian <bruce@momjian.us> wrote: > On Mon, Oct 13, 2014 at 11:35:18AM -0400, Bruce Momjian wrote: >> On Mon, Oct 13, 2014 at 05:21:32PM +0200, Andres Freund wrote: >> > > If we have it, we should improve it, or remove it. We might want to use >> > > this code for something else in the future, so it should be improved >> > > where feasible. >> > >> > Meh. We don't put in effort into code that doesn't matter just because >> > it might get used elsewhere some day. By that argument we'd need to >> > performance optimize a lot of code. And actually, using that code >> > somewhere else is more of a counter indication than a pro >> > argument. MAP_NOSYNC isn't a general purpose flag. >> >> The key is that this is platform-specific behavior, so if we should use >> a flag to use it right, we should. You are right that optimizing >> rarely used code with generic calls isn't a good use of time. > > I have adjusted Sean's mmap() options patch to match our C layout and > plan to apply this to head, as it is from August. Looks great, thanks for taking care of that. -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company
On Thu, Mar 19, 2015 at 12:16:07PM -0400, Bruce Momjian wrote: > On Mon, Oct 13, 2014 at 11:35:18AM -0400, Bruce Momjian wrote: > > On Mon, Oct 13, 2014 at 05:21:32PM +0200, Andres Freund wrote: > > > > If we have it, we should improve it, or remove it. We might want to use > > > > this code for something else in the future, so it should be improved > > > > where feasible. > > > > > > Meh. We don't put in effort into code that doesn't matter just because > > > it might get used elsewhere some day. By that argument we'd need to > > > performance optimize a lot of code. And actually, using that code > > > somewhere else is more of a counter indication than a pro > > > argument. MAP_NOSYNC isn't a general purpose flag. > > > > The key is that this is platform-specific behavior, so if we should use > > a flag to use it right, we should. You are right that optimizing > > rarely used code with generic calls isn't a good use of time. > > I have adjusted Sean's mmap() options patch to match our C layout and > plan to apply this to head, as it is from August. Patch applied. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. +