Обсуждение: Does PstgreSQL run on the Cray XMT?
Hello,
I work for Cray, the supercomputer company, and we have a new machine
that has some unique remote memory latency hiding features and a very
large shared memory. <www.cray.com/products/xmt>
I'm new to datebases, but would like to try and port PstgreSQL to
this machine. First, has anyone else looked at PstgreSQL on the
Cray XMT? Second, where can I get a tar ball of the source and some
instructions on how to compile the code?
Strenski
Dave, If you don't get a response here after a day - you may want to do a quick post to either the General or Hackers list. Sounds pretty cool ....... On Fri, May 21, 2010 at 4:01 PM, Dave Strenski <stren@cray.com> wrote: > > Hello, > > I work for Cray, the supercomputer company, and we have a new machine > that has some unique remote memory latency hiding features and a very > large shared memory. <www.cray.com/products/xmt> > > I'm new to datebases, but would like to try and port PstgreSQL to > this machine. First, has anyone else looked at PstgreSQL on the > Cray XMT? Second, where can I get a tar ball of the source and some > instructions on how to compile the code? > > Strenski > > -- > Sent via pgsql-novice mailing list (pgsql-novice@postgresql.org) > To make changes to your subscription: > http://www.postgresql.org/mailpref/pgsql-novice > -- Mike Ellsworth
On Fri, May 21, 2010 at 4:01 PM, Dave Strenski <stren@cray.com> wrote:
Hello,
I work for Cray, the supercomputer company, and we have a new machine
that has some unique remote memory latency hiding features and a very
large shared memory. <www.cray.com/products/xmt>
I'm new to datebases, but would like to try and port PstgreSQL to
this machine. First, has anyone else looked at PstgreSQL on the
Cray XMT? Second, where can I get a tar ball of the source and some
instructions on how to compile the code?
Check the website as a start. There is a download link. On linux/unix, building is often as simple as extracting the tarball and then doing ./configure && make && make install. I have no idea how that will translate to a cray.
Sean
Sean Davis <sdavis2@mail.nih.gov> writes:
> On Fri, May 21, 2010 at 4:01 PM, Dave Strenski <stren@cray.com> wrote:
>> I'm new to datebases, but would like to try and port PstgreSQL to
>> this machine. First, has anyone else looked at PstgreSQL on the
>> Cray XMT? Second, where can I get a tar ball of the source and some
>> instructions on how to compile the code?
>>
> Check the website as a start. There is a download link. On linux/unix,
> building is often as simple as extracting the tarball and then doing
> ./configure && make && make install. I have no idea how that will translate
> to a cray.
If it's a Unix-oid system then configure/make should pretty much work.
(If it's not, a port is going to be somewhere between impractical and
impossible, I'm afraid.) What you are mostly going to have to add,
if as I suspect this is a Cray-specific CPU type, is an implementation
of the spinlock primitives. Look in src/include/storage/s_lock.h
for the existing ones. You can make the code fall back on using SysV
semaphores as a stopgap, but that's going to be entirely unacceptable
from a performance standpoint.
regards, tom lane
Tom, Sean,
That's for the help.
Strenski
On Thu, 27 May 2010, Tom Lane wrote:
> Sean Davis <sdavis2@mail.nih.gov> writes:
>> On Fri, May 21, 2010 at 4:01 PM, Dave Strenski <stren@cray.com> wrote:
>>> I'm new to datebases, but would like to try and port PstgreSQL to
>>> this machine. First, has anyone else looked at PstgreSQL on the
>>> Cray XMT? Second, where can I get a tar ball of the source and some
>>> instructions on how to compile the code?
>>>
>> Check the website as a start. There is a download link. On linux/unix,
>> building is often as simple as extracting the tarball and then doing
>> ./configure && make && make install. I have no idea how that will translate
>> to a cray.
>
> If it's a Unix-oid system then configure/make should pretty much work.
> (If it's not, a port is going to be somewhere between impractical and
> impossible, I'm afraid.) What you are mostly going to have to add,
> if as I suspect this is a Cray-specific CPU type, is an implementation
> of the spinlock primitives. Look in src/include/storage/s_lock.h
> for the existing ones. You can make the code fall back on using SysV
> semaphores as a stopgap, but that's going to be entirely unacceptable
> from a performance standpoint.
>
> regards, tom lane
>