sgmr* vs. md*
I just changed the call from mdunlink to smgrunlink, but this brings up
a good point.
smgr is a generic i/o interface layer that allows multiple storage
managers. Currently, we always use DEFAULT_SMGR as a parameter to smgr*
functions, causing calls to the md* routines. Is there any value in
just removing the smgr layer completely. It was originally for a CD
jutebox i/o layer in addition to our current disk i/o layer.
--
Bruce Momjian | http://www.op.net/~candle
maillist@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026
On Sun, 16 May 1999, Bruce Momjian wrote:
I just changed the call from mdunlink to smgrunlink, but this brings up
a good point.
Thanks, I should have noticed that myself...
smgr is a generic i/o interface layer that allows multiple storage
managers. Currently, we always use DEFAULT_SMGR as a parameter to smgr*
functions, causing calls to the md* routines. Is there any value in
just removing the smgr layer completely. It was originally for a CD
jutebox i/o layer in addition to our current disk i/o layer.
I think that extra layer is a very good idea. Some new kind of storage
might come along that someone wants to use, and md.c wouldn't do the right
thing.
Since it's such a thin layer, performance doesn't really suffer. Doesn't
hurt to keep it...
Ole Gjerde
smgr is a generic i/o interface layer that allows multiple storage
managers. Currently, we always use DEFAULT_SMGR as a parameter to smgr*
functions, causing calls to the md* routines. Is there any value in
just removing the smgr layer completely. It was originally for a CD
jutebox i/o layer in addition to our current disk i/o layer.
Wouldn't this be the interface for a tablespace i/o manager ?
A tablespace has the advatage of only needing a number of files
for thousands of small tables, and reduces the overhead of many
open file handles. A tablespace is also needed before raw devices
can be efficiently exploited.
Andreas
Import Notes
Resolved by subject fallback
smgr is a generic i/o interface layer that allows multiple storage
managers. Currently, we always use DEFAULT_SMGR as a parameter to smgr*
functions, causing calls to the md* routines. Is there any value in
just removing the smgr layer completely. It was originally for a CD
jutebox i/o layer in addition to our current disk i/o layer.Wouldn't this be the interface for a tablespace i/o manager ?
A tablespace has the advatage of only needing a number of files
for thousands of small tables, and reduces the overhead of many
open file handles. A tablespace is also needed before raw devices
can be efficiently exploited.
Good point.
--
Bruce Momjian | http://www.op.net/~candle
maillist@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026