T.R | Title | User | Personal Name | Date | Lines |
---|
788.1 | Yes | MAIL::DUNCANG | good 'O' boys make me throw up | Thu Nov 08 1990 19:53 | 4 |
| Yes, but oracle probably does not have the kind of i/o bottlenecks
that would warrant the use of disk striping. You can specifiy that
Oracle tables be split among many physical storage areas which would
control any i/o per second and transfer rate bottlenecks.
|
788.2 | Stiping | TRCO01::MCMULLEN | Ken McMullen | Thu Nov 08 1990 20:42 | 7 |
| What is the Oracle page size? If it is only 1024 or 2048 it most likely
will not benefit from disk striping. There is a paper available on the
network (if I can find the reference I will post it) that looked at the
benefits of disk striping Rdb. The conclusion was that in most
cases there would not be a benefit. This should be the same for Oracle.
Ask Oracle ... I would like to hear their answer!
|
788.3 | ORacle Pages aka Blocks
| HGOVC::DEANGELIS | Momuntai | Fri Nov 09 1990 03:52 | 9 |
|
Ken,
Oracle pages are known as Blocks and their sizes are controlled by an Oracle
parameter DB_BLOCK_SIZE. These can be adjusted on a per database basis.
For Ultrix the range of block sizes available are 512 - 8192 bytes.
Regards,
John.
|
788.4 | Disk striping and Ingres | ARRODS::FINNI | | Fri Nov 09 1990 11:21 | 9 |
| On a related topic, I have just finished working on a benchmark with
Ingres, and their people were very keen to use disk striping for the
disk containing the log file (=before-image journal). They claimed that
it had given them significant improvements in other benchmarks. Ingres only
has one such file per database, so it is an obvious bottleneck.
However, we didn't use it in the end because they couldn't get V6.3 to
work with VMS V5.4 and had to go back to V5.3, so I can't report on its
value.
|
788.5 | Striping also helps small transfers | WIBBIN::NOYCE | Bill Noyce, FORTRAN/PARALLEL | Tue Nov 13 1990 17:12 | 13 |
| Re .2 - striping with small transfers
Disk striping helps in two different situations. When transfers
are much larger than the stripe interval, striping speeds up each
I/O by transferring two or more streams of data in parallel. This
is the traditional benefit seen in supercomputer applications.
But striping can also help when a disk is being beaten to death
with many small I/O's. It does this by distributing the I/O's about
evenly across two or more disks, thus reducing the load on any single
disk. I would expect Oracle to achieve this benefit of striping.
Note that Rdb/VMS can already spread I/O across multiple files (on
multiple disks), with a bit more intelligence than striping provides,
so it doesn't really need striping to benefit.
|
788.6 | Oracle has physical areas too | HGOVC::DEANGELIS | Momuntai | Wed Nov 14 1990 03:40 | 24 |
| � <<< Note 788.5 by WIBBIN::NOYCE "Bill Noyce, FORTRAN/PARALLEL" >>>
� -< Striping also helps small transfers >-
� But striping can also help when a disk is being beaten to death
� with many small I/O's. It does this by distributing the I/O's about
� evenly across two or more disks, thus reducing the load on any single
� disk. I would expect Oracle to achieve this benefit of striping.
� Note that Rdb/VMS can already spread I/O across multiple files (on
� multiple disks), with a bit more intelligence than striping provides,
� so it doesn't really need striping to benefit.
Note that Oracle has a similar feature to Rdb's physical areas. In Oracle
they're called TABLESPACES. A table can be placed in a named tablespace, or
the default SYSTEM tablespace. A tablespace can consist of one or more physical
files. If a tablespace fills, you can add files to the tablespace. The files
can, of course, be located on any disk/disks you choose. You cannot partition
data across files in a tablespace using column(s) as easily as in Rdb, but it
may be possible using Oracle clusters. You can however, spread rows across n
files of a tablespace by using a % free parameter for the tablespace.
Therefore I don't believe Oracle will benefit from striping if the database is
designed optimally.
John.
|
788.7 | | BNKSIA::MORAN | Vincent Moran | Wed Nov 14 1990 06:28 | 7 |
| Re .6
JDA you are a man of many caps. Its good to see that we now have an
ORACLE expert to call on !
Vince ;^)
|
788.8 | who're you pointing at... | HGOVC::DEANGELIS | Momuntai | Thu Nov 15 1990 04:58 | 8 |
| � <<< Note 788.7 by BNKSIA::MORAN "Vincent Moran" >>>
� ORACLE expert to call on !
^^^^^^^^^^^^^
Hey Vince let's not get carried away! Next you'll be saying I'm an expert in
Cantonese cuisine :-).
John.
|
788.9 | RDB and Stripping | TRCO01::MCMULLEN | Ken McMullen | Mon Dec 10 1990 19:16 | 3 |
| The document on STRIPPING and RDB can be found at:
COOKIE::AIM$PUBLIC:[STRIPPING]RDB_STRIPPING.LN03
|
788.10 | No (strip)-ing allowed, (stripe)-ing only! | MBALDY::LANGSTON | assimpleaspossiblebutnotsimplr | Mon Dec 10 1990 21:54 | 11 |
| Re: .9
"stripping" should be "striping",
i.e.
The document on STRIPING and RDB can be found at:
COOKIE::AIM$PUBLIC:[STRIPING]RDB_STRIPING.LN03
Bruce
|
788.11 | "strip" the mail header | MBALDY::LANGSTON | assimpleaspossiblebutnotsimplr | Mon Dec 10 1990 22:24 | 4 |
| The file COOKIE::AIM$PUBLIC:[STRIPING]rdb_striping.ln03 has a mail header on it.
You will have to "strip" the header before you print the file.
Bruce
|