Migrating to Oracle on Windows
gooooog...@yahoo.com
I am looking into migrating a database from DB2 on z/OS to Oracle on
Windows.
I am quite new to databases and would appeciate any information on
free programs that do this or a good process for carrying this out.
Thanks,
Noel
Ana C. Dent
@q75g2000hsh.googlegroups.com:
Let's review.
You are "new to databases".
This means that you don't know DB2.
This means that you don't know Oracle.
Please explain what qualifications you bring to this task & why you were
chosen to complete it?
Do you know & can you program in SQL?
Do you expect a point & click GUI to do EVERYTHING for you?
DA Morgan
If your organization has enough money to have DB2 on z/OS there is
essentially no excuse for moving to Windows, definitely no excuse
for looking for free software, and absolutely no excuse for passing
this to someone with zero training.
More information and version numbers required. Before I would offer
help some assurances that this isn't just a troll. Because what you
wrote above is outrageous and an absolute guarantee of failure.
--
Daniel A. Morgan
University of Washington
damo...@x.washington.edu
(replace x with u to respond)
Puget Sound Oracle Users Group
www.psoug.org
Frank van Bortel
Hash: SHA1
Ana C. Dent wrote:
> Do you expect a point & click GUI to do EVERYTHING for you?
Why else would you go to Windows?
- --
Regards,
Frank van Bortel
Top-posting is one way to shut me up...
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (MingW32)
iD8DBQFGUA8GLw8L4IAs830RAqdOAKCIe4SnToYg6vvNpThLcuhldwq+QgCgjkN8
hHexevxCTK34hEnbT1n2FUY=
=WoSq
-----END PGP SIGNATURE-----
gooooog...@yahoo.com
> goooooglegro...@yahoo.com wrote:
> > Hi,
>
> > I am looking into migrating a database from DB2 on z/OS to Oracle on
> > Windows.
>
> > I am quite new to databases and would appeciate any information on
> > free programs that do this or a good process for carrying this out.
>
> > Thanks,
>
> > Noel
>
> If your organization has enough money to have DB2 on z/OS there is
> essentially no excuse for moving to Windows, definitely no excuse
> for looking for free software, and absolutely no excuse for passing
> this to someone with zero training.
>
> More information and version numbers required. Before I would offer
> help some assurances that this isn't just a troll. Because what you
> wrote above is outrageous and an absolute guarantee of failure.
> --
> Daniel A. Morgan
> University of Washington
Hi,
The example I gave of migrating from DB2 on z/OS to Oracle on Windows
is just one example - in general I want to be able to move a database
from Oracle to DB2, and vica-versa, on both Windows and z/OS.
Even a process to move the data from DB2 on z/OS to Oracle on Windows
would be a good start as I already have the database structure setup
on both, but ideally I would like to take a database on DB2 (z/OS) and
create a similiar one on Oracle and then copy across all the data.
I see Oracle have a tool
http://www.oracle.com/technology/tech/migration/workbench/index.html
which handles most of this, however it does not handle migrating from
DB2 on z/OS to Oracle on Windows, hence why I am tackling this
particular setup.
The version of Oracle I am using is 10.2.0 the version of DB2 I am
using is 9.
I would have thought that this would be a common thing to do and thus
would have a solution that was well known among the database community
- so I am just looking for some help\guidance.
Thanks,
Noel
DA Morgan
You haven't given any indication of the amount of data or the type of
application. OLTP or DSS? Megabytes or Petabytes? You've given no
indication of the migration windows or any other business rules. Nor
have you indicated whether what you intend is migrating only heap tables
or whether other object types, code, etc. are involved.
The only thing you have clearly communicated is that a very large and
financially successful organization has turned a critical task over to
someone with zero experience: A recipe for failure.
I have done what you are asking about numerous times. And it is
relatively straight-forward. But I'll not help someone fake their
credentials and ruin a company. You need to do far better to get help
from me and I suspect others here are reacting the same way.
Doesn't your management have a single person on the team who is
qualified to do the work? If not hire a consultant. I could easily
recommend one if you want and if so contact me off-line.
--
Daniel A. Morgan
University of Washington
Mladen Gogala
> Why else would you go to Windows?
Because Windows are, contrary to the popular belief, superior to Linux
in many aspects. One of the things Windows does far better then Linux
is I/O. On Linux, SCSI drivers are an emulation used to access things
like ATAPI DVD burners. That means an interrupt more for every SCSI disk
access on Linux which slows things down. Also, NTFS can do both direct and
async I/O and can be used in a cluster. It's far superior to Ext3.
Linux, on the other hand, has better memory management, primarily because
of MS idiotic refusal to implement shared memory and the usual IPC
primitives. The thread architecture of Windows means that the only way
that that two tasks have to share access to the same part of memory is to
run within the same process, as tasks. In that case, they share the entire
address space and their combined address space cannot exceed 4GB (32 bit).
On Windows, your SGA size is limited by the OS architecture, but
synchronization mechanisms are quite a bit cheaper. Mutexes are not
implemented as a separate device and are quite a bit faster then
semaphores. For a small database, with very few users, Windows will
probably be faster then Linux.
I'm a long time Unix user, having started with now forgotten Ultrix
and having worked with a large variety of exotic Unix varieties like
Wyse Unix, Irix or Bosix (Groupe Bull), I am very disappointed with the
quality of Linux. I am a Red Hat user since the version 5 and I am using
Linux because Unix-like operating systems are more familiar to me then
Windows, but I have no illusions about the quality of Linux.
Unfortunately, overall quality of Windows isn't much better, either.
Unix, on the other hand, is an ancient OS, in desperate need of a thorough
overhaul. SUN realized that and their Solaris is a significant departure
from the accepted Unix standards. Unix still can only return 8 bits as a
status code, Unix still doesn't have record management system and a lock
manager, Unix still doesn't have but a rudimentary system of privileges
and access control, Unix still doesn't do per process I/O accounting and
has sorely inadequate monitoring tool. It doesn't take a genius to
recognize the importance of the question "which process is responsible for
the most of I/O", but it is impossible to answer that question using only
standard monitoring tools on Unix.
OS religious wars are totally besides the point and are now more a part
of the hip hop culture then a part of debate among professionals. I am a
part of R&R generation, not hip hop, I have no objections to someone using
Windows. It's an OS, neither better nor worse then the other ones for the
same platform. If you want X11 server and Perl, there is always Cygwin.
--
http://www.mladen-gogala.com
hast...@hotmail.com
wrote:
> MS idiotic refusal to implement shared memory and the usual IPC
> primitives.
Hmm.. Not sure what you mean here, Mladen.
Memory-mapped files can be used to share memory between processes.
--- Raoul
Frank van Bortel
Hash: SHA1
Mladen,
you totally missed the point.
- --
Regards,
Frank van Bortel
Top-posting is one way to shut me up...
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (MingW32)
iD8DBQFGUUc0Lw8L4IAs830RAnzIAJ4u5de/oRHlOr9JV9D6/yTMIIW9AgCfW13f
Be+ljZeGTHHHVZE0d8MiYTs=
=i0zL
-----END PGP SIGNATURE-----
Mladen Gogala
> Hmm.. Not sure what you mean here, Mladen. Memory-mapped files can be
> used to share memory between processes.
There are no calls like shmat, shmget or shmdet. Moreover, memory mapped
files are not shared in real time. For the change to be visible by other
processes, it has first to be flushed to the disk. This type of sharing
is completely inadequate. What is needed are standard IPC primitives, not
supported by windows. That is why Oracle has multi-threaded architecture
on Windows, as opposed to the independent processes everywhere else.
EscVector
My testing shows Windows operations are faster than linux in exactly
same hardware.
Also, I almost never use a GUI in windows unless it is the better way
to go for a particular task.
Get to know PowerShell. It's very nice.
hast...@hotmail.com
wrote:
> On Sun, 20 May 2007 12:10:51 -0700, hasta_l3 wrote:
> > Hmm.. Not sure what you mean here, Mladen. Memory-mapped files can be
> > used to share memory between processes.
>
> For the change to be visible by other processes, it has first
> to be flushed to the disk. This type of sharing is completely
> inadequate.
I think that is incorrect, Mladen
" With one important exception, file views derived from a single file
mapping object
are coherent or identical at a specific time. If multiple processes
have handles of
the same file mapping object, they see a coherent view of the data
when they
map a view of the file." (http://msdn2.microsoft.com/en-us/library/
aa366537.aspx)
Recall that mapped files are backed up in virtual memory. You need to
flush if
you want to update the underlying file. But people using mapped files
for shared
memory usually dont even have such a file (you can map directly the
paging file)
This said, I would really *love* to be shown wrong...
> What is needed are standard IPC primitives, not supported by windows.
Which primitive do you feel is missing ?
> That is why Oracle has multi-threaded architecture on Windows,
> as opposed to the independent processes everywhere else.
Well, I suspect that oracle chose the standard multi-threaded
architecture for efficiency reasons alone.
For example, some reports on the improvements of bulk collect
lead me to suspect that the PL/SQL -> SQL switch is faster
under windows than under unix.
--- Raoul
Mladen Gogala
> " With one important exception, file views derived from a single file
> mapping object
> are coherent or identical at a specific time. If multiple processes have
> handles of
> the same file mapping object, they see a coherent view of the data when
> they
> map a view of the file." (http://msdn2.microsoft.com/en-us/library/
> aa366537.aspx)
What is a "coherent view"? Does it reflect any changes to the mapped
area immediately? My information tells me that it is not the case.
The "consistent view" means that all users will see anything written
to the file immediately, which is useless.
>
> Recall that mapped files are backed up in virtual memory. You need to
> flush if
> you want to update the underlying file. But people using mapped files
> for shared
Unless you flush, other users will not see your changes. This is not mmap
because Windows doesn't support shared memory.
> memory usually dont even have such a file (you can map directly the
> paging file)
>
> This said, I would really *love* to be shown wrong...
>
>> What is needed are standard IPC primitives, not supported by windows.
>
> Which primitive do you feel is missing ?
Standard IPC primitivies. Look into manual pages under IPC:
shmat, shmget,shmop,shmctl, semget,semop and all the rest:
| ATTRIBUTE TYPE | ATTRIBUTE VALUE |
|_____________________________|_____________________________|
| Availability | SUNWipc (32-bit) |
|_____________________________|_____________________________|
| | SUNWipcx (64-bit) |
|_____________________________|_____________________________|
SEE ALSO
ipcrm(1), savecore(1M), msgget(2), msgrcv(2), msgsnd(2),
semget(2), semop(2), shmctl(2), shmget(2), shmop(2), attri-
butes(5), environ( 5)
The manual page is ipcs on Solaris:
$ uname -a
SunOS lpo-oracle-01 5.8 Generic_117350-39 sun4u sparc SUNW,Ultra-80
$
hast...@hotmail.com
wrote:
> On Mon, 21 May 2007 10:21:40 -0700, hasta_l3 wrote:
> > " With one important exception, file views derived from a single file
> > mapping object
> > are coherent or identical at a specific time. If multiple processes have
> > handles of
> > the same file mapping object, they see a coherent view of the data when
> > they
> > map a view of the file." (http://msdn2.microsoft.com/en-us/library/
> > aa366537.aspx)
>
> What is a "coherent view"? Does it reflect any changes to the mapped
> area immediately?
Yes. Assume process X writes value V at address A. If it reads back
the content of address A, it will get back value V. By the coherency
guarantee above, process B reading address B mapped to the same
virtual memory cell will also read value V.
You may want to browse
http://msdn2.microsoft.com/en-us/library/ms810613.aspx
notably the first sentences of the section on shared memory
> My information tells me that it is not the case.
Well, I would really like to browse a reference
(life teached me to be a die-hard skeptic :-)
> > Which primitive do you feel is missing ?
>
> Standard IPC primitivies. Look into manual pages under IPC:
> shmat, shmget,shmop,shmctl, semget,semop and all the rest:
Unfortunatly, I dont know unix...
Windows synchronization objects include events, semaphores
and mutexes, which can be used across processes.
Communication include shared memory and queues
--- Raoul
Mladen Gogala
> Unfortunatly, I dont know unix...
>
> Windows synchronization objects include events, semaphores and mutexes,
> which can be used across processes.
Not the way POSIX defines them. POSIX IPC primitives very well
defined standard which specifies arguments, behaviors and return
values. Microsoft chose not to implement that. Actually, Microsoft has
a POSIX layer and the corresponding section is here:
Also, this is an additional layer, not native callse, and this software
has a very bad rap. Portability is simply not Microsoft's MO.
Serge Rielau
> I would have thought that this would be a common thing to do and thus
> would have a solution that was well known among the database community
> - so I am just looking for some help\guidance.
One thumb rule that some go by is that any decision on a DBMS for a
given app remains in place for seven years (one may argue the exact
number but long by software standards it is).
After those seven years the app is reasonably dependent on the DBMS and
any change is quite expensive. Hence you see some interesting vendors
popping up like Ants (Sybase -> Oracle)or EDB (Oracle -> Postgress).
The regulars in this group will be the first to tell you that an app
should exploit the DBMS intricacies to the fullest....
Cheers
Serge
--
Serge Rielau
DB2 Solutions Development
IBM Toronto Lab
joel garry
> goooooglegro...@yahoo.com wrote:
> > I would have thought that this would be a common thing to do and thus
> > would have a solution that was well known among the database community
> > - so I am just looking for some help\guidance.
>
> Actually switching data servers is not common at all.
> One thumb rule that some go by is that any decision on a DBMS for a
> given app remains in place for seven years (one may argue the exact
> number but long by software standards it is).
Talk Like Yoda Day, it is! http://www.yodaspeak.co.uk/index.php
> After those seven years the app is reasonably dependent on the DBMS and
> any change is quite expensive. Hence you see some interesting vendors
> popping up like Ants (Sybase -> Oracle)or EDB (Oracle -> Postgress).
>
> The regulars in this group will be the first to tell you that an app
> should exploit the DBMS intricacies to the fullest....
>
I agree that it is not common, but it certainly is a big selling point
among many app vendors.
On the other hand, when it is done, it can be a lot of work. Some
people can get quite lucrative work specializing in such things.
Large organizations are likely to be db-heterogenous and have a
backlog of old apps that could benefit from a migration like the OP
asks about. I got an unsolicited PDP job email yesterday... yeesh.
jg
--
@home.com is bogus.
"Your update subscription (customer number xxxxx-xxxx) expired on
December 31, 1969." - F-Prot
Mladen Gogala
> I got an unsolicited PDP job email yesterday
PDP? Like PDP-11, RSX, KED and alike? Somebody is still using that?
It looks like the old computers never die. They live forever in our
nightmares. Don't get me wrong, I spent significant part of my career
being a VAX/VMS system administrator and, as a student, I was using
FORTRAN and PASCAL compilers on PDP-11. We had some fun with the
DEFINE FILE command and old RZ drives but I am now 46 years old,
quarter of a century after the last use of an RSX box that I can
remember. Anyone who has kept these boxes in a corporate IT after
all these years deserves to be shot, regardless of my nostalgia and
Peter Pan complex.
joel garry
wrote:
I dunno about corporate IT, but a quick google at the pdp groups shows
people are still using them (I assume for data collection or some sort
of control activities).
I have a couple of RSTS machines in my basement (one I bought in 1984,
I think, must have the paperwork around somewhere, saved the other
from a dumpster in the '90s), but I haven't even tried to turn them on
in a decade. I started my paid career on an 11/34. I first saw
Oracle on a VAX around 1983, but at the time, 11/780's still were
better for lots of I/O. Around that time, VAXen started to be cheaper
than the support+electricity for the big PDP's... I think DEC must
have taught Larry something about forcing upgrades...
jg
--
@home.com is bogus.
joel garry
Of course this made me totally curious, so I had to ask:
http://groups.google.com/group/vmsnet.pdp-11/browse_thread/thread/5cebb706bfbe2076/a1ffeff443537e76?lnk=st&q=author%3Ajgar&rnum=1
jg
--
@home.com is bogus.
Last words of Robert Charles Comer before execution: "Yes, go
Raiders."
The Boss
Given the fact that DB2 9 for z/OS GA'd only 2 months ago (March 16th), I
can't believe this is a Real Life issue.
Installing DB2 (or any serious software) on z/OS isn't something you can do
overnight.
If the OP (or his organisation) was able to:
- order DB2 9 for z/OS
- wait for the delivery
- install it
- implement their application incl the database data on top if it
- in spite of all the costs involved so far, decided to go for Oracle/Win
- redesigned/transformed the db layout from DB2/zOS to Oracle/Win
in such a relatively short time, I can't believe they are not able to simply
unload the DB2/zOS data into flat files and transform these a little so they
can be used as input for a load into Oracle/Win.
I would also like to see the bussiness rationale for such an outrageous
move.
The only two times I have been involved in a move like this (from z/OS to
Windows) were cases where the application was being phased out and there was
a need to keep the data in a read-only archive for audit-like purposes.
Neither of those cases required a DBMS change, i.e. one of them was
DB2/zOS --> DB2/Win, the other Oracle/zOS --> Oracle/Win.
--
Jeroen