I'd like to pose this question here and see where it leads. If this is not the appropriate forum, please let me know what is. I'm in a group where we need to implement a data warehouse. Our design so far is to purchase two 840 filers each with one terabyte of storage using the 36gb drives. On top of that we are going to use either a 420R or Netra 1405 sun server as the database server (solaris 8, Oracle 8.1.6). What we need to do is grab files off of a 760 filer that is in CDR (comma delimited record) format and load that data into an instance of Oracle. Our goal then is to use some method of getting the data off of the 'load' server to another server that will be essentially 'read only' for reporting purposes. These so far are separate servers and I'd like to stay away from Oracle Parallel Server (OPS) at this point because of the extra maintenance and design work that would need to be done to handle this. What I would like to know is what methods of data movement some of you have used. In our quote for the filers we have asked for SnapMirror but I have some reservations on how that might work with Oracle because of the SCN's that are in each separate instance of Oracle.
John
On Tue, 19 Dec 2000, Schiel, John wrote:
What we need to do is grab files off of a 760 filer that is in CDR (comma delimited record) format and load that data into an instance of Oracle. Our goal then is to use some method of getting the data off of the 'load' server to another server that will be essentially 'read only' for reporting purposes.
http://www.netapp.com/tech_library/3041.html describes this pretty well.
Using SQL*Loader w/ direct path loading, you should be able to get those records in there extremely quickly, and since you're not interacting with the 'real' instance, the negative side-effects of DPL are moot.
You won't need SnapMirror for this application --
On your 'load' instance, use sqlloader to toss the data in rawdata:/vol/vol1, into a database living on db1:/vol/vol1 db2:/vol/vol2. When you're done, set the tablespaces read only, go to warm backup, create a snapshot manually, revert to r/w, and repeat.
For the 'read-only/reporting' instance, mount the snapshots of db1:/vol/vol1 and db:/vol/vol2 that you took previously, and start it up.. At whatever update interval you need, stop the instance, re-mount with the latest snapshot, then restart. (Works great for backups, and kids love it too)
Not perfect, as it does require bouncing the server occasionally, but it should work fine.
There's some other options, which would weigh heavier on the database side. (would a slightly larger server permit you to load directly into the live database?, etc).
enjoy.
..kg..
-
kevin graham -
Schiel, John