Hy together,

I have a question I can't find the answer for. I found a new way to export and import Oracle databases. We're running new Oracle 10g databases on RHEL 4 and we're migrating databases (8i) from HPUX which can only be done with exporting and importing the database.

So, "my new way", rather to export the file, move it to the new server and import it there, is as followed:

I created a pipe file (mknod) on the HPUX server and one on the Redhat server. Now, I'm exporting into the first pipe file, writing it's content through an SSH tunnel to the second pipe file and import it to the new database. It has the advantage that I don't have to wait for export to finish bevore import and I can eliminate the disk I/O on this old server which means a massive speed up.

So, the import ist faster than the export (new server beats the old one). But what is if the export would be faster than the export? Where does the exported data go to? Is it stored in the RAM? (Evil with a 60GB database and only 16GB RAM). Or is it stored on the disk? Or, and that would be my favourite, can the export write just so fast as the other server can import?

I often use pipes, but I never did with that amount of data. Can anybody tell my how that exactly works?

Thanks so far and regards.

Tobi

When you fill up ram you will start to use the swap partition, if you have one! (recommended but not compulsory) At that point you will slow down dramatically and you could loose friends if you are on a multi-user network.

Since I'm SysAdmin on our Linux machines and Oracle DBA, I really made no friend. But who cares, I don't need friend at work, I need some in my spare time.

How ever. The "cheat" worked fine. I asked a mate which suggested, as I did, that pipes work synchronous, and it did. The RAM were in use as always, swap wasn't used at all and the export import session took about 10 hours, 6 less than in the usual way.

But thanks anyway for your answer.