Oracle backup to networked disk (Windows)

Question

I'm trying to back up my database to a Windows share. Currently, Oracle is running under the SYSTEM account, in a two node cluster that uses Fail Safe. I know it should be running as Administrator, but when I set the service to run as Administrator, it only starts the service, not the database. At that point, I can log in through sqlplus to start the database manually, but when I tried that it was down a couple of hours later, which was probably the next time someone tried to use it.
I've done some checking on permissions. I know Administrator has full control over all the datafiles and everything I checked in the ORACLE_BASE folder (c:\oracle). I've started a service request, but was hoping someone here might have some insight for a faster answer. I have had a few less than stellar experiences with service requests recently.

Answer 1

Our database is currently a little above 1TB. We are currently moving from 147GB 10k disks to 300GB 15k disks. We will probably make another improvement to our storage before the end of the year, and still use up most of everything. There is even talk of implementing a system that would take ~35TB of disk space. Unfortunately, I'm just a developer who got stuck with the DBA work.
Did it occur to you it is much cheaper to
BUY EXTRA DISKSCheaper than what? Writing software correctly so you can do a backup directly to another machine, in case the power supply blows or some other disaster fries all the drives? That doesn't sound cheaper. Adding another clustered RAID array isn't my idea of cheap.
I consider requiring a customer to buy extra disk space to write a file so it can be moved elsewhere a stupid design, especially at the insane price of Oracle software. If I can get this working, impress everyone with the volume of data being thrown around, and show a speed issue with using the Windows share, then I'm sure buying more disk is an option.
However, telling someone that because of software issues, they need to spend another several thousand just so they can use their current, valid backup solution that cost a few thousand will make them less than happy. Me basing this on, "I can't get this to work," also sounds like a bad idea. It took them 6 months to get a backup solution better than sticking a USB hard drive on one node.
I'm aware that running an export isn't a good way to backup data, but it beats not having any other copy of your data at all. At least, it would if it would run without getting an "error in writing to export file."