I am new to a corporation that is running Redhat Enterprise AS with Oracle on a 16Tb disk array. They need to expand and are moving to a 50Tb disk array. I've been tasked with making sure Redhat Enterprise AS is still the right choice. They are sticking with Oracle and so I'm just looking to get honest reviews on when Redhat Ent. AS is good, when it is bad, and why it is always better than the alternatives. I'm on a time crunch so I admit to not doing extensive internet searches.

Any help you guys can give from personal experience would be great. I would also appreciate being pointed to websites or persons that have already looked at this issue

You're running a 16TB database? On x86? I'd like to hear details of that. I'd have thought you'd not have enough processing power. Are you doing RAC? A server farm?

Or do you mean you have multiple instances sharing a single 16TB array? This seems more likely. If so then it would depend a lot on the size of each instance as well as how many instances you run on each system.

We run our large (2TB soon to be 4TB) Production system and the related test/development/training instances on HP-UX PA-RISC. For performance in Production we use the EMC DMX array. The test/development/training we run on Clariion ATA because peformance isn't quite as important there.

The most important thing for performance however is the disk array you use and the way you lay it out. Also of course the disks make a difference. A drive running at 15k rpm will be better than one running at 10k rpm. A fibre connected drive will outperform a ATA drive. (Clariion offers both). Having a high cache configuration on the array helps a lot because it will report the writes as complete faster than going directly to disk then do the disk write in the background. Also it allows reads to find much data in cache preventing the slower read of a disk fetch.

Processing power is still important and there's a max you'll get on a single Intel box even with dual cores (today - multicore CPUs are going past dual core so this is likely to change). Another important factor is the memory on the host itself. Oracle SGA should be as large as possible as it has its own caching and buffering mechanisms. In fact on our HP systems we've disabled buffering on Oracle filesystems because it actually interferes with the efficiency of the SGA.

We do use RH AS 3 for a RAC installation of about 300 GB and so far don't have a lot of performance issues with it. We run it on Clariion fibre drives.

Yes. We currently have two instances and are upgrading to the same with 2 50 TB instances.

I'm getting quotes on servers now. We are looking at ten servers with 4 dual core 64-bit processors in each. Each will have 2 - 4GB Fibre channel cards and 4 -1GB ethernet ports.

Can you forsee any reason that RH AS will not scale to this level?