In order to save some money on licensing costs, our Global Systems Director (a non technical pharmacist!) is proposing rolling up our 3 exisiting Oracle 9i systems onto one server, and proposing we put a new fourth (as yet unknown) large system on the same machine. Don't underestimate his clout within the organisation - he rarely listens to others for advice (and never asks for it).
I'm agog and aghast at the prospect: I can see performance dipping hugely and the prospect of a minor problem taking out all 4 production systems at once.
I need to be fully prepared for when we meet and discuss this. I have no idea what this new fourth system does or how it does it so I can't explain to him how it might affect the other systems. Hows the best way of preparing for this? Are there any obvious arguments I can use? Any good technical resources I can look at?
Tom Kyte (http://asktom.oracle.com) often says one-box-one-instance. That way you'll reduce the wasted disk space and memory associated with many instances on one box. That would alleviate some of the problems of them sharing the same machine provided you structure your security and working practises correctly.
Lots of instances on one box is fine provided the box is sized properly. I would consider using RAC because it works well and it allows for quick scaling should the need arise.
PS. Nothing wrong with not coming from an IT background. I did a PhD in genetic engineering before I became a computer geek.
Didn't say that not having an IT background was the issue. He's a pharmacist - he has no IT experience/knowledge/sense at all. Just knows the business well, and thinks he can direct IT strategy - down to writing specs for systems - based on that. Like I said he has a lot of clout so gets away with it.
I myself am an English Lit graduate, but at least I've worked with Oracle for 14 years since I graduated.
Anyway - point is what is the best way to estimate the effects of moving two systems onto one server? I wasn't responsible for the design of either. One (our main production system) is very stable and we haven't had 1 minutes unplanned downtime in 3 years. A second one really needed to be integrated with this main system but we couldn't afford to revalidate the system so the functionality was built as a seperate system that uses a db link to drag huge amounts of data from the main system and uses views based on this data to do what it does. One proposal is to put these two systems on the same box. The current box has just 4 usable 18Gb disks, One of which houses the archived redo. The datafiles for the system, therefore, are spread only over 3 disks. I would have to spread the second database over these disks too. I'm worried about the contention at the disk head stage, especially when one of the systems is reading large amounts of data from disks that the main prod system is trying to query/write to.
We could buy a larger server I suppose, but how would I need to go about estimating whats needed? What indicators need to be examined so that I can make a realistic evaluation.