Sometimes things cross your desk that make you just stop and wonder if anyone else just stops and wonders…

I was asked to provide a quote for some NT servers today. 6 of them. So I did (25K total). I was looking at the rest of the budget for this project. The rest of the hardware for it is all VMS / AIX. Big iron stuff. You know, fully redundant, huge IO, lots of … important stuff. Part of the budget for the project was to put more RAM in some of the existing VMS / Galaxy hardware. An extra 2GB/server being used, and there are 2 servers being used from our cluster. So, 4GB of RAM.

What’s the cost?

75,000$/GB!!!

Yes, more than 300,000$ will net you a cool 4GB of RAM.

Which got me to thinking. What kind of cluster could I build in NT for that kind of cash? In fact, I took the total hardware budget for the project (2,500,000$) and built out a nice little solution:

15 x Dell 4800′s for data warehousing / versioning: 400,000$
SAN / Shared Storage solution (100TB): 100,000$
150 x Dell 1850′s for application: 450,000$
100 x Dell 2850′s for database / abstraction: 300,000$
50 x Dell 1850′s for interfacing: 150,000$
10 x load balancers: 500,000$
5 x condenser / eq’s: 400,000$
40 racks: 90,000$
“Stuff” for 40 racks: 120,000$

So, yeah. Even if half of the hardware failed we’d still be well over the current processing power, redundancy, storage…

I know there are lots of reasons to use the big iron, I’m not arguing that. But, we could easily run VMWare and emulate VMS / AIX in Windows 2003 so that software didn’t need to be rewritten.

…

Just a thought. I mean, as I said, when you’re spending 300,000$ on RAM sometimes you just gotta wonder.