A Few System Design Principles Pt. 2

OK, we're back to looking at some principles of system design. If you missed yesterday's post, here is the Reader's Digest version: First find the real needs, next determine your timeline. That's it, you're caught up. On to principle number three.

Don't be afraid to re-work the plan. Blow it up if needed.

Last Friday, I met with the business rep at the Apple store. While I was waiting for him, I started cruising the Apple store online. I went to the refurbished section. I came across a MacPro with the same specs as the XServe for $500 less. I also noticed that AppleCare was 1/4 the price it was for an XServe. And I could add simple SATA drives into it easily without having to buy the expensive sleds. And it had 3 full-length PCI-e slots. Hmmm....

I also learned the day earlier that Astaro was having an amazing year-end sale that would save me $1,000 on the price of my firewall. The Apple store also had a re-furb'd iMac that would save another few hundred dollars. Hmm....

I went home and started re-working my plan. I talked with CPC's IT guy and he reminded me to check out Newegg for hard drives. Hey, that saved another $120. And by the way, I discovered that the Seagate drives I was thinking of (750 Gig ES2's) weren't faring well in the user ratings. WD's new RE3's however, looked really good. They cost less, and are specially designed for a RAID. 

After a few hours, I blew up the whole plan. I will spare you the details, but we end up with almost double the storage on our RAID (and it's now a RAID 6 instead of a 5), the RAID will be faster since I found another enclosure with a faster interface, I'll get both AppleCare and Server Upgrade protection, I get more coverage on my firewall, my off-site storage will be easier and faster (by a factor of almost 4), I get a new editing computer that's almost twice as fast as my current one, we get a server that will last us for years and I saved $1,000 off the original budget.

All because I was willing to blow it all up and start again. Interestingly, the basic system is very similar to what I started off with, but every element has been tweaked, upgraded and re-priced to find the best deal. 

Trust your gut.

I had been given the go-ahead to place the order the day before I went to the Apple Store. I could have ordered what I originally wanted right then and there and been fine with the leadership. However, something within me said, "Hold off." I'm really glad I did. I had earlier dismissed re-purposing my current MacPro as a server (as suggested by CPC's IT guy) as too expensive because I have to buy Leopard Server. However, when I re-ran the numbers, it actually worked out better. We don't need the server density afforded by the XServe, and the MacPro will actually be easier for me to work on. And I get a faster editor.

Play with different scenarios.

Say this with me, "Spreadsheets are my friend." If you were to look at the spreadsheet I have built to analyze this, it might make your head explode. But I kept running scenarios until I found one that worked. I have a half-dozen different ways to do network storage. Three ways for off-site backup. Two server platforms. And a myriad of combinations of the above. It's sort of like solving a Rubick's Cube. You twist and turn, spin this way and that, then suddenly, Aha! it all comes together. That was my Friday night (OK, I got obsessed with it...so what).

In the end, it cost me a day. For the next 3-5 years, we'll have a system that we'll grow into gracefully. One that will scale as we grow without having to be re-configured every six months. And we saved money. And did I mention my new 2x faster editor? I think that was worth a day...