Step 3: More Database Speed, Please
|
|||||||||||||||||||||||||||||||||||||||||||
Main Page Technical Docs Scaling WebObjects Example Apps Step 0 Step 1 Step 2 Step 3 Step 4 Step 5 Step 6 Step 7 Step 8 Summary Postscript |
Its hard to say in general what we need to do to get our database
tpm up. Do we need a RAID array? More RAM? More CPU? Hard to say
in general. In real life, you measure and add more of whatever
you need, and then measure again.
But lets go with a general solution. We're going to get a midrange Solaris box, with Oracle 8 (FrontBase doesn't run on Solaris, so that's not an option, but it served us well in the meantime.) So that will be say, $35,000 in additional hardware, plus another week of engineering, for $41,000 in incremental cost. If we used FrontBase in the previous step, we now need to get Oracle 8, so that's an additional $15,000. At this point, we can double our tpm license for WebObjects up to 100 tpm for $5000. Since we've spent $65,000 so far, this isn't a big deal, nile.com makes that in a couple of hours. To get to the next step, lets look at some economics. If our new setup can max out our WO license, then we're serving 100 pages/minute on nile.com. That's 10 sales per minute. Or $120000/week in profit. Conversely, if we're down a week, we lose $120000 in lost profit.
Plus lost future sales, which we'll ignore. Plus another week
of engineering at $6K as you run around madly trying to rebuild
everything. Total: $126,000/week. Maybe we should start building in some redundancy. Next Page: Step 4, Backups? Who needs backups! Ooops...
|