Re: You guys need to hit Google up for a new serve
Fortunately (as has been said somewhere, I think) the issue is not number of servers.
The issue is re-engineering a very large and complex centralized database to allow synchronized access from <n> servers, for <n> greater than one.
I've worked on some problems with some similarities to this one; to be able to listen to an overview of the process and suggest feasible improvements is a most unusual skill.
In the absense of knowledge of the algorithm, about all you can _know_ for certain is that choice of algorithms is far more significant than number of servers. In fact, more servers added to the wrong algorithm, or a buggy algorithm, can actually slow down the process!
More generally, you should assume that any kind of synchronization is designed for a particular kind of access patterns; and can degrade painfully in the face of distinctively different access patterns. At this point, some suspect (but we do not know) that this is happening also.
Think of a rush hour commuter traffic jam. The "cars" are "servers" carrying "loads." It would be an act of incredible stupidity to look at the commute time and say, "the problem is, there aren't enough taxis out there to get people to work on time." To say, "it's obvious we need BIGGER servers -- that is, buses" is unfortunately more credible, but no less stupid. The fact is, performance problems are often subtle, and an experienced analyst will expect many of the solutions to be counterintuitive. In fact, depending on the KIND of traffic patterns, buses may well make the traffic run slower, or faster, or have very little effect.
One of the things that is going on right now is, technical staff is experimenting with, um, different sizes of taxis and buses, different shapes and locations of parking lots and bus stops, and various other interrelated parameters.