Microsoft Office SharePoint Server (MOSS) has seen tremendous take-up rates, not just here in Singapore, but all over the world. I came across many kinds of architectural topology designs during field work with my customers.
One of the most common confusions I have come across arises from the topology design when it comes to the deployment of the MOSS Farms. Some people advocate that the Index server be placed in the same box as the Web-Front-End (WFE) servers with the view that idle power is wasted resources.
There is logic behind this:- Indexing/Crawling not just takes up processing power, it places a certain load on bandwidth as well. The typical organization will set it to crawl at night when online transactions are fairly low. Of course, this is very subjective and differs from environments to environments. A huge load doesnt mean just crawling thousands of websites and fileshares but sets it at a very frequent and short period for incremental crawling. Most of my customers set it to crawl 2-3 times a day (morning, lunch-time, wee-hours at night). That is really not a huge load.
If those are your requirements, you may think having a dediated Indexing Server may be an overkill. If you can bulk-up one of the WFE servers on roids (using a combination of RAM and CPUs), that particular WFE may be able to double-up as a Index server. This will save some costs as well as processing power on that one machine that is just expected to work 3 times a day. Mind you, that machine is expected to be of some decent build as well.
An indexing server can neither be load-balanced nor clustered. What most people want is availabilty at the Query servers, not the Index servers. The built indices are propagated to the Query/Search servers.
There are certain things you need to be aware of, especially when it comes to high-availability in your environment. As said in the referenced link, if an Index server doubles up as a Query service as well, it will not propogate its indices to the other Query servers - that is the gotcha. So, if you have already scoped out 2 WFEs (with the Query service ON) and 1 Index server together with a pair of clustered SQL boxes (one of the most common 5-server MOSS setups), you should try to see if you can run WFE on that Index server and turn the Query service of that particular server OFF. If you can successfully do that, you would have a 3 WFE, 2 Query and 1 Index logical server deployment. You should have the best of both worlds this time around. Hardware resources on that Index server will be more efficiently utilized and you can take one of the WFE servers offline and still enjoy redundancy.
Do take note that this is my own personal advice only. No two customers have exactly the same functional and non-functional requirements and a few of the cases I have seen actually run the crawlers on hundreds of sites, thousands of documents and fileshares and build up the indices up to 10 times a day. In those cases, you should have a dedicated Index server and not have any resource processing contention issues with it. Follow the principles to make sure you max out the hardware resources and costs.