I would like to know how you determine when is time to optimize something at the proper location in regards to an ASP.NET application.

We process about one million hits a day on IIS. We have one single Web server and one SQL Server server. There has been a few latency, not enough to affect the work of the users, but enough for me to detect it.

The Web site application pool has 4 worker process loaded. I moved it up to 8 today. However, it is not clear if this is the increase of hits going to SQL Server which is slowing things down. So, there are CPU cores on those servers, memory, IIS configuration and scalability related items. It is difficult to find the perfect combination. I know this can be optimized now before we enable the load balancing. How would you determine the optimized setup? Should I ask the net team to add more cores to the servers, quadruple the memory and so on?

At the IIS level, the application pool in production is stuck in 32 bit. That can only be optimized to 64 bit on the next deployment. I know the application pool set to 64 bits would be a major asset when the volume of hits is large. But, then again, how much a gain of performance would that provide.

So, sometimes, we have about 20 to 40 hits per second. Could the situation of today be really related to the fact that I did not have enough worker process loaded? The hits average 0.6 second in response time. Assuming it would take a second per hit, I assume it would then mean that 8 worker process would not be enough and a queue would be created.

We have 40% more traffic in regards to a 0.01 hit, which is the smallest hit on the Web site. This is a slave hit, such as iPhone, waiting to see if it has to push something to the client. This increase is not what takes time on SQL Server. But, as I got more, I see everything started to be slower. Is it simply that because I would not have enough worker process loaded, that I could end up in this situation? I have those metrics in the ASP.NET application so it seems to indicate that even if the increase of load is pretty much subject to the smallest hit as possible, the fact that too many hits are reaching IIS might be related to a slow down across the entire infrastructure. Could it be? Because, once the hit is started, this is offloaded from IIS, rerouted to .NET, then to SQL Server and back to IIS. I am logging the time of the process. And, that is the part that is getting slower. So, I have some doubts it is because there are too many hits reaching IIS. And, if this is at SQL Server level, it would then mean that the increase of requests has reached a limit.