I've a database running on a windows server I'm monitoring with Hyperic. I see the writes per minute ramping up during the day as we approach our peak load. The peaks are increasing gradually over time as we add more users to the system. With other resources (ie CPU, Memory or Disk Space) it's pretty easy to see what the maximum would be and make sure we get additional resources ready before we run out. Is it possible to get even a rough prediction of what the maximum writes per minute is likely to be? Alternatively, do I just need to puh the number of disk writes up and see when things start failing, I assume my disk queue will be the best indicator that I'm exceeding the capabilities of the drive/array.
By way of some additional details, I'm running SQL Server 2005 and the reason I'm so interested in the writes is that the writes on the disk currently outstrip the reads by about 30x. I'm assuming that writes then are likely to be a bottleneck before any other disk activity. The application it's supporting is mainly OLTP although we do report from the data in there as well.