Let us break this down.
Let us take performance first. A million hits per month is such a low volume that you could use pretty much any old chip to handle that. You can run benchmarks on your laptop or desktop to see if that chip can handle your peak loads. Don't forget to populate your database with a few months worth of data first. Most likely your disk performance will be a bottleneck, not your CPU.
Then you are concerned with reliability, i.e running for months and months. Even with the most expensive systems, there are other factors than the CPU that will decide your reliability.
Most important is availability of disk storage, which is the most likely thing to fail. You want to ensure you don't lose any data if disk fails. This is commonly achieved by using RAID, either mirror, rAID10 or RAID 5. Next, you want to prepare for disaster, so that you still have your data if the whole datacenter goes down. If you don't want to lose any data at all, you need to replicate your database to another site. If losing half a day or so is no big deal, just schedule backup over network instead.
If your host goes down, either because teh power supply blows, network card, memory , cpu cooing fans or other components fail, you need to have some sort of failover mechanism. This is generally achieved by running in a cluster, where 2 or more systems are hooked up to the same storage. To set up a cluster, you simply define file systems, ip adresses, and applications start/stop/monitoring scripts, and when your server blows, the cluster will ensure that the end user wont even notice. I'd say investing $1k or more for an expensive Xeon does not give you any more reliability. Use the money on buying a 2nd server for standby instead. Or if the hosting company is providing the HW, familiarize yourself with what High availability solutions they offer. If they are professional, they would have this type of support down to a science, including off-site backups, and quick recovery from a server failure. As long as the system options they offer has sufficient performance, then you should be able to sleep at night. If they don't offer any of these features, but only sell standalone server with internal storage, then you should write your own solution, and back up the full OS and application config once, and grab a copy of your DB frequently, so that you can redeploy at same site or different site, quickly. Spending extra dollars on powerful chips does not give you any of the piece of mind you are seeking. Replacing a chip in a server is done in a few minutes, trying to recover from disk failure if you have no backup/recover strategy would take you days, or knock you out forever, or in the case of a business bring them to the brink of bankrupcy. Either your data and uptime is so valuable that you would invest the extra time and money to ensure continuous uptime, or you can go with cheaper solutions such as hoping for the best, but have frequent backups. n Define your requirements, and how much you want to spend, and if that is not enough, pick what features are most important to you. If you set up everything yourself, and run the servers on your own network, you could be up and running with a 2 used desktops or laptops un a cluster for $2-300, plus whatever you need for storage. You can even use internal storage, and make frequent replication from server 1 to server 2 and not even have to get a NAS or other shared storage.