There are two aspects to understanding what will fit on a server. How big the server is, and what usage it is experiencing.

In terms of how ‘big’ a server is, we are most interested in memory, disk space, and CPU power. Network connectivity should also be considered, but is usually not a limiting factor compared to the others.

• CPUs are sized in terms of both their speed and the number of cores. The nature of the clearString system and the relational database sitting behind it mean that it works best with around 4 cores, but beyond that extra cores will most likely be idling, so just contributing cost and no benefit.
• RAM memory is more scalable, but again it is possible to over-provision and waste resources.
• Hard disk storage is most scalable. Aside from having sufficient for the data being stored (and backup snapshots etc), one of the most important aspects is to have storage with as fast an access time as possible, as this will become a pinch-point in the time taken to serve information to users.

The other side of the equation is the amount of usage the server is experiencing. Often the question is posed in terms of how many sites, or how many users, a server can handle. Neither metric is a limiting factor. Instead, what matters is how much time the server is spending working responding to user interaction.

This is based on two factors: the number of requests (which for these purposes can be thought of as page requests, but there are other requests too), and the complexity of those requests.

For example, a site might experience 360 page requests an hour, each taking 0.1 second to respond to. That is 36 seconds of work, in an hour. So in that case, working flat-out, the server could handle 100 sites that are that busy. In practise, requests are unlikely to come in at an even rate, so allowance for burstiness needs to be made, so perhaps halve that to allow 50 sites of that usage. But this is just an example: every site will be different, both in terms of usage, and in the complexity of requests (some will be much quicker, a few might take longer), so it may be quite reasonable to fit 100 sites onto a single server, while in other cases you may want to have a server dedicated to a single site.

 

 

Note: The system provides measurement tools to monitor how long requests take to process, and it also provides a range of indexing and caching tools to speed up processing.

What it doesn’t depend on directly are the number of Users visiting a site, or the number of Pages or Components in a site. (You could charge based on these metrics, but they don’t in of themselves make much difference to server loading).