How to choose a VPS service among many hosting companies, on whose servers the site will work as comfortably as possible? There is no absolute right guide, everything is relative. But there are several points that you should definitely pay attention to, as well as those that may, on the contrary, misinform you.
Since the VPS (virtual server) shares the resources of the node (server) with other VPS, and each VPS has some kind of load, that is, some of the resources are consumed, the question arises - do we get all the resources declared in the tariff. In fact, the answer is almost always "yes". If you ordered one processor core, then you will always have it, but the performance will be different. And it depends on three main factors:
- Physical server parameters (nodes);
- How "densely" the virtual servers are placed on the node;
- The type of virtualization to use.
Physical characteristics have not changed much lately, except for High CPU tariffs, when small servers with maximum processor performance are used.
The "density" of placement determines how much lower the CPU performance on your VPS will be compared to an empty server.
Much depends on the type of virtualization. Software (such as OpenVZ, Virtuozzo) are not able to localize resources within a single virtual server (in this case, a "container"). It is not uncommon for resource consumption within one container to have a very large impact on the operation of all other containers. Highly deprecated type of virtualization and deprecated. The second type is hardware virtualization (KVM, XEN, VMWare), in which resources are more intelligently and clearly distributed between VMs (virtual machines).
Many webmasters draw conclusions based only on the speed of their site. But almost always these are erroneous conclusions due to the fact that very rarely webmasters set up servers on their own. As a rule, VPS sets up the support of the hosting company or even the server is already configured in the template used. It would be incorrect to say that the server is configured incorrectly. It would be correct to say that the server is not configured for a specific site and its load. If you want to comparing performance in this way, then you need to choose the same OS everywhere and configure everything yourself, without using ready-made pre-installed templates.
Synthetic VPS performance tests also don't give you any particularly useful information, because they only measure peak values, not even averages. And this says absolutely nothing about how neighboring VPS can influence you. In some cases, lower peak values indicate reasonable limits on purpose, which, in the event of an abnormal load, will not allow disrupting the performance of all other VPS on the node. So these are just "parrots" that it is wrong to run after.
The only recommendation that can be given from all of the above is not to use software virtualization. Everything else is built on trust and the company's attitude towards the client.
Also, do not forget to choose a location that is closer to the audience of your site.