Now the market of hosting services is growing rapidly. And change is not only quantitative parameters (volume of disk memory, etc.) but also quality there are fundamentally new services. From a user perspective, it’s all very well, because high competition leads to lower costs and improve the quality of service, but the company is very often under the same name offer very different services.
In this article, I have tried to classify the main proposals that are now on the market. Naturally, the emphasis is on services, with whom I worked, but I think any fundamental aspects are not overlooked.
I also want to warn you that the name of the hosting services is rather arbitrary. Now many companies for promotional purposes manipulate the names as they want. Basically, one way or another, a play on the term “cloud”.So I decided to classify the services of the opportunities that they provide.The result is a next plate.
1) Tight resource constraints
It means that you pay a certain amount of resources and cannot go beyond them. Restrictions can be very different. Amount of disk space, the amount of memory, time and CPU usage script (shared hosting), etc. options. When this value is not dependent on actual resource consumption.Hoster monitors only the fact that you have not exceeded the limit.
There just want to make a reservation at the expense of “unlimited” tariff. Nothing infinite hosting market does not exist. Just sometimes, in order to understand what is the restriction is necessary to carefully read the “Terms of Service”. For example, “unlimited traffic” means in practice that if he exceeds NNNN GB, the speed of the port for your account will be reduced from 100 to 10 MB / s.
If you want to change the amount of resources you have, or to switch to another tariff plan (in the case of shared hosting and VPS), or move to a different server.
Tight resource constraints typical of the “classic” services (shared, VPS and dedicated server). Cloud technologies mean that the host is a system that automatically keeps track of the current resource consumption of your account, and if necessary, one way or another their change (more on that below). The charging is done by actual consumption.
2) Binding to the physical hardware
Quite a controversial option. All the data anyway, they are on some physical servers. The question is, what happens in the event of server failure.
In the case of “classic” WEB HOSTS you will need to move to a different server. Transfer can be done either by you or the service. Hosting support, it does not matter. The point is that for some time the site will be “down”, and all work on the move will be carried out manually.
Using the cloud means that the host is a system that automatically “raise” the site on a different physical server. Technically, this may implement in different ways. For example, Amazon’a you create a virtual machine image (Amazon Machine Images – AMIs) you with the right settings. Their system is able to automatically create instances (virtual server) on THE BASIS OF these images, and to monitor their work. IE if one instance, for some reason falls, it automatically creates a second. Such a system is used for the service “server in the cloud.”
“Hosted in the cloud” usually means that your account is placed in the server cluster, with the possibility of “hot” replacement of equipment (IE without powering down). All information is stored on disk RAID-arrays, IE, the failure of a single server can affect the overall performance, but will not lead to a complete shutdown of the site.
Naturally, there may be situations such as power failure throughout the data center (and such cases have been), but their probability is much lower than the failure of a single server.
It implies the ability to install / uninstall, and configure apsplications. Available only for dedicated servers. Moreover, if the server is in the cloud, then the setup is slightly more complicated than for a normal server. The problem is that the instances are started automatically and their addresses changed. For example, Amazon you need to select the AMI-image (which you want to configure), set up and run it as a new instance, see the address in the control panel, connect to it via SSH and make the desired settings. Then create a new AMI, based on the configured instance.
In the case of WEB HOSTING you need to select one of the configured hoster configurations. Some settings (but not all) can be changed using the control panel. The advantage here is you do not need to spend time configuring web servers, databases, and other applications. But if the configuration host something is not satisfied, the root-access needs.
4) Dynamic vertical scaling
In this case, it means that the resources allocated to your account must vary automatically depending on current load. And all the selected resources must be available to the application. IE Amazon EC2 in this category does not fall, because he just automatically starts new instances (this scaling), and change the type of instance is possible through the admin panel in manual mode.
These parameters have only “hosting in the cloud.” The name may not be very successful, as such services are often referred to as PaaS (Platform as a Service – Platform as a Service).
Consider a couple of examples.
First – Cloud Control. This service uses the horizontal and vertical scaling the same time. Resources are allocated in the form of containers (horizontal scaling) the amount of which may vary (vertical scale) from 1x to 8x, where x = 128 MB of RAM.
Second – rate “flexible” from HostPro. Announced the service last Saturday at the seminar (see slides of the presentation can be here). In this embodiment, three parameters are dynamically changed – disk space (it is believed actually used), RAM and CPU.
5) Dynamic horizontal scaling
Break your system into a smaller structural component Hoster own (without your help) can not, so in this case we are talking only about the change of the number of servers in parallel processing requests.
To implement such services Hoster should provide balancer, IE server, which will distribute incoming requests, and a mechanism for cloning of virtual servers. It is also convenient if the hosting company provides tools for deploya an at code (no need to invent them himself).
However, there is a problem. Technical implementation of such services, each hosting its own. Therefore, moving becomes painful event.
For example, for Amazon’a deploy a code uses a special service – Elastic Beanstalk. Cloud Control provides a special utility that lets you zadeploit code directly from the Git repository. Although it is possible and without utilities since technically on their servers for your project creates a Git repository, which is taken from the code to create instances.HostPro (if I have correctly understood the information on the seminar) provides only the API, IE deploy need to configure yourself.
6) Scaling on demand (manually)
Implemented in the form of a transition to a more expensive (cheapest) tariff plan. At the same time you change the amount of resources available.
Table I indicated that the isolated (dedicated) server does not have this property, but of course you can order another server and move to it. Just compare the transfer to another server with the change of tariff plan is not quite correct.
7) Scaling on demand (manually)
Such a possibility is some cloud hosting, for example, Amazon’a. It operates as follows. You order a certain number of so-called Reserved Instances. This is completely normal EC2 instances, only charge for actual use, but for 1 or 3 years ahead. The price in terms of 1 hour, of course, is much lower.IE if you have a more or less uniform load, then this option allows you to save.
It’s all simple. All the services for which the name is the word “cloud” are more expensive than usual:)
Count on the fact that they are much cheaper, too, is not necessary, because the Included in their development and support of the infrastructure (balancers, system load monitoring and resource allocation,billing, etc.).
But at the same resiliency in clouds above. Furthermore, there is an opportunity to save if the load on the application varies widely. For example, the application used only during the working day.Or, for example, at the site during the holidays attendance increased tenfold.
Unfortunately, universal recipes in the choice, I do not know. Everything is very much dependent on the particular application.
The appearance of the clouds to some extent the situation easier, because Now hosters provide services, which previously was not simple. But often even the banal definition of their value is not a simple task. Really only help to experiment with the settings on the working site.
More nice to see that local hosting companies are trying to catch up with their Western colleagues. The same seminar HostPro made a positive impression. By some measures, they are still lagging behind, but, for example, their CDN looks attractive. Not least because that Amazon’ovskogo CloudFront’a no servers in Russia and Ukraine:)