Colocation as an option to save on the deployment of its own infrastructure and solving technical problems

This can happen with every company. At first, everything works as it should, the equipment functions perfectly, there are no problems and warnings from the monitoring system. But here comes the moment when, for some reason, everything stops. It happens that the reason is the human factor, often unforeseen situations arise, which must happen once in a thousand years. Well, for example, somewhere someone damaged the Internet backbone, and the connection to the network disappeared. Now most companies depend on a wide and reliable Internet channel.

Everything is online - customer information, web applications needed by the company to work. Well, that's all. And suddenly the connection disappears, and with it the access to the information necessary for the normal functioning of the company. And only then you begin to understand that the “cloud” is, of course, very convenient. But if the cloud disappears, then the work of the company can simply be paralyzed. Do you think this can not be? Well, with "Wikipedia" in 2012 happened. Summed up the Amazon data center with a 40-minute crash.

There are original stories. For example, the case when the Internet and communication with the cloud database disappeared in the hospital. The drunk hunter began firing a rifle at an Internet cable passing through the air lines. And hit. The result is the Internet that has disappeared for a long time.
')
Problems with providers happen quite often. If you still have not experienced disconnection from the vital "cloud", then everything can be ahead. But it is better not to wait for a catastrophe, but simply to prepare, so that later it would not be excruciatingly painful.

We, the Kingservers company, recently talked to some customers about emergency cases in their companies. As it turned out, these companies have clear rules in case of loss of the network. Some of these rules I want to bring here, especially since we ourselves use some of the points below.

So, if something happens to a network connection, employees should know what to do:

• Contact the administrator, reporting the problem. Well, this is the most obvious thing an employee can do, right?
• Record the start time of the problem, describing the situation, recording the date, time and other information. All this may come in handy for the support team;
• Send someone to find out what happened, and whether it is possible to solve something in the near future;
• Inform your customers about the problem so that they too are ready. Communication with customers should be kept through different channels so that as many people as possible know about the problem.

Unfortunately, if the central cable is damaged, then almost nothing can be done by yourself. But you can prepare for the problem before it occurs. Here it is worth taking the following steps:

• Connecting to alternative communication channels (optical fiber, twisted pair, wireless networks);
• Use Border Gateway Protocol (BGP);
• Sign a contract with various telecom operators, so if one is unavailable, all the rest will work.

Own infrastructure or collocation?

All of the above, mainly concerns companies that have their own data centers. The main problem is that if a company would like to use redundancy in its infrastructure, this is not always possible for a number of reasons. Home - finance. Installing backup lines of whatever beats is pretty expensive. However, the data center of the company where critical services / applications are hosted should have an agreement with at least two different providers who have different Internet channels. Yes, you have to spend money on equipment and additional services, but sometimes it's worth it. If we are talking about a large company with its own data center, then, in most cases, such a company has the means to maintain its own infrastructure in good condition.

Small and medium-sized companies have a harder time. Most of them do not have money to create a reliable telecommunications infrastructure. Therefore, it is worth advising such an opportunity as a collocation. This service removes the need to develop and maintain your data center. The equipment is hosted by a third-party provider who does all the work. Such a service as colocation, as a rule, is available to almost all hosting providers, albeit large ones, albeit small ones.

The advantage of it, in most cases, is that the majority of colocation providers have a developed infrastructure with redundancy. These include expensive professional routers and switches, optical fiber, and more. As an example, the company Huricane Electric, we did a small overview of one of its data centers.

In addition, if you place equipment somewhere under the responsibility of the contractor, then the contractor will be responsible for everything (but you still have to look at the contract). Judging from the analysis of the situation conducted by Emerson Network Power , the average cost of failure of an unwanted disconnection from the network of a normal company is now $ 8,851 per minute. That's about $ 740357 for each trip (average).

This is a significant amount even for large companies. If you are a SaaS provider, think about how much customers can lose by idle time, and try to increase the reliability of your equipment. Otherwise, in case of failure, customers will start to leave, someone will sue. And the operator will have a “fun time” when management is not engaged in work, but sitting up his pants in the courts (plus we don’t forget about the high costs that accompany any legal process).

As output

Generally speaking, such problems in the IT world occur with enviable regularity. But this few people learn. If you already have your own equipment, then it is best to use the redundancy of various infrastructure elements in your data center. If you don’t have your own equipment or feel sorry for creating your own infrastructure, it’s better to use a colocation in a company where there are several communication channels at once. And then you can not worry about communication with the network and the performance of your services (well, almost).

---

Colleagues, we, Kingservers company, are now looking for IT authors for technical sites (not Habrahabr and not Geektimes, third-party Russian-language and English-language resources). A condition - a person who writes an article (it must be literate and technical) must also have the opportunity to publish it on such a resource. If you are such an author, write in a personal.