Stable hosting - myth or reality?
Many hosting providers that provide hosting services experience various problems with this type of service. In most cases, this is either “lagging hosting” or a service that takes a lot of financial and labor resources, complex programs for monitoring consumption and constant conflicts with users that exceed the limits. But with the right approach, these factors can be minimized.
In this article I will try to share my experience, primarily logistics, alas, sometimes completely absent, both from beginner hosters and large hosting providers. Probably, with this article I will improve the affairs of competitors, but the market is large, enough for everyone. In the end, the most important goal of our work is not so much to make money as to make at least some contribution to the development of the Internet as a whole, because it is hosting providers who “build” the Internet and this is a big responsibility.
Whether or not to perceive this publication as a guide to action is up to you. I’ll just tell you about the approach we are pursuing, a relatively small hosting provider that has actively begun to build up its client base in this segment of services recently, which is an indicator of the real effectiveness of the methodology. Previously, we also sometimes had problems with stability or a costly part, but now hosting clients practically do not bother us, and we do them :) For statistics - for 1000 hosting clients we now have only 1 request in those. support during the day on average, and the reason is far from a technical problem on our side.
Simple or terrible work of the hosting server can cause even 1 client. No matter how you limit resources, optimize your server, no matter how powerful it is, sooner or later a client will get there, which will certainly cause a problem on the entire server, breaking all levels of your protection against "overload". Powerful servers at first glance increase stability, but lead to the need to place more clients on one server to depreciate, which increases not only the risk of a problem, but also the number of complaints that are sent to your support if such a server is unavailable. And as we know, a server can become inaccessible not only because of overload.
Conclusion - it is better to use several servers of lower performance, instead of one over-performance server. Since it is:
- more reliably (in case of problems with iron or problems with load, there will be problems on a much smaller part of your infrastructure, since the number of less productive units is greater);
- it’s easier to find the source of the problem (not always advanced load monitoring systems clearly indicate the culprit, sometimes you have to look manually, which is easier to do on a less efficient server, where there are fewer clients by definition);
- fewer dissatisfied clients if the server crashes, fewer support requests;
- there is less likelihood of server overload due to the development of projects of current clients, the risk of the need to rotate accounts between servers is reduced (sometimes some client projects can increase the consumption of resources during their development by several times, the number of such happy and successful clients on less efficient servers will be clearly less than on one super-productive one, moreover, the dependence is non-linear!).
In a good inexpensive restaurant you will never find a huge menu, the larger the variety - the more expensive it is to maintain a food base, which must always be fresh and stored in proper conditions. In addition, restaurant "unlimited" (receptions) - this is actually harmful, but not useful. Obviously contribute to the acquisition of excess weight, and the quality of the food itself is undoubtedly lower. Sometimes it is cheaper and tastier to eat by choosing a limited order. Similarly, in hosting. Let us, starting from this, form the basic principles:
- there are no unlimited resources except traffic (everything should be limited, except traffic, since “unlimited traffic” is a concept that is very important in marketing because of the greatest clarity for users among other restrictions, users do not want to think about how much gigabytes per month consumes their site);
- in fact, traffic should not be explicitly limited, leave a path to retreat (all your other restrictions should be imposed by an “invisible” limiter on traffic, so that you do not have to worry about one user scoring the entire server channel without violating other limits , since server resources are consumed during traffic generation);
- minimalism in the variety of tariff plans (by reducing the number of possible options for tariff plans, it will be easier for you to distribute them between servers, the likelihood of customers needing to rotate from one server to another is minimal, and if possible, the gap in the allowable consumption of resources of the minimum and maximum tariff plans should not be very large).
The most reliable deterrent for spammers and other unscrupulous customers who want to use your services is price. This specific contingent expects to “shit” and leave, as you will certainly block it. Accordingly, such people try to pay as little as possible (or not pay at all, but ask for testing) and certainly do not expect long-term cooperation.
Think ten times before hosting for a dollar. You can earn a dollar, but get a hundred problems! Inexpensive tariff plans may be, but not with a monthly payment, but with payment for a long time. Nevertheless, if you want to provide a stable service, and so that your customers are satisfied, there can’t be any tariffs for the dollar, in principle, I will explain why.
As for the test policy, there should not be any tests. The services are tested either by spammers or other ball lovers and lovers of eating the brain for free, to test on your admins. Appreciate your admins, protect them from these problems! The price of hosting today is at the level of a glass of beer, the user can easily find the money to pay for the service for a month. He doesn’t ask to test beer in a pub!
If the user relies on long-term cooperation, then in an exceptional case, with proper justification for the need for testing, you can go to a meeting - to guarantee moneyback, if the service does not work for justified reasons and within the allotted time, but not a free test.
It just so happened that newcomers usually take the cheapest hosting option that is available without knowledge of how much resources they really need. This is the source of all the problems! Complaints begin that “I was transferred to a more expensive tariff, although there is no attendance on the site, the hoster wants to rip off more money”, which is often the case when beginners use heavy scripts such as WordPress, which, with plugins and due user inexperience, can create a significant load to the server even with one visitor on the site. In addition, by providing such tariffs, you simply begin to donate the resources of your server if you do not take the load into account properly. To avoid this, your minimum tariff should include enough resources, both for a business card site and for a site with an audience of several thousand people a day!
But how then to ensure the low cost of the minimum tariff, which is especially critical for a client with a business card site? Everything is very simple! Give the opportunity to get a good price for long-term payment, say when paying for a year, make a 2-fold discount, and when paying for a 3-month period, the price should be the same as when paying for 2 months, if you pay for the service monthly. This encourages even beginners to pay for the service for at least 3 months and reduces the number of failures. However, with the annual payment option, the price of the minimum tariff plan should be at least $ 4 / month. Someone will say that it can be too expensive for a business card website, but a large number of resources included and a guarantee of stability of the service work without fail. In addition, you are not tired of selling a service that also includes the work of your admins, cheaper than a pint of beer?
Advantages:
- You often receive funds immediately for a year;
- fewer refusals from hosting monthly, less forgetfulness of renewals;
- higher income from minimum tariff plans, as they cost more;
- a larger reserve of resources on the server as a whole, due to the fact that not all users of minimum tariff plans consume the limit allotted to them, which increases stability, reducing the likelihood of server congestion by any process in a critical situation, as it will ultimately remain on it more free resources than was planned;
- doubting customers, seeing the prospect of receiving the service very cheaply with a long payment term, in order to check “the truth works fine or not” pay initially for a month, as well as users who do not have the full amount for payment initially for a long period when ordering, but plan extend the service as soon as the money is collected.
The number of domains that can be placed on your account must be limited. First of all, in order to stimulate the user’s transition to a more expensive tariff plan, which provides for the possibility of placing more sites.
But even the most expensive tariff plan should not allow you to add domains unlimitedly.
Because in the absence of restrictions on the number of domains:
- more server resources are used, the probability of exceeding the load limit is greater;
- there is no additional incentive to switch to a more expensive tariff if the load is within the normal range;
- Dovers and other fans of black SEO will become your customers;
- server configuration files are clogged with slag entries.
The rental cost is average for server performance, and in my understanding this configuration is not lower than Intel Xeon E3-1230 / 8-16GB DDR3 / 2-4x1TB HDD, nowadays it is so low that you can recoup the cost of depreciation by more than 2 times by placing it on server up to 50 clients. Of course, you need to stop earlier if the total load of all clients from the server starts to peak at 50%, but so far I have not observed this :) In addition, you also need to determine the maximum possible number of clients with a tariff plan above the minimum. For example, we have only 3 tariffs, while the maximum differs in resources by 5 times from the minimum. Accordingly, among our 50 customers should not be more than 10 with a maximum tariff plan.
If, as a result of placing 50 clients, even half of the server’s resources are not used, it’s okay, you don’t need to be greedy, the payback and the necessary level of income from the server are achieved, oversell is evil. It’s better not to try to save money and fill the server to the maximum, but to install an additional server. Especially with a large number of orders per day (more than 10) - it will be difficult to calculate the real load. This will save you from unnecessary support costs, from the need to rotate clients between servers due to their development (upgrade of the tariff plan) and increase stability. Well, and most importantly, with a critical problem on the server, only 50 clients will experience inconvenience, and not more. By the way, the benefits of a minimum number of clients on one server have already been mentioned above, so I won’t repeat myself.
Sometimes a user may ask to make an individual tariff - to give more resources than the maximum tariff provides, for example, to make 2 maximum tariff plans in one for the price of two. You should not give in to this provocation. Such a user is already initially aimed at high consumption, and large consumers of resources consume at times and in critical cases many times more resources, the progression is far from arithmetic. That is, if we denote the critical load created by the user of the maximum tariff plan with a unit, then the user with the individual tariff described above is likely to create a load of not 2, but at least 4 in a critical case.
In addition, breaking standardization makes it harder to plan resource allocation. The likelihood of problems with a large consumer of resources is much greater than with a small one. Such users can be "tricked" and try to break, since they probably already have a large project with such consumption, the competition is higher. Therefore, you need to clearly understand that such clients have no place on the hosting, they can afford a dedicated server.
Need more resources than the maximum tariff provides? Need unlimited domains, etc.? Welcome to a dedicated server, but not to hosting. Risking the stable work of the remaining 49 people because of the prospect that 1 will pay more is stupid. Let him pay even more and take a dedicated server;)
Unfortunately, the fantasy of some webmasters has been so degraded recently that they don’t even bother to fill their own sites, but simply parse content to their site from other sites, increasing the volume of databases to a gigabyte or more, while the architecture of such a database is not quite perfect character. When a table breaks in a large database or a bunch of processes hang, the server is unlikely to be fine. Yes, and slag sites - this is not the Internet that I would like to see. Create restrictions for fans of slag sites - do not allow them to be hosted, and users with databases of more than gigabytes are better to honestly and immediately refuse service, unless you are completely sure that the user will not kill the server and the SDL portal with optimized scripts.
Yes, amateurs of cheap warehouses will forgive me, but now renting a server in a reliable Data Center with a real guaranteed channel without limiting traffic is cheaper as a result than low-cost airlines, where nothing is guaranteed at all. Desktop iron is not applicable here, unless you want to carry out those. work on their servers several times a year, thereby creating inconvenience to their customers. Well, iron that is not designed for a constant server load cannot work properly in such conditions.
However, some low-cost airlines now offer servers with 64 GB RAM, even server configurations at first glance, and at a low cost. The secret is that some components there are still desktop and this is self-assembly, the server, perhaps, is only a case. Do not chase cheapness and megacities. You also do not need them. Do you know a lot of projects that consume 64 GB of RAM? Most projects will actually have a server with 4-8 GB of RAM with a head, even a lot will be. But why take it when there is 64 GB RAM for less money, many think, this is not cool! This is stupid marketing and nothing more, which, alas, works. Even if the configuration is really server-wide, everything else will be none.
Quality can’t be too cheap, it’s better to take a seemingly less productive server for the same money, but in a good Data Center and ultimately use its resources fully, as there will be all the possibilities. Reliability, speed of replacement (up to the whole server within an hour) and, of course, location is important. It should be a reliable platform with excellent connectivity and support, with an adequate abus department, which will not block your server from the first complaint from someone unclear, but will consider it, analyze it and, if necessary, will contact you and give you time to respond. Loukosetry are not able to maintain the staff of such employees, therefore they can send you absolutely all complaints, whether confirmed or not, not to disdain your disconnection.
Complaints about hosting servers are common. That is why it is not enough to rent a server in a reliable Data Center, it is necessary that the Data Center itself be in a reliable country by law. This will save you from being taken out of the brain by law enforcement agencies, with or without law, since the laws of some countries are moronic, for example, the ban on the placement of adult sites and the criminal prosecution of such webmasters, while in many countries this is absolutely legal. It is also better not to be located in those countries where laws are not always in force, and your server can be seized not only by court order, but simply by request as evidence or under some other pretext.
- the actual load on the hosting servers for the above configuration is within 20-30% with 50 clients on the server, a huge reserve of resources, traffic consumption in the range of 5-25 Mbit / s;
- for every 1000 customers, an average of 1 request per. support per day;
- servers are stable, customer sites always work, problems are an incredibly rare occurrence, the need for time spent on support is minimal;
- exceeding the permissible load due to high limits is an extremely rare occurrence and switching to a more expensive tariff plan does not cause a storm of emotions, since at the time of exceeding the webmaster already makes enough money and is no longer a beginner.
It makes no sense to build cluster, cloud hosting services, to develop ultra-expensive solutions for monitoring the load and accounting for consumed resources. Even cloud hosting services are falling, absolute stability is not possible, but this is possible. In this case, it is useful to apply the Occam Razor principle to hosting: a diversified solution on non-overloaded servers will be much cheaper, more stable, easier to implement, complicating this scheme is the way not to stability, but to problems. Sometimes the simplest solution is the most correct, do not complicate your life without emergency, I believe that it is better to adhere to this principle.
Sergii V. Sikorskyi, co-owner of ua-hosting.com.ua
In this article I will try to share my experience, primarily logistics, alas, sometimes completely absent, both from beginner hosters and large hosting providers. Probably, with this article I will improve the affairs of competitors, but the market is large, enough for everyone. In the end, the most important goal of our work is not so much to make money as to make at least some contribution to the development of the Internet as a whole, because it is hosting providers who “build” the Internet and this is a big responsibility.
Whether or not to perceive this publication as a guide to action is up to you. I’ll just tell you about the approach we are pursuing, a relatively small hosting provider that has actively begun to build up its client base in this segment of services recently, which is an indicator of the real effectiveness of the methodology. Previously, we also sometimes had problems with stability or a costly part, but now hosting clients practically do not bother us, and we do them :) For statistics - for 1000 hosting clients we now have only 1 request in those. support during the day on average, and the reason is far from a technical problem on our side.
No mega-servers for hosting
Simple or terrible work of the hosting server can cause even 1 client. No matter how you limit resources, optimize your server, no matter how powerful it is, sooner or later a client will get there, which will certainly cause a problem on the entire server, breaking all levels of your protection against "overload". Powerful servers at first glance increase stability, but lead to the need to place more clients on one server to depreciate, which increases not only the risk of a problem, but also the number of complaints that are sent to your support if such a server is unavailable. And as we know, a server can become inaccessible not only because of overload.
Conclusion - it is better to use several servers of lower performance, instead of one over-performance server. Since it is:
- more reliably (in case of problems with iron or problems with load, there will be problems on a much smaller part of your infrastructure, since the number of less productive units is greater);
- it’s easier to find the source of the problem (not always advanced load monitoring systems clearly indicate the culprit, sometimes you have to look manually, which is easier to do on a less efficient server, where there are fewer clients by definition);
- fewer dissatisfied clients if the server crashes, fewer support requests;
- there is less likelihood of server overload due to the development of projects of current clients, the risk of the need to rotate accounts between servers is reduced (sometimes some client projects can increase the consumption of resources during their development by several times, the number of such happy and successful clients on less efficient servers will be clearly less than on one super-productive one, moreover, the dependence is non-linear!).
The principle of forming hosting hosting plans is identical to the principle of building a menu of a good restaurant
In a good inexpensive restaurant you will never find a huge menu, the larger the variety - the more expensive it is to maintain a food base, which must always be fresh and stored in proper conditions. In addition, restaurant "unlimited" (receptions) - this is actually harmful, but not useful. Obviously contribute to the acquisition of excess weight, and the quality of the food itself is undoubtedly lower. Sometimes it is cheaper and tastier to eat by choosing a limited order. Similarly, in hosting. Let us, starting from this, form the basic principles:
- there are no unlimited resources except traffic (everything should be limited, except traffic, since “unlimited traffic” is a concept that is very important in marketing because of the greatest clarity for users among other restrictions, users do not want to think about how much gigabytes per month consumes their site);
- in fact, traffic should not be explicitly limited, leave a path to retreat (all your other restrictions should be imposed by an “invisible” limiter on traffic, so that you do not have to worry about one user scoring the entire server channel without violating other limits , since server resources are consumed during traffic generation);
- minimalism in the variety of tariff plans (by reducing the number of possible options for tariff plans, it will be easier for you to distribute them between servers, the likelihood of customers needing to rotate from one server to another is minimal, and if possible, the gap in the allowable consumption of resources of the minimum and maximum tariff plans should not be very large).
Protection against dishonest customers
The most reliable deterrent for spammers and other unscrupulous customers who want to use your services is price. This specific contingent expects to “shit” and leave, as you will certainly block it. Accordingly, such people try to pay as little as possible (or not pay at all, but ask for testing) and certainly do not expect long-term cooperation.
Think ten times before hosting for a dollar. You can earn a dollar, but get a hundred problems! Inexpensive tariff plans may be, but not with a monthly payment, but with payment for a long time. Nevertheless, if you want to provide a stable service, and so that your customers are satisfied, there can’t be any tariffs for the dollar, in principle, I will explain why.
As for the test policy, there should not be any tests. The services are tested either by spammers or other ball lovers and lovers of eating the brain for free, to test on your admins. Appreciate your admins, protect them from these problems! The price of hosting today is at the level of a glass of beer, the user can easily find the money to pay for the service for a month. He doesn’t ask to test beer in a pub!
If the user relies on long-term cooperation, then in an exceptional case, with proper justification for the need for testing, you can go to a meeting - to guarantee moneyback, if the service does not work for justified reasons and within the allotted time, but not a free test.
Hosting should be inexpensive not at the expense of a low limit on resources
It just so happened that newcomers usually take the cheapest hosting option that is available without knowledge of how much resources they really need. This is the source of all the problems! Complaints begin that “I was transferred to a more expensive tariff, although there is no attendance on the site, the hoster wants to rip off more money”, which is often the case when beginners use heavy scripts such as WordPress, which, with plugins and due user inexperience, can create a significant load to the server even with one visitor on the site. In addition, by providing such tariffs, you simply begin to donate the resources of your server if you do not take the load into account properly. To avoid this, your minimum tariff should include enough resources, both for a business card site and for a site with an audience of several thousand people a day!
But how then to ensure the low cost of the minimum tariff, which is especially critical for a client with a business card site? Everything is very simple! Give the opportunity to get a good price for long-term payment, say when paying for a year, make a 2-fold discount, and when paying for a 3-month period, the price should be the same as when paying for 2 months, if you pay for the service monthly. This encourages even beginners to pay for the service for at least 3 months and reduces the number of failures. However, with the annual payment option, the price of the minimum tariff plan should be at least $ 4 / month. Someone will say that it can be too expensive for a business card website, but a large number of resources included and a guarantee of stability of the service work without fail. In addition, you are not tired of selling a service that also includes the work of your admins, cheaper than a pint of beer?
Advantages:
- You often receive funds immediately for a year;
- fewer refusals from hosting monthly, less forgetfulness of renewals;
- higher income from minimum tariff plans, as they cost more;
- a larger reserve of resources on the server as a whole, due to the fact that not all users of minimum tariff plans consume the limit allotted to them, which increases stability, reducing the likelihood of server congestion by any process in a critical situation, as it will ultimately remain on it more free resources than was planned;
- doubting customers, seeing the prospect of receiving the service very cheaply with a long payment term, in order to check “the truth works fine or not” pay initially for a month, as well as users who do not have the full amount for payment initially for a long period when ordering, but plan extend the service as soon as the money is collected.
No tariffs with unlimited domains
The number of domains that can be placed on your account must be limited. First of all, in order to stimulate the user’s transition to a more expensive tariff plan, which provides for the possibility of placing more sites.
But even the most expensive tariff plan should not allow you to add domains unlimitedly.
Because in the absence of restrictions on the number of domains:
- more server resources are used, the probability of exceeding the load limit is greater;
- there is no additional incentive to switch to a more expensive tariff if the load is within the normal range;
- Dovers and other fans of black SEO will become your customers;
- server configuration files are clogged with slag entries.
No more than 50 clients per server under any conditions
The rental cost is average for server performance, and in my understanding this configuration is not lower than Intel Xeon E3-1230 / 8-16GB DDR3 / 2-4x1TB HDD, nowadays it is so low that you can recoup the cost of depreciation by more than 2 times by placing it on server up to 50 clients. Of course, you need to stop earlier if the total load of all clients from the server starts to peak at 50%, but so far I have not observed this :) In addition, you also need to determine the maximum possible number of clients with a tariff plan above the minimum. For example, we have only 3 tariffs, while the maximum differs in resources by 5 times from the minimum. Accordingly, among our 50 customers should not be more than 10 with a maximum tariff plan.
If, as a result of placing 50 clients, even half of the server’s resources are not used, it’s okay, you don’t need to be greedy, the payback and the necessary level of income from the server are achieved, oversell is evil. It’s better not to try to save money and fill the server to the maximum, but to install an additional server. Especially with a large number of orders per day (more than 10) - it will be difficult to calculate the real load. This will save you from unnecessary support costs, from the need to rotate clients between servers due to their development (upgrade of the tariff plan) and increase stability. Well, and most importantly, with a critical problem on the server, only 50 clients will experience inconvenience, and not more. By the way, the benefits of a minimum number of clients on one server have already been mentioned above, so I won’t repeat myself.
No individual tariff plans
Sometimes a user may ask to make an individual tariff - to give more resources than the maximum tariff provides, for example, to make 2 maximum tariff plans in one for the price of two. You should not give in to this provocation. Such a user is already initially aimed at high consumption, and large consumers of resources consume at times and in critical cases many times more resources, the progression is far from arithmetic. That is, if we denote the critical load created by the user of the maximum tariff plan with a unit, then the user with the individual tariff described above is likely to create a load of not 2, but at least 4 in a critical case.
In addition, breaking standardization makes it harder to plan resource allocation. The likelihood of problems with a large consumer of resources is much greater than with a small one. Such users can be "tricked" and try to break, since they probably already have a large project with such consumption, the competition is higher. Therefore, you need to clearly understand that such clients have no place on the hosting, they can afford a dedicated server.
Need more resources than the maximum tariff provides? Need unlimited domains, etc.? Welcome to a dedicated server, but not to hosting. Risking the stable work of the remaining 49 people because of the prospect that 1 will pay more is stupid. Let him pay even more and take a dedicated server;)
No big databases and hosting parsers
Unfortunately, the fantasy of some webmasters has been so degraded recently that they don’t even bother to fill their own sites, but simply parse content to their site from other sites, increasing the volume of databases to a gigabyte or more, while the architecture of such a database is not quite perfect character. When a table breaks in a large database or a bunch of processes hang, the server is unlikely to be fine. Yes, and slag sites - this is not the Internet that I would like to see. Create restrictions for fans of slag sites - do not allow them to be hosted, and users with databases of more than gigabytes are better to honestly and immediately refuse service, unless you are completely sure that the user will not kill the server and the SDL portal with optimized scripts.
Your server must be reliable and located in a reliable Data Center, and not in stock
Yes, amateurs of cheap warehouses will forgive me, but now renting a server in a reliable Data Center with a real guaranteed channel without limiting traffic is cheaper as a result than low-cost airlines, where nothing is guaranteed at all. Desktop iron is not applicable here, unless you want to carry out those. work on their servers several times a year, thereby creating inconvenience to their customers. Well, iron that is not designed for a constant server load cannot work properly in such conditions.
However, some low-cost airlines now offer servers with 64 GB RAM, even server configurations at first glance, and at a low cost. The secret is that some components there are still desktop and this is self-assembly, the server, perhaps, is only a case. Do not chase cheapness and megacities. You also do not need them. Do you know a lot of projects that consume 64 GB of RAM? Most projects will actually have a server with 4-8 GB of RAM with a head, even a lot will be. But why take it when there is 64 GB RAM for less money, many think, this is not cool! This is stupid marketing and nothing more, which, alas, works. Even if the configuration is really server-wide, everything else will be none.
Quality can’t be too cheap, it’s better to take a seemingly less productive server for the same money, but in a good Data Center and ultimately use its resources fully, as there will be all the possibilities. Reliability, speed of replacement (up to the whole server within an hour) and, of course, location is important. It should be a reliable platform with excellent connectivity and support, with an adequate abus department, which will not block your server from the first complaint from someone unclear, but will consider it, analyze it and, if necessary, will contact you and give you time to respond. Loukosetry are not able to maintain the staff of such employees, therefore they can send you absolutely all complaints, whether confirmed or not, not to disdain your disconnection.
Complaints about hosting servers are common. That is why it is not enough to rent a server in a reliable Data Center, it is necessary that the Data Center itself be in a reliable country by law. This will save you from being taken out of the brain by law enforcement agencies, with or without law, since the laws of some countries are moronic, for example, the ban on the placement of adult sites and the criminal prosecution of such webmasters, while in many countries this is absolutely legal. It is also better not to be located in those countries where laws are not always in force, and your server can be seized not only by court order, but simply by request as evidence or under some other pretext.
results
- the actual load on the hosting servers for the above configuration is within 20-30% with 50 clients on the server, a huge reserve of resources, traffic consumption in the range of 5-25 Mbit / s;
- for every 1000 customers, an average of 1 request per. support per day;
- servers are stable, customer sites always work, problems are an incredibly rare occurrence, the need for time spent on support is minimal;
- exceeding the permissible load due to high limits is an extremely rare occurrence and switching to a more expensive tariff plan does not cause a storm of emotions, since at the time of exceeding the webmaster already makes enough money and is no longer a beginner.
It makes no sense to build cluster, cloud hosting services, to develop ultra-expensive solutions for monitoring the load and accounting for consumed resources. Even cloud hosting services are falling, absolute stability is not possible, but this is possible. In this case, it is useful to apply the Occam Razor principle to hosting: a diversified solution on non-overloaded servers will be much cheaper, more stable, easier to implement, complicating this scheme is the way not to stability, but to problems. Sometimes the simplest solution is the most correct, do not complicate your life without emergency, I believe that it is better to adhere to this principle.
Sergii V. Sikorskyi, co-owner of ua-hosting.com.ua