Choosing a hosting: speed and reliability
The hosting industry has existed in Russia for over 20 years. It would seem that there is enough time for the market to stabilize, there are clear leaders and simple selection criteria for customers. However, in solving the problem of choosing a hosting service, it is still necessary to rely on our own experience, recommendations, and “spear method”.
In this article, we will look at a methodical approach to choosing a hosting provider from the standpoint of high-quality web project work, both in terms of speed and reliability. I’ll clarify that we’ll talk only about dedicated servers or VPS (or their cloud counterparts), completely virtual (shared) hosting is left out of the box as an unacceptable option.
When choosing a company providing hosting services, you need to look at several criteria.
Specialization: it is desirable that the company is engaged in the hosting and / or management of data centers. The presence of specialization gives two advantages: higher competencies and result orientation. For example, it is often possible to meet the situation when a telecom company leases its data centers, and this service is not the key for business and receives the appropriate attention from management.
')
Direct provider: it is better to work directly with the company managing the data center. Only in this case the most prompt solution of problems is possible (fewer links in the incident response chain).
Experience: any IT services suggest a dynamic development of the market, so you should not be afraid of new players. However, it’s worthwhile to bypass the companies created less than two years ago: at best, you will encounter “childish” problems, at worst - the company will disappear from the market (examples are Oversan, Makhost and some others).
Scale: the optimal size of the company from small to medium scale. Micro-companies should be avoided, as they may not have the necessary personnel, reserve resources and most likely they will save on quality of services. Large companies will be more difficult to interact with: in the case of a need for an individual solution and generally less sensitive to customer problems.
Openness: an additional advantage will be the active participation of the company in communication with customers. This activity can take the form of blogging, surveys, participation in professional conferences and other events. As a rule, this is a sign of a desire to improve the quality of services and a willingness to engage in dialogue with customers.
Geography: it is important where the company's data centers are located - the data should be as close as possible to users. If distributed infrastructure is required from a single supplier, the company must have a network of data centers.
That data center is key to the quality of your project. No organization or technical support will be able to compensate for poor connectivity or power failures in the DC. To evaluate the criteria for selecting a data center, it is required that the company discloses information about its data center. In case there is no such data, it is worth asking: why is this information hidden?
Connectivity: one of the most important characteristics. It is determined by the number and quality of Internet channels in the data center, as well as the traffic exchange policy (peering). It is important that the data center is located as close as possible to your users (site visitors). Check the network latency to the DC can be a simple ping command, which must be performed from various networks in the region of the target audience. For a more in-depth analysis of connectivity, you can use the global connectivity graph, which is presented on the radar.qrator.net service (Radar by Qrator). This graph shows the routing policies between all autonomous systems (AS) and the Internet. It is worth looking at the graph itself (section Graph), providers, peering. You can also see the major clients of the hoster (those who have their own AS). It is necessary to be afraid of the companies at which the graph of connectivity rests against a single node - this means that the hoster is connected through another network and is completely dependent on its work.
Launch date: preference should be given to new data centers, as now more thoughtful projects are used, more power is allocated, a large bandwidth is laid. At the same time, you need to remember about the possibility of "children's" problems in data centers running "yesterday." For old data centers, it is also possible to replace / maintain infrastructure equipment, which often leads to interruptions in service.
Certification: in theory, the level of certification (Tier II, III, IV) should show the reliability of the data center. In general, the idea is correct, the presence of such certification (Tier II, better Tier III) can be considered as an advantage. However, you need to remember that as a rule, only the data center project is certified. It’s not at all the fact that the project will be fully implemented. In addition, the most important aspect - maintenance and operation - is usually not certified in Russia.
Power supply: the connected power should be enough for your needs (in the case of a rack rental). Redundancy should be organized either in the classical scheme: electrical UPS + diesel generator, or in the form of a dynamic UPS. It is desirable to power the DC from two independent rays. In fact, the reliability of the power supply can be checked only by personal experience (there have been cases of DC shutdown due to UPS).
Cooling and ventilation: there are various standards for temperature in the engine room of the DC. The current trend in order to increase energy efficiency is to increase the working temperature of the air, so it is worthwhile to clarify the needs of your equipment in terms of temperature. It is important that the air conditioning system has a reserve for maintenance and power reserve. In addition, it is important that the air in the engine room is filtered from dust, otherwise the equipment becomes clogged with dirt very quickly.
Additional fault tolerance: for critical applications, additional measures may be required to increase reliability - connecting the server via two power supplies, several independent network links, or organizing fault tolerance between the DC (for example, VRRP).
Availability and cost of channels: when assessing the cost of accommodation, you need to accurately determine the conditions of connection. First, the channel can be with a guaranteed or non-guaranteed band. Secondly, there may be a limit on the amount of traffic or incoming / outgoing ratios. It is also worthwhile to find out in advance the possibility and cost of channel expansion (the standard is 100 Mbit / s).
Processes and security: when deciding on the placement of equipment in the data center, it is useful to visit it in advance and look at the security system, as well as the organization of work with the equipment. The entrance to the computer room should be strictly regulated and limited. In the service area of ​​the equipment - order and cleanliness. If during your visit to the DC you found a mess and lack of organization, you should not expect high availability.
Actually, we come to the hosting provider for the services. What you should pay attention to?
Specialization: look at the range of services offered, highlight the main direction and compare with their tasks. If the flagship services promoted by the company are exactly what you need - great. And vice versa, if you want to place your server with a company that mainly deals with VPS, wait for problems.
Reserve for growth: the needs of your project now and through may vary greatly. Is there a reserve in the service line to grow your project? If so, how is the migration from one service to another organized (automated, using a hoster, independently)?
DDoS counteraction: for any commercial project, the risk of DDoS attacks is relevant. What can the provider offer? Is there a ready-made solution within the DC or cloud? How exactly is the protection organized, what are the tariffs and terms of connection? In the absence of protection on the side of the DC, you need to find out the possibility of quickly changing the IP address and access to the machine via a spare communication channel.
The quality of support is difficult to assess in advance, but there are several signs.
Mode of operation support: should be really around the clock. This can be indirectly verified by the reaction time to requests at night and on weekends.
Service Level Agreement (SLA): this is a useful thing, but you shouldn’t rely on it completely. The fact is that compensation for a simple service is usually limited to the cost of the service for the period, that is, from a financial point of view does not cover the real losses of the business. On the other hand, the very existence of SLA disciplines the company and shows a systematic approach to the quality of services.
Accounting applications: it is better that the work with applications was implemented in the form of a web panel with duplication to e-mail. The history of the correspondence should be saved. Systems that work only on mail are not reliable enough, moreover, your contact mail can be located on the problem server.
The above criteria are based on our experience with hosting providers. I hope this experience will be useful. If you have your own work on the choice of hosting provider, share them in the comments.
In this article, we will look at a methodical approach to choosing a hosting provider from the standpoint of high-quality web project work, both in terms of speed and reliability. I’ll clarify that we’ll talk only about dedicated servers or VPS (or their cloud counterparts), completely virtual (shared) hosting is left out of the box as an unacceptable option.
1. Company, what should it be?
When choosing a company providing hosting services, you need to look at several criteria.
Specialization: it is desirable that the company is engaged in the hosting and / or management of data centers. The presence of specialization gives two advantages: higher competencies and result orientation. For example, it is often possible to meet the situation when a telecom company leases its data centers, and this service is not the key for business and receives the appropriate attention from management.
')
Direct provider: it is better to work directly with the company managing the data center. Only in this case the most prompt solution of problems is possible (fewer links in the incident response chain).
Experience: any IT services suggest a dynamic development of the market, so you should not be afraid of new players. However, it’s worthwhile to bypass the companies created less than two years ago: at best, you will encounter “childish” problems, at worst - the company will disappear from the market (examples are Oversan, Makhost and some others).
Scale: the optimal size of the company from small to medium scale. Micro-companies should be avoided, as they may not have the necessary personnel, reserve resources and most likely they will save on quality of services. Large companies will be more difficult to interact with: in the case of a need for an individual solution and generally less sensitive to customer problems.
Openness: an additional advantage will be the active participation of the company in communication with customers. This activity can take the form of blogging, surveys, participation in professional conferences and other events. As a rule, this is a sign of a desire to improve the quality of services and a willingness to engage in dialogue with customers.
Geography: it is important where the company's data centers are located - the data should be as close as possible to users. If distributed infrastructure is required from a single supplier, the company must have a network of data centers.
2. Data center
That data center is key to the quality of your project. No organization or technical support will be able to compensate for poor connectivity or power failures in the DC. To evaluate the criteria for selecting a data center, it is required that the company discloses information about its data center. In case there is no such data, it is worth asking: why is this information hidden?
Connectivity: one of the most important characteristics. It is determined by the number and quality of Internet channels in the data center, as well as the traffic exchange policy (peering). It is important that the data center is located as close as possible to your users (site visitors). Check the network latency to the DC can be a simple ping command, which must be performed from various networks in the region of the target audience. For a more in-depth analysis of connectivity, you can use the global connectivity graph, which is presented on the radar.qrator.net service (Radar by Qrator). This graph shows the routing policies between all autonomous systems (AS) and the Internet. It is worth looking at the graph itself (section Graph), providers, peering. You can also see the major clients of the hoster (those who have their own AS). It is necessary to be afraid of the companies at which the graph of connectivity rests against a single node - this means that the hoster is connected through another network and is completely dependent on its work.
Launch date: preference should be given to new data centers, as now more thoughtful projects are used, more power is allocated, a large bandwidth is laid. At the same time, you need to remember about the possibility of "children's" problems in data centers running "yesterday." For old data centers, it is also possible to replace / maintain infrastructure equipment, which often leads to interruptions in service.
Certification: in theory, the level of certification (Tier II, III, IV) should show the reliability of the data center. In general, the idea is correct, the presence of such certification (Tier II, better Tier III) can be considered as an advantage. However, you need to remember that as a rule, only the data center project is certified. It’s not at all the fact that the project will be fully implemented. In addition, the most important aspect - maintenance and operation - is usually not certified in Russia.
Power supply: the connected power should be enough for your needs (in the case of a rack rental). Redundancy should be organized either in the classical scheme: electrical UPS + diesel generator, or in the form of a dynamic UPS. It is desirable to power the DC from two independent rays. In fact, the reliability of the power supply can be checked only by personal experience (there have been cases of DC shutdown due to UPS).
Cooling and ventilation: there are various standards for temperature in the engine room of the DC. The current trend in order to increase energy efficiency is to increase the working temperature of the air, so it is worthwhile to clarify the needs of your equipment in terms of temperature. It is important that the air conditioning system has a reserve for maintenance and power reserve. In addition, it is important that the air in the engine room is filtered from dust, otherwise the equipment becomes clogged with dirt very quickly.
Additional fault tolerance: for critical applications, additional measures may be required to increase reliability - connecting the server via two power supplies, several independent network links, or organizing fault tolerance between the DC (for example, VRRP).
Availability and cost of channels: when assessing the cost of accommodation, you need to accurately determine the conditions of connection. First, the channel can be with a guaranteed or non-guaranteed band. Secondly, there may be a limit on the amount of traffic or incoming / outgoing ratios. It is also worthwhile to find out in advance the possibility and cost of channel expansion (the standard is 100 Mbit / s).
Processes and security: when deciding on the placement of equipment in the data center, it is useful to visit it in advance and look at the security system, as well as the organization of work with the equipment. The entrance to the computer room should be strictly regulated and limited. In the service area of ​​the equipment - order and cleanliness. If during your visit to the DC you found a mess and lack of organization, you should not expect high availability.
3. Services
Actually, we come to the hosting provider for the services. What you should pay attention to?
Specialization: look at the range of services offered, highlight the main direction and compare with their tasks. If the flagship services promoted by the company are exactly what you need - great. And vice versa, if you want to place your server with a company that mainly deals with VPS, wait for problems.
Reserve for growth: the needs of your project now and through may vary greatly. Is there a reserve in the service line to grow your project? If so, how is the migration from one service to another organized (automated, using a hoster, independently)?
DDoS counteraction: for any commercial project, the risk of DDoS attacks is relevant. What can the provider offer? Is there a ready-made solution within the DC or cloud? How exactly is the protection organized, what are the tariffs and terms of connection? In the absence of protection on the side of the DC, you need to find out the possibility of quickly changing the IP address and access to the machine via a spare communication channel.
4. Support and SLA
The quality of support is difficult to assess in advance, but there are several signs.
Mode of operation support: should be really around the clock. This can be indirectly verified by the reaction time to requests at night and on weekends.
Service Level Agreement (SLA): this is a useful thing, but you shouldn’t rely on it completely. The fact is that compensation for a simple service is usually limited to the cost of the service for the period, that is, from a financial point of view does not cover the real losses of the business. On the other hand, the very existence of SLA disciplines the company and shows a systematic approach to the quality of services.
Accounting applications: it is better that the work with applications was implemented in the form of a web panel with duplication to e-mail. The history of the correspondence should be saved. Systems that work only on mail are not reliable enough, moreover, your contact mail can be located on the problem server.
Results
The above criteria are based on our experience with hosting providers. I hope this experience will be useful. If you have your own work on the choice of hosting provider, share them in the comments.
Source: https://habr.com/ru/post/317012/
All Articles