Data Center Experience and Problems: How to Verify Data Center Reliability
Nowadays, designing and creating data centers is a fairly well-studied area. Were developed and set standards that are followed when building data centers.
They contain the mandatory and recommended requirements for the infrastructure of the machine rooms, cable systems and network design. However, despite all the experience, this occupation is still incredibly difficult.
/ Photo by Robert CC
')
The first question to ask before starting the construction of a data center is “where to build it?”. Very often, old abandoned warehouses or supermarkets turn into data processing centers, since all the necessary communications have already been connected to them. If there are no suitable facilities nearby, then you need to build closer to power plants, since you will need a lot of electricity. This is confirmed: the new data center, which in 2012 was built in the county of Crook, USA, the social network Facebook, consumes 28 megawatts . The funny thing is that the same amount consumes the entire county Crook.
Therefore, the builders and owners of many modern data centers approach this issue responsibly. For example, Apple is constantly seeking to reduce energy costs. According to the company's reports , since 2013, 100% of their data centers and 94% of offices operate on renewable energy sources. The data center in Maiden is surrounded by solar panels, which occupy about 400,000 square meters. square meters and produce 42 million kilowatt-hours of electricity per year. In the Digital Beijing data center in Beijing, LED lamps are used exclusively, consuming 60% less electricity, and a special glass wall prevents the penetration of heat into the interior, reducing cooling costs.
Speaking of cooling. It is not a secret that data centers emit a huge amount of heat, therefore, to avoid overheating of vital systems, it has to be removed. It is considered that the cooling of equipment consumes a large part of the entire power consumption. Water and sophisticated intelligent air conditioning systems are used to remove heat from each server and router. Sometimes, however, there are situations opposite: for example, when two “independent” power supplies come from a single source.
The collected thermal energy is often simply dissipated in the air, but the situation can be radically changed. “It is necessary to connect the data center to the central heating system,” said Jari Innanen, chairman of the coordination center for data center workers in Finland. According to Jari, the water used to cool Finnish data centers can provide heating for 300,000 private houses in the country. This approach is already being used in Finland: the data center of the Russian Yandex is heating the city of Mäntsälä.
When the data center is built and functioning, the question arises, how many employees do you need to support its work? If we are talking about the number of people "on the spot", then quite a bit. According to Facebook, their data center, built in Prineville in 2010, employs 35 people. If we take into account the size of the entire building, then this is only 1 worker per 790 sq. meters Their responsibilities include the repair of equipment or its replacement, maintenance of diesel generators and assessment of the condition of the entire facility as a whole.
However, these, so to speak, "local" technicians are supported by hundreds and even thousands of system administrators, engineers and programmers working remotely. They do not need direct access to the equipment (with the exception of rare critical situations), and they can manage virtually all processes on the network using software.
And what about security? After all, the servers can be stored not just your photos of cats in Instagram. Here, a variety of information security management systems, data backup and recovery systems, and information transmission channel protection systems are used - everything is standard. Much more interesting is the situation with protection from external influences and penetration. According to Brandon Berryhill, a data center engineer, some of the data center buildings are able to withstand wind gusts of up to 321 km / h and earthquakes of up to 9 points.
Armored cabin with a biometric data reader in the data center where the IT-GRAD cloud is located
Of course, some objects are much better protected than others. There are data centers in which between the attacker and the servers there is only a simple employee card scanner, but there are places with thick walls, bullet-proof doors and barbed wire, where biometric scanners are installed, and armed guards stand at the entrance. “One of my colleagues said that there is a data center in Europe, around which a moat was dug. It’s a pity without crocodiles, ”says Brandon.
Unfortunately, in Russia (and abroad) not all data centers work in such conditions. Not all commercial data centers are certified for reliability, not to mention an independent external audit confirming certification. Sometimes the company that owns the data center is limited to saying that they conducted an “internal audit” and are based on an independent assessment of the level of reliability of the data center.
First of all, this is due to the fact that the certification process of the data center costs a lot of money. Moreover, even the drawings of the future data center are subject to certification, and after it has been built, Uptime Institute experts (independent auditor of the data center compliance with fault tolerance requirements) check the result, assess the compliance of the data center systems with the declared values ​​and place information about the data center and the Tier certificate assigned to it in the UTI catalog .
Very often, such verification is not performed: in addition to organizing the certification process, the receiving party (the owner of the data center) needs to ensure the arrival and residence of auditors, as well as lengthy and thorough inspections, so sometimes companies are limited to assessing reliability based on their own ideas. In this case, the assessment is sometimes carried out for each subsystem separately, and the highest value is chosen as the final one.
Suspicious levels of reliability with “pluses” like Tier 3+ can also be suspicious. In fact, this means that one of the elements of the system is more reliable than the others, which does not increase the reliability of the entire system, but suggests that the data center meets higher reliability requirements in comparison with the standard. Such “pluses” are often a marketing ploy (we wrote about it in more detail here ).
First of all, you should trust your data to certified data centers. In the absence of such, the client, of course, may try to independently assess the state of the data center, but this usually presents a number of difficulties: detailed information about the data center will be needed, including data on engineering communications and so on. On the one hand, the analysis and evaluation of this information will take a lot of time (and the time of a qualified specialist), on the other hand, it is likely that the management, who saved on the construction of the data center, will not be ready to provide such data to the client.
In addition to certification issues, it is worth examining the legal status of the data center (does the provider have all the necessary permits from state institutions), and also find out whether the provider company has entered into service and support systems contracts in the event of an emergency — for example, in the event of a weather anomaly ( this is often forgotten, reducing all questions to the reliability of the system itself).
Pay attention to the possibility of visiting the data center: if you are denied a tour, you may have a reasonable cause for concern. And vice versa: if the company is ready to show you the machine room and engineering premises, you have more reasons to trust the provider. For example, the provider company, whose data center uses IT-GRAD, not only conducted a tour of its data center, but also provided detailed information for our material , in which we talk about where our cloud is located, how communication is provided, cooling, security of the object and other nuances associated with our chosen data center.
They contain the mandatory and recommended requirements for the infrastructure of the machine rooms, cable systems and network design. However, despite all the experience, this occupation is still incredibly difficult.
/ Photo by Robert CC')
Facts from the life of the data center
The first question to ask before starting the construction of a data center is “where to build it?”. Very often, old abandoned warehouses or supermarkets turn into data processing centers, since all the necessary communications have already been connected to them. If there are no suitable facilities nearby, then you need to build closer to power plants, since you will need a lot of electricity. This is confirmed: the new data center, which in 2012 was built in the county of Crook, USA, the social network Facebook, consumes 28 megawatts . The funny thing is that the same amount consumes the entire county Crook.
Therefore, the builders and owners of many modern data centers approach this issue responsibly. For example, Apple is constantly seeking to reduce energy costs. According to the company's reports , since 2013, 100% of their data centers and 94% of offices operate on renewable energy sources. The data center in Maiden is surrounded by solar panels, which occupy about 400,000 square meters. square meters and produce 42 million kilowatt-hours of electricity per year. In the Digital Beijing data center in Beijing, LED lamps are used exclusively, consuming 60% less electricity, and a special glass wall prevents the penetration of heat into the interior, reducing cooling costs.
Speaking of cooling. It is not a secret that data centers emit a huge amount of heat, therefore, to avoid overheating of vital systems, it has to be removed. It is considered that the cooling of equipment consumes a large part of the entire power consumption. Water and sophisticated intelligent air conditioning systems are used to remove heat from each server and router. Sometimes, however, there are situations opposite: for example, when two “independent” power supplies come from a single source.
The collected thermal energy is often simply dissipated in the air, but the situation can be radically changed. “It is necessary to connect the data center to the central heating system,” said Jari Innanen, chairman of the coordination center for data center workers in Finland. According to Jari, the water used to cool Finnish data centers can provide heating for 300,000 private houses in the country. This approach is already being used in Finland: the data center of the Russian Yandex is heating the city of Mäntsälä.
When the data center is built and functioning, the question arises, how many employees do you need to support its work? If we are talking about the number of people "on the spot", then quite a bit. According to Facebook, their data center, built in Prineville in 2010, employs 35 people. If we take into account the size of the entire building, then this is only 1 worker per 790 sq. meters Their responsibilities include the repair of equipment or its replacement, maintenance of diesel generators and assessment of the condition of the entire facility as a whole.
However, these, so to speak, "local" technicians are supported by hundreds and even thousands of system administrators, engineers and programmers working remotely. They do not need direct access to the equipment (with the exception of rare critical situations), and they can manage virtually all processes on the network using software.
And what about security? After all, the servers can be stored not just your photos of cats in Instagram. Here, a variety of information security management systems, data backup and recovery systems, and information transmission channel protection systems are used - everything is standard. Much more interesting is the situation with protection from external influences and penetration. According to Brandon Berryhill, a data center engineer, some of the data center buildings are able to withstand wind gusts of up to 321 km / h and earthquakes of up to 9 points.
Armored cabin with a biometric data reader in the data center where the IT-GRAD cloud is located
Of course, some objects are much better protected than others. There are data centers in which between the attacker and the servers there is only a simple employee card scanner, but there are places with thick walls, bullet-proof doors and barbed wire, where biometric scanners are installed, and armed guards stand at the entrance. “One of my colleagues said that there is a data center in Europe, around which a moat was dug. It’s a pity without crocodiles, ”says Brandon.
Modern realities: problems of data centers
Unfortunately, in Russia (and abroad) not all data centers work in such conditions. Not all commercial data centers are certified for reliability, not to mention an independent external audit confirming certification. Sometimes the company that owns the data center is limited to saying that they conducted an “internal audit” and are based on an independent assessment of the level of reliability of the data center.
First of all, this is due to the fact that the certification process of the data center costs a lot of money. Moreover, even the drawings of the future data center are subject to certification, and after it has been built, Uptime Institute experts (independent auditor of the data center compliance with fault tolerance requirements) check the result, assess the compliance of the data center systems with the declared values ​​and place information about the data center and the Tier certificate assigned to it in the UTI catalog .
Very often, such verification is not performed: in addition to organizing the certification process, the receiving party (the owner of the data center) needs to ensure the arrival and residence of auditors, as well as lengthy and thorough inspections, so sometimes companies are limited to assessing reliability based on their own ideas. In this case, the assessment is sometimes carried out for each subsystem separately, and the highest value is chosen as the final one.
Suspicious levels of reliability with “pluses” like Tier 3+ can also be suspicious. In fact, this means that one of the elements of the system is more reliable than the others, which does not increase the reliability of the entire system, but suggests that the data center meets higher reliability requirements in comparison with the standard. Such “pluses” are often a marketing ploy (we wrote about it in more detail here ).
How to avoid risks
First of all, you should trust your data to certified data centers. In the absence of such, the client, of course, may try to independently assess the state of the data center, but this usually presents a number of difficulties: detailed information about the data center will be needed, including data on engineering communications and so on. On the one hand, the analysis and evaluation of this information will take a lot of time (and the time of a qualified specialist), on the other hand, it is likely that the management, who saved on the construction of the data center, will not be ready to provide such data to the client.
In addition to certification issues, it is worth examining the legal status of the data center (does the provider have all the necessary permits from state institutions), and also find out whether the provider company has entered into service and support systems contracts in the event of an emergency — for example, in the event of a weather anomaly ( this is often forgotten, reducing all questions to the reliability of the system itself).
Pay attention to the possibility of visiting the data center: if you are denied a tour, you may have a reasonable cause for concern. And vice versa: if the company is ready to show you the machine room and engineering premises, you have more reasons to trust the provider. For example, the provider company, whose data center uses IT-GRAD, not only conducted a tour of its data center, but also provided detailed information for our material , in which we talk about where our cloud is located, how communication is provided, cooling, security of the object and other nuances associated with our chosen data center.
Source: https://habr.com/ru/post/276103/
All Articles