Recently, the Internet "turned" 30 years old. During this time, the information and digital needs of the business have grown to such an extent that today we are no longer talking about a corporate server room or even the need to be located in a data center, but about renting a whole network of data centers with an accompanying set of services. Moreover, we are talking not only about global projects with big data (the giants have their own data centers), but even about medium-sized companies with frequent updates of database positions (for example, online stores) and services with high-speed data exchange (for example, banks).
Why does a business need a system of distributed data centers
Such a system consists of IT complexes, geographically distributed according to the principle: main data center and regional data centers. They are initially equipped taking into account possible information flows and business processes of modern developing companies and ensure the uninterruption of these flows and processes.
▍Why distributed?
First, because of the risk of breaking all the eggs in one basket. Nowadays, there is a demand for fault-tolerant solutions that can ensure absolutely continuous operation of corporate applications, services and websites in any conditions. Even at the end of the world. Such computing infrastructures should not only store data efficiently, but also minimize downtime for the company’s (read: business) IT services, both during an epidemic of blocking by Roskomnadzor, and during natural disasters, and during a real man-made disaster, and in any other force majeure circumstances . It’s not for nothing that these solutions are called disaster recovery.
To do this, the sites of computing systems that work for the company must be removed from each other at a safe distance according to a certain scheme (see the table and illustration below). If necessary, a disaster recovery plan (DR-Plan) is used and the automatic transfer of customer services to another network site using fault-tolerant methods and software solutions that are optimal for each specific case (data replication, backup, etc.).
Secondly, to improve performance. In normal mode (not force majeure, but with peak loads), distributed data centers are designed to increase company productivity and minimize information losses (for example, during DDoS attacks). Here, load balancing complexes between computing nodes are activated: the load is redistributed evenly, and if one of the nodes fails, its functions will be taken over by other nodes of the complex.
Thirdly, for the efficient operation of remote branches. For companies with many divisions, solutions for centralized storage and processing of information with geographically distributed replication are used. At the same time, each branch can work with its own amount of data, which will be consolidated into a single database of the central office. In turn, changes in the central database are reflected in the bases of subdivisions.
▍Structure of distributed data centers
Geographically distributed data centers are divided into four types. For an external user, they look like a single system: management occurs through one services and support interface.
Geographically distributed data centers
▍Goals for which businesses need distributed data centers:
Continuity of data processing. Continuity is required to solve inevitably arising technical problems without stopping business processes, even if some communication channels and a significant part of the system fail. By the way, the ability of the system to perform its functions within the planned time, taking into account the average time indicator of safe operation and the time frame for restoring functionality () determines the level of reliability of the data center. There are four levels in total: TIER1, TIER2, TIER3, TIER4; the higher the indicator, the more reliable the equipment of the center and the higher the standard of its entire infrastructure.
Increased productivity and capacity. If necessary (peak loads), the ability to increase capacity and increase the efficiency of backup data centers due to economies of scale: maximum use of computing resources of the entire distributed system. Scalability provides flexible, on-demand computing capabilities through dynamic configuration.
Disaster resistance. It is provided by redundant computing power at a remote site. System performance is achieved by a certain setting of the RPO recovery point and RTO recovery time (the degree of security and recovery speed depend on the tariff).
distributed services. The company's IT resources and services are decoupled from the underlying infrastructure and delivered in a multi-tenancy environment on demand and at scale.
Geographic localization of services. To expand the target audience of the brand and enter the company into new geographic markets.
Cost optimization. Creating and maintaining your own data center is very project. For most companies, especially large geographically distributed ones and those planning new points of presence in the market, outsourcing the IT infrastructure will help to save significantly.
Why is it profitable for a business to have a data center "at hand"
For many modern services and business applications, the speed of access to the site is critical. This speed depends, first of all, on the distance between the sites of the distributed data center system. If it is small, then communications are simplified and productivity is increased due to the fact that signal delay (latency) is reduced. This is especially important when making reservations. In a fiber optic cable, the propagation delay of light is approximately 5 ms/km. Latency affects the execution time of an I/O operation, which is approximately 5-10 ms.
Since services must operate constantly, while they must have a high degree of availability and minimal downtime, it is beneficial for a business to rent IT infrastructure geographically close to users of target markets.
The speed of access to the site also depends on the equipment. For example, in our new data center in the IT Park of Kazan, you can get a 100 Mbit/s Internet channel for your virtual server with the most comfortable access.
For a business with a large international reach, it is good to use foreign sites to host data in order to save traffic costs and reduce the response time of website pages for foreign users. Long response time is the reason and, more importantly, the reason why your target audience is fleeing your sites (high bounce rate leading to loss of leads).
What are the advantages of backup data centers?
Given the often unstable situation in Russia in the field of information security (for example, the same recent massive blocking of IP addresses by Roskomnadzor, which affected even sites that are not related to Telegram), it is convenient to locate part of the IT infrastructure of a business outside the Russian legal field. For example, by renting servers in a Swiss data center, you are subject to Swiss data protection laws, which are quite strict. Namely: neither the state structures of Switzerland itself (with the exception of the government in special cases), nor the law enforcement agencies of other countries have access to any information on the "Swiss" servers. Without the knowledge of the client, data cannot be requested from data centers and providers.
Deployment of a backup data center (or hosting) on a remote () site is strategically justified if there is a need for painless migration of business-critical services for their uninterrupted operation.
A little more about the Kazan data center
Since we are already talking about a data center in Kazan, let's allow ourselves a small advertising block. "IT-Park", which houses the data center - the largest technology park in the field of high technologies in Tatarstan. It is a 3 MW TIER2,5 square kilometer data center with the capacity to accommodate more than 300 racks.
Security at the physical level is provided by two armed security circuits, video cameras around the perimeter, a passport access system at the entrance, an ACS biometric system (fingerprints) in the engine room, and even a dress code for visitors (robes, special shoe covers with a machine for putting on).
All technical rooms and server rooms are equipped with a gas-fired automatic fire extinguishing system with smoke detectors, which makes it possible to eliminate the source of ignition without damaging high-tech equipment. Energy-saving, cooling, ventilation systems are implemented at the highest level, and the key elements of these systems are located in separate rooms.
We put into operation our own containment area in the IT Park data center. The data center has an SLA of 99.982%, which means it fully complies with the high international requirements for the operational stability of data centers. It has FSTEC and FSB licenses, a PCI-DSS certificate, which allows you to place equipment of organizations working with personal data (banks and others). And, as always, the prices for virtual servers from the RUVDS hosting provider in this data center do not differ from the prices for in our other data centers in Moscow, St. Petersburg, London, Zurich.
Source: habr.com