Actual innovation: what to expect from the data center market in 2019?

Construction of data centers is considered one of the fastest growing industries. Progress in this area is enormous, but whether some breakthrough technological solutions will appear on the market in the near future is a big question. Today we will try to consider the main innovative trends in the development of the world TSODostroeniya to answer it.

Course on Hyperscale

The development of information technology has led to the need to build very large data centers. Mostly hyper-scalable infrastructure is needed by cloud service providers and social networks: Amazon, Microsoft, IBM, Google and other major players. In April 2017, there were 320 such data centers in the world, and in December there were already 390. By 2020, the number of hyper-scalable data centers should increase to 500, according to forecasts by Synergy Research specialists. Most of these data centers are located in the United States, and this trend continues, despite the fast pace of construction in the Asia-Pacific region, as noted by analysts at Cisco Systems.

All hyper-scalable data centers belong to corporate and space in the racks do not pass. They are used to create public clouds related to the Internet of things and technology of artificial intelligence services, as well as in other niches that require processing of huge amounts of data. Owners are actively experimenting with an increase in power density per rack, unpackaged servers, liquid cooling, temperature increase in machine rooms and a variety of specialized solutions. Given the increasing popularity of cloud services, Hyperscale in the foreseeable future will be the main driver of growth in the industry: here we can expect the emergence of interesting technological solutions from leading manufacturers of IT equipment and engineering systems.

Boundary computing

Another noticeable trend is just the opposite: in recent years, a huge number of micro data centers have been built. According to the forecasts of Research and Markets, this market will grow from $ 2 billion in 2017 to $ 8 billion by 2022. This is attributed to the development of the Internet of things and the industrial Internet of things. Large data centers are located too far from the automation systems of production processes in the field. They are engaged in tasks for which the readings of each of the millions of sensors are not required. Primary data processing is best done where they are generated, and only then send useful information along long routes to the cloud. To designate this phenomenon, a special term was invented - boundary computing or edge computing. In our opinion, this is the second most important trend in the development of TSODostroeniya, which leads to the emergence of innovative products on the market.

Battle for PUE

Large data centers consume a huge amount of electricity and generate heat, which must somehow be disposed of. The share of traditional cooling systems accounts for up to 40% of the energy consumption of the object, and in the struggle to reduce energy consumption, the main enemy is the compressors of refrigerating machines. Gaining popularity to completely or partially abandon their use of the solution with the so-called. free cooling In the classical scheme, chiller systems with water or aqueous solutions of polyatomic alcohols (glycols) are used as a coolant. In the cold season, the condenser unit of the chiller does not turn on, which significantly reduces energy consumption. More interesting solutions are based on a two-circuit air-to-air scheme with rotary heat exchangers and with or without adiabatic cooling section. Experiments are underway with direct cooling with outside air, but these solutions can hardly be called innovative. Like classical systems, they involve air cooling of IT equipment, and the technological limit of efficiency of such a scheme is almost reached.

Further reductions in PUE (the ratio of total energy consumption to energy consumption of IT equipment) will come at the expense of the increasingly popular liquid-cooling schemes. It is worth remembering the project launched by Microsoft for the creation of modular submarine data centers, as well as the concept of floating data centers Google. The ideas of technological giants are still far from industrial implementation, but less fantastic liquid cooling systems are already working on various objects from supercomputers from Top500 to micro data centers.

In case of contact cooling, special heat removers are installed in the equipment, inside which liquid circulates. Immersion cooling systems use a dielectric working fluid (usually mineral oil) and can be made either as a general sealed container or as individual enclosures for computing modules. Boiling (two-phase) systems at first glance look like submersible ones. They also use dielectric liquids in contact with electronics, but there is a fundamental difference - the working fluid begins to boil at temperatures of about 34 ° C (or slightly higher). We know from the physics course that the process goes with energy absorption, the temperature stops rising and with further heating the liquid evaporates, i.e. a phase transition occurs. At the top of the sealed container, the vapors come into contact with the radiator and condense, and the droplets return to the common reservoir. Liquid cooling systems make it possible to achieve fantastic PUE values ​​(around 1.03), but they require serious modifications of computing equipment and cooperation of manufacturers. Today they are considered the most innovative and promising.

Results

To create modern data centers came up with a lot of interesting technological approaches. Manufacturers offer integrated hyperconvergent solutions, software-defined networks are built, and even the data center itself becomes software-defined. To improve the efficiency of facilities, not only innovative cooling systems are installed, but also DCIM-class hardware and software solutions, which allow optimizing the performance of engineering infrastructure based on data from a variety of sensors. Some innovations do not justify the hopes placed on them. Modular container solutions, for example, could not replace traditional data centers from concrete or pre-fabricated steel structures, although they are actively used where computing power needs to be deployed promptly. At the same time, traditional data centers themselves become modular, but on a completely different level. Progress in the industry is very fast, although without technological leaps - the innovations we mentioned first appeared on the market several years ago. 2019 in this sense will not be an exception and will not bring obvious breakthroughs. In the era of numbers, even the most fantastic invention is quickly becoming an ordinary technical solution.