The IT industry market is the largest consumer of uninterruptible power supplies (UPS), using approximately 75% of all UPS produced. Annual global sales of UPS equipment in all types of data centers, including corporate, commercial and ultra-large, is $ 3 billion. At the same time, the annual increase in sales of UPS equipment in data centers is approaching 10% and it seems that this is not the limit.
Data centers are getting bigger and bigger and this in turn creates new challenges for the power infrastructure. While there is a long debate about where static UPSs are superior to dynamic ones and vice versa, there is one thing that most engineers will agree on - that the higher the power, the more suitable electrical machines are for working with it: it is generators. used to generate electricity in power plants.
All dynamic UPSs use motor generators, however they are of different designs and certainly differ in features and performance. One such fairly common UPS is the mechanically connected diesel engine solution, the diesel rotary UPS (DRIBP). However, in the world practice of building data centers, the real competition is between static UPS and another dynamic UPS technology - rotary UPS, which is a combination of an electric machine that generates a sinusoidal voltage of natural form and power electronics. Such rotary UPSs are electrically connected to energy storage devices, which can be either batteries or flywheels.
Modern advances in control technology, reliability, efficiency and power density, as well as lower unit cost of UPS power, are factors that are not unique to static UPSs. The recently introduced Piller UB-V series is a worthy alternative.
Let's take a look at some of the key criteria for evaluating and selecting a UPS system for a modern large data center in the context of which technology seems to be the preferred one.
1. Capital costs
It is true that static UPSs can offer a lower price per kW for smaller UPS systems, but that advantage quickly evaporates when it comes to high power systems. The modular concept that static UPS manufacturers inevitably have to adopt revolves around paralleling a large number of low power UPSs, such as the 1kW size shown in the example below. This approach achieves the required system output power, but due to the complexity of the many redundant elements, it loses a 250-20% cost advantage compared to the cost of a solution based on rotary UPSs. Moreover, even this parallel connection of modules has limits on the number of units in one UPS system, after which the parallel modular systems themselves must be parallel, which further increases the cost of the solution due to additional switchgears and cables.
Tab. 1. Solution example for 48MW IT load. The larger size of the UB-V monoblocs saves time and money.
2. Reliability
In recent years, data centers have become more and more commoditized enterprises, while reliability is increasingly taken for granted. In this regard, fears are growing that this will lead to problems in the future. Because operators aim for the highest fault tolerance rating (number of 9) and it is assumed that the disadvantages of static UPS technology are best overcome by the short time to repair (MTTR) through the ability to quickly and hot-swap UPS modules. But this argument can be self-destructive. The more modules involved, the higher the probability of failure and, more importantly, the higher the risk that such a failure will lead to a loss of load in the overall system. It's better to have no crashes at all.
An illustration of the dependence of the number of equipment failures on the value of the time between failures (MTBF) during normal operation is shown in fig. 1 and the corresponding calculations.
Rice. 1. Dependence of the number of equipment failures on the MTBF index.
The probability of equipment failure Q(t) during normal operation, in section (II) of the graph of the normal failure curve, is quite well described by the exponential law of distribution of random variables Q(t) = e-(Ξ»x t), where Ξ» = 1/MTBF is the intensity failures, and t is the operating time in hours. Accordingly, after a time t in a trouble-free state there will be N(t) installations from the initial number of all installations N(0): N(t) = Q(t)*N(0).
The average MTBF of static UPSs is 200.000 hours, while the MTBF of rotary UPSs of the UB-V Piller series is 1.300.000 hours. The calculation shows that over 10 years of operation, 36% of static UPSs will be in an accident, and only 7% of rotary UPSs. Taking into account the different number of UPS equipment (Table 1), this means 86 failures out of 240 static UPS modules and 2 failures out of 20 Piller rotary UPSs, in the same data center with a 48MW IT payload over 10 years of operation.
The experience of operating static UPSs at data centers in Russia and in the world confirms the reliability of the above calculations, based on the statistics of failures and repairs available from open sources.
All Piller rotary UPSs, and in particular the UB-V series, use an electric machine to generate a pure sine wave and do not use power capacitors and IGBT transistors, which are very often the causes of failures in all static UPSs. Moreover, a static UPS is a complex part of the power supply system. Complexity reduces reliability. UB-V rotary UPSs have fewer components and a more robust system design (motor-generator) for improved reliability.
3. Energy efficiency
Modern static UPSs have much better online (or "normal" mode) energy efficiency than their predecessors. As a rule, with peak efficiency values ββββat the level of 96,3%. Higher figures are often given, but this is only achievable when the static UPS is operating, switching between online and alternative modes (e.g. ECO-mode). However, when using the alternative energy-saving mode, the load is operated from the external network without any protection. For this reason, in practice in data centers in most cases only online mode is used.
The Piller UB-V series of rotary UPSs do not change state during normal operation while still providing up to 98% efficiency online at 100% load and 97% at 50% load.
This difference in energy efficiency allows you to get significant savings on electricity during operation (Table 2).
Tab. 2. Energy cost savings in data center 48 MW IT load.
4. Occupied space
General purpose static UPSs have become significantly more compact with the transition to IGBT technology and the elimination of transformers. However, even taking into account this circumstance, UB-V series rotary UPSs give a gain of 20% or more in terms of space occupied per unit of power. The resulting space savings can be used both to increase the capacity of the energy center, and to increase the "white", useful, space of the building to accommodate additional servers.
Rice. 2. Space occupied by 2MW UPS of different technologies. Real installations to scale.
5. Accessibility
One of the key indicators of a well-designed, built and operated data center is its high fault tolerance factor. While 100% uptime is always the goal, reports indicate that more than 30% of the world's data centers experience at least one unplanned outage per year. Many of these are caused by human error, but energy infrastructure also plays an important role. The UB-V series utilizes years of proven Piller monobloc rotary UPS technology that is far superior in reliability to any other technology. What's more, UB-V UPSs themselves in data centers with a properly controlled environment do not need to be shut down annually for maintenance.
6. Flexibility
Often, data center IT systems are updated and upgraded within 3-5 years. Therefore, power and cooling infrastructures must be flexible enough to meet this and have sufficient future prospects. Both conventional static UPS and UB-V UPS can be configured in various ways.
However, the composition of solutions based on the latter is wider, and, speaking in general, since it is beyond the scope of this article, it makes it possible to implement uninterruptible power supply systems at medium voltage 6-30 kV, to work on networks with renewable and alternative generation sources, to build cost-effective, ultra-reliable, Tier IV UI compliant IP Bus systems in an N+1 configuration.
As a conclusion, several conclusions can be drawn. The more data centers develop, the more difficult the task of their optimization becomes when it is necessary to control economic indicators, aspects of reliability, reputation and minimization of environmental impact at the same time. Static UPSs have been and will continue to be used in data centers in the future. However, it is also undeniable that there are alternatives to existing approaches in the field of power supply systems that have significant advantages over the "good old statics".
Source: habr.com