Along with the shift we’re seeing in the marketplace from traditional datacenter builds to those including more aspects of prefabricated modular (PFM) datacenters, there is a general openness to rethinking conventional technologies and build methods. Broadly speaking, we expect to see a ‘phasing out’ of traditional uninterruptible power supply (UPS) topologies/technologies as the shift toward shorter ride-through times of UPSs continues. Operators are increasingly accepting or requiring shorter ride-through times, and we predict this will catalyze a change in the types and configurations of UPSs that are deployed. Some technologies that were considered inferior or niche in the past may be preferred in the future. The gradual but inevitable proliferation of PFM datacenters will help spur this change.
A few years ago, the new thing with UPSs was eco-mode, which essentially operated a double-conversion UPS in bypass or as a line-interactive UPS (these technologies are explained in more detail later in this report). Additionally, initiatives like Facebook’s Open Compute project, which eliminates the centralized UPS and replaces it with distributed batteries or rack-based UPSs, are likely to become more prevalent. This will happen naturally as datacenter fault tolerance shifts to software and algorithmic controls rather than through straight (and expensive) physical-equipment redundancy.
Not everyone is as progressive as the big datacenter guys like Facebook, Google and Microsoft, however. The most likely first step for the mainstream market as far as UPS technologies go will be a gradual reduction of required backup time. To an extent, this has already been happening for a while. While 30-minute runtime battery strings were commonplace even just a few years ago, runtimes of 15 minutes or less are now common. This has been mostly as a result of striking that fine balance between cost and reliability, and those building datacenters having to decide if they really need 30 minutes of backup time, or if something shorter would be acceptable. In addition to the growing acceptance of shorter runtimes, the maturation of PFM datacenters is ushering in a period of less concern about engineering nuances of the datacenter itself and greater focus on more broad performance characteristics – regardless of how they are actually achieved. Broadly speaking, this means that convention is being disregarded, and loyalty to a particular piece of gear or technology method is softening. Consequently, UPS technologies and build methods are already starting to be disrupted.
As the market continues to mature and technology shifts occur, it is important to outline the different types of UPSs that may be seen in this transition. This discussion is intended to be from a high-level, non-engineering point of view. There are more intricacies to the UPS discussion than this report will cover, but it aims to give an elementary summary of the different types of UPSs.
A static UPS is one that makes use of solid-state circuitry (rather than a rotating mass) within the UPS itself to provide ride-through time when utility is lost. Static UPSs are the most common type of UPS in datacenters. They provide load isolation, are relatively straightforward to maintain, and can give a range of ride-through times at the loss of utility power with different battery string configurations/sizes. Battery strings are generally expensive and require climate control and monitoring/exhaust (e.g., for hydrogen detection and exhaust for wet cell batteries). For the purposes of this report, there are two main types of static UPSs: double-conversion and line-interactive.
• Double-conversion UPSs do just as their name indicates: they take incoming alternating current (AC) , convert it to direct current (DC) and then convert this DC back to AC power on the output. This conversion offers complete isolation, generating the signal independent of incoming power. Double-conversion UPSs are able to perform frequency conversion (e.g., 50Hz in, 60Hz out) as well as compensate for what are referred to as ‘non linear’ loads, which can damage sensitive electronic equipment. The battery-based double-conversion UPS is the conventional UPS model.
• Line-interactive UPSs, during normal operation, in essence act as isolation transformers and don’t provide constant AC/DC/AC conversion. Line-interactive UPSs are more passive in nature and ‘activate’ only when something goes wrong (e.g., out-of-tolerance input, power surges, power dips, etc.).
A less common type of UPS is the rotary UPS, which uses a motor/generator system to create the output. Rotary UPS units (e.g., those from Hitec Power Protection, Piller Power Systems and E1 Dynamics) can be configured in a number of different ways involving synchronous or asynchronous motors, rectifiers or inverters. While the different configurations can be confusing, the main point of differentiation is that there is a motor/generator and rotating mass connected to an engine. These differ from flywheel kinetic storage UPSs, often called flywheel UPSs (see below).
Rotary UPSs are capable of providing high fault-clearing capabilities (peak/maximum current to blow a fuse ) without going to bypass. This means that the unit is able to provide ‘short circuit current’ to blow a fuse or trip a protection switch downstream instead of ‘protecting itself’ as static UPSs do. Rotary UPSs also provide superior current inrush handling for inductive loads such as motor startup or compressor loads. In terms of maintainability, the total life of rotary UPSs is generally longer than static UPSs, but they do require periodic downtime for mechanical maintenance, while static UPSs may not if batteries can be hot-swapped.
Rotary UPSs are the second-most-common UPS found in the market and have a relatively loyal following (proponents of rotary UPSs generally scoff at static UPSs as inferior and less robust – and vice versa). Rotary UPSs are more common in Europe than the US, however, and have been adopted by several large, well-known datacenter operators.
Oftentimes the terms ‘flywheel UPS’ and ‘rotary UPS’ are used interchangeably, though they are quite different. While the rotary UPS does make use of a spinning mass, or flywheel, the overall technology design (described above) is different from the ‘flywheel UPS.’ A flywheel UPS (e.g., from Active Power or Eaton/Powerware) is really a static UPS that replaces (or supplements) a DC battery string with a DC-powered kinetic flywheel energy-storage system. As far as the UPS itself is concerned, it does not care if the DC bus is supplied from batteries or flywheels.
Flywheel UPSs generally have the shortest ride-through times of all the UPSs (15-30 seconds as compared to minutes with battery-based backup systems). This has been the major sticking point with flywheel UPSs in the past. The main argument in favor of flywheel UPSs is that if your generator system has not started within the time frame that the flywheel can carry the critical load, it is unlikely that 15 or even 30 minutes more will give operators a chance to correct the issue to prevent a total outage. While this point is debatable, there is a general trend in the direction of reduced backup/ride-through times of UPSs, so flywheels are starting to gain some traction.
The UPS vendor market is broad-reaching with many players. There are incumbents in each type, and some have a greater geographical reach than others. For static UPSs, the main players are Eaton, Emerson Network Power/Liebert, GE, Minuteman from Para Systems, Mitsubishi International, Schneider Electric/APC/MGE, Socomec Group, Toshiba, Riello, KST, Energetix/Pnu Power and Tripp Lite. The rotary UPS market has fewer suppliers than the static market and includes E1 Dynamics (manufactured by EURO-DIESEL), Hitec Power Solutions and Piller Power Systems. As far as flywheel-based UPSs go, the only vendor that focuses on flywheel UPSs exclusively is Active Power, though many of the static UPS manufacturers offer flywheel-based ‘battery’ systems through partnerships with kinetic energy storage equipment vendors like Vycon (which does not offer a UPS, just kinetic energy-storage equipment).
UPS technology is well-entrenched in the datacenter industry with many loyalty camps. Debates over static vs. rotary and flywheels vs. batteries have lasted years (and are a common discussion topic at industry conferences even still). The advent of cloud computing has helped lessen the weight placed on datacenter and UPS design/technology methods, however, and traditional datacenter providers and their philosophies have increasingly considered cloud service providers as competition.
This competition will continue, and the PFM market will likely be the impetus to a disruption in datacenter resiliency/uptime design methods. Facebook’s Open Compute project, while it won’t be the silver bullet for the datacenter industry, is a good indicator of the trends to expect for the mainstream (i.e., not the cutting-edge companies like Facebook, Google, et al.) as technology catches up and mainstream users can adopt methods/technologies that the leaders have blazed the trail for.
UPSs will not likely be phased out overnight (if they’re ever really eliminated at all), but as acceptable ride-through times decline, different battery/energy-storage technologies will start to gain momentum. This includes new batteries like GE’s sodium-based Durathon NaMx line, lithium batteries and potentially even compressed air or supercapacitors.
Analyst: Jason Schafer, Bron: 451 Research
Special from 451 Research’s Market Insight Service – DCT – Datacenter Technologies.