The recent Amazon Web Services (AWS) outage highlights the growing dependency we have on data centres and more importantly the web-based services they provide. The service outage lasting five hours was put down to human error by a coder debugging of all things the S3 billing service; this led to the accidental removal of crucial subsystems within the IT infrastructure.
Today, around 20% of errors within a datacentre are due to human error. The remaining 80% are linked to hardware failures and principally within critical power and critical cooling systems as well as general IT systems. As systems become more automated the room for such errors will reduce and thankfully the S3 AWS datacentre outage is still a very rare occurrence.
In the business world a datacentre outage like this affects those operating cloud-based services and for some the loss of five hours of service can lead to major economic loss. Aside from core operating technologies such as those used by cloud-based companies like Slack, Trello, Splitwise and Medium, most cloud-based services operated by companies cover CRM (Customer Relationship Management) and Accounting packages. However with the Internet of Things (IoT) and Industry 4.0 the potential impact of a datacentre outage could be far wider.
Industry 4.0 is a ‘smart factory’ related concept that relates to the adoption of automation and data exchange within manufacturing technologies. Industry 4.0 includes cyber-physical systems, linked to the Internet of things and cloud-based computing and is another concept that is reliant on critical power and cooling systems.
Within industry the adoption of microprocessor-based systems and the wider use today of cloud-based and ‘smart’ systems has lead to the need to revisit how companies look at their power protection strategies. In a typical industrial environment, the electrical power supply is more polluted than within a general office environment due to the use of motor driving inductive loads (leading to brownouts, sags and surges) and systems than can generate spikes and electrical noise. Power outages can also take manufacturing systems off-line but the typical need for battery runtime is more commonly around several milliseconds to ride through small breaks in the mains power supply than the longer times associated with datacentres.
An uninterruptible power supply used within an industrial environment should therefore be more robust in terms of its overall design. Industrial UPS systems may be transformer-based and include galvanic isolation to completely protect the loads from mains borne spikes and electrical noise. The UPS may be fitted with fan filters for dusty environments and may even have different case design.
A typical industrial UPS has an IP20 case design and may be built-into a sub-system control panel. IP20 UPS cabinets are the type most commonly found in office environments. In an industrial environment, the IP rating may be higher with the UPS sat on a plinth (common in the water industry) and fitted with a protective hood to protect it from overhead condensation and water droplets.
In terms of runtime, the uninterruptible power supply may have a smaller battery set or even use super capacitors or a dc flywheel to provide the runtime for a manufacturing process to ride through a momentary power outage. Where longer runtimes are required the battery set may be based on valve regulated lead acid (VRLA) or even Ni-cad (NiCad) battery technologies.
Lithium-ion (Li-ion) batteries provide another alternative battery type and UPS DC power source than opens up a world of connections to renewable power sources and more rapid charge/discharge cycles. If the UPS has a bi-directional converter, the UPS could also act as a virtual power plant (VPP) storing locally generated energy from solar and wind turbines in the battery set connected to the UPS and using this take some of the industrial machinery off-line from the mains power supply grid.
Within an industrial environment a UPS system can provide protection from a wide range of power problems acting as a voltage stabiliser and power conditioner. For some sites, the UPS itself can even optimise the output voltage supplied to the industrial machinery loads to improve overall operating efficiency (voltage optimisation).
The Amazon S3 power outage shows that datacentres are still vulnerable and it is rare for a datacentre operator to comment and pinpoint the exact cause of the outage. Hats off to Amazon for their more open approach. There are many factors to consider when designing a power protection strategy, whether it’s for a cloud-service operating datacentre or an industrial application. Not all commercial UPS systems are suitable for an industrial environment and care has to be taken to ensure the right system is used to keep Industry 4.0 applications running, even when operator error risk potential downtime.