Entropy is a fact of nature, and as things wear out or operate below specifications, they need to be repaired or replaced. In a business-critical operation such as a data center, the objective is to identify and address potential points of failure before equipment actually fails. That’s why predictive maintenance should be part of any IT operation.
The Uptime Institute says that the average cost of one hour of data center downtime is $138,000. For larger e-commerce providers such as Amazon, the company loses an average of $1,104 for every second it is down. And among Fortune 500 companies, 59 percent said they experience a minimum of 1.6 hours of downtime every week, which translates into a loss of $46 million in labor costs alone.
With so much at stake, it’s no wonder that CIOs are investing more in predictive maintenance technology. Given the cost of data center downtime, investing in new strategies to prevent system failure more than justifies the expense, and smart solution providers are finding new ways to provide the right tools to help with both preventative and predictive maintenance.
The Difference Between Predictive and Preventative
Every data center has preventative maintenance: scheduled maintenance to assess equipment and software performance, just like you have scheduled data backup. Maintenance teams will schedule system upgrades and maintenance on a weekly or monthly basis so that revisions and repairs create minimal operational disruption.
Many issues can be anticipated , so preventative maintenance, performed to head off these types of problems, is a cost-effective way to keep the data center operational and extend the useful life of data center equipment. However, to keep the data center at peak performance, you need predictive maintenance too.
Predictive maintenance applies analytics to assess equipment and determine potential points of failure based on actual system performance. A computerized maintenance management system (CMMS) monitors machine components, looking for anomalies prior to failure. CMMS is used in conjunction with scheduled maintenance and can be invaluable for ensuring uninterrupted service, because it measures the actual condition of the machine rather than using historical data. And the cost of fixing an issue once it manifests itself is much higher than that of predictive maintenance.
Harnessing the Internet of Things
Predictive maintenance is largely driven by sensors in data center equipment that report operating status. Monitoring critical mechanical and electrical systems, for example, requires tracking the performance of assets such as:
- Heating, ventilation and air-conditioning (HVAC) systems for temperature control and humidity
- Chiller performance
- Cooling towers that reject heat from chillers or HVAC systems
- Pumps that support cooling and dehumidifiers
- Generators that provide back-up power
- Uninterruptible power supplies and automatic transfer switches
- Power distribution panels
Each of these components has internal sensors that measure performance and report back to a centralized CMMS. When the sensors indicate that a component is operating below a designated threshold, it flags that component for assessment and repair.
The same principle can be applied to almost every aspect of the data center. Leveraging information such as run time, energy consumption, temperature and output identifies potential points of failure before equipment actually fails. In essence, you are applying big data predictive modeling techniques to the Internet of Things (IoT) in order to predict points of failure. For example, by modeling current equipment functions against historical data, you can measure performance against an operational baseline and reduce data center disruptions.
Implementing Predictive Maintenance
In order to help customers implement a predictive maintenance model, you need to develop a strategy centered on data gathering and analysis. Chances are that your clients are already gathering data in order to assess data center operations, but you need to help them go deeper and start developing predictive analytics based on IoT data. This may mean a cultural as well as an operational shift for the IT department to start gathering that IoT data.
Here are three key steps to consider as you implement a predictive maintenance model for your customers’ data centers:
- Investing in IoT – To support predictive maintenance, your customers will need to start investing in more intelligent hardware as well as CMMS software. Recommend adding new hardware that has IoT-monitoring capability so that you can incorporate it into the preventative maintenance strategy. Investing in smart machines from the outset will mean fewer unpredicted failures in the future, so even if intelligent systems are more expensive, the lower total cost of ownership should warrant the higher capital expenditure.
- Provide data experts – The facility managers can set up the predictive maintenance infrastructure, but they shouldn’t be expected to interpret the data. Provide the analytics expertise in order to facilitate data collection and analysis, just as you would with any big data project. Data can be stored in the cloud for access and analysis, so your experts can be anywhere, which makes it easier to offer predictive maintenance as a service.
- Adopt a collaborative maintenance strategy – Data gathering and analysis comprise the first step, but you have to make the analytics actionable. Be sure to share the data with the right parties. Data cannot be siloed if the predictive maintenance model is to work. Gather data from all sources, isolate pending problems and then work with the IT team in order to address pending failures. A reporting infrastructure needs to be established with alarms and notifications to prioritize predictive maintenance events. Also, be sure to establish a maintenance hierarchy so you can alert IT staff when they can deal with routine problems, as opposed to calling in a specialist.
Predictive maintenance can be a real lifesaver for your clients and a new value-added service for your business. Whether you help your customers develop their own predictive maintenance strategies or offer predictive maintenance as a service, both you and your customers will realize huge returns.