Cloud computing, virtualization, solid-state data storage, and other technologies have become the building blocks of today’s data centers, but how these building blocks are assembled and applied varies with each industry. Manufacturing, for example, is applying the latest data center technologies and best design practices in order to increase productivity through automation, maintain closer control of the supply chain, and use data processing to more closely manage manufacturing processes.
The manufacturing sector is continuing to grow, accounting for more than 16 percent of global gross domestic product and 14 percent of employment. However, manufacturing services have become much more diversified over time, encompassing everything from supply chain to advertising. In the United States, for example, every dollar spent on advertising includes 19 cents spent on ancillary services. In some manufacturing sectors, more than half of employees work in service functions such as R&D and support staff.
As part of manufacturing services, data center functions are being used increasingly to consolidate operations and integrate diverse manufacturing and service functions. Data centers also are finding a larger role in automating routine manufacturing functions and analyzing data in order to improve operations and support services.
The Internet of Things and Software-Defined Networking
One innovation that has had a significant impact on manufacturing is the Internet of Things (IoT). Machine-to-machine communication enables new possibilities in monitoring and managing production and in predictive analysis and maintenance.
When you consider the cost of production compared to losses due to an outage or factory downtime, the use of predictive analytics is compelling. Data center operations are going to be extended in order to gather machine performance metrics for real-time management as well as later analysis. For example, control and telemetry systems use tags or data points for values such as operating temperature, amperage, and flow rates. IoT technology is enabling additional operational tasks such as lubrication temperature, vibration, and torque values.
In a large manufacturing facility, you can typically have 1 million tags at 4 bytes each, or 32 bits per tag. If you have to sample those tags every minute, then the data rate is about 534 kbps. As you add more tags in order to monitor operating functions, your tag count can increase to 5 million tags or more, which demands a data rate of 160 Mbps or more. When you add even more functions, such as 3D printing, then data processing requirements will increase exponentially. That’s going to mean a new type of data processing architecture that makes the most of the cloud, virtualization, and localized processing.
In order to be effective for plant operations monitoring and management, these tags have to be processed in real time. You now need an infrastructure that has the elasticity to handle more data, the speed to process data in real time, and the ability to issue automated maintenance and operating instructions based on analytics findings. This is going to require handling some analytics locally for real-time response and incorporating cloud computing and virtualization in order to gain the flexibility and storage capacity to accommodate ever-larger data sets.
The same principles of software-defined networking (SDN) in the data center are now being used in manufacturing facilities. By abstracting management tools from physical resources, the data center manager can now move applications to any physical system for local data processing and real-time response without having to reconfigure the application itself. Using SDN abstraction allows you to accommodate real-time monitoring and administration and still use centralized management.
A Foundation for Big Data
With the boom in IoT and the exponential growth of machine data gathered from the factory floor, there are also new possibilities for big data analytics. According to LNS Research and MESA International, factory managers are looking to big data in order to improve production in a number of ways:
- 46 percent want better forecasts for product and production demands
- 45 percent want to better understand plant performance across multiple metrics
- 39 percent want to provide service and support to customers faster
And just as real-time analytics can help with systems operations, the ability to measure performance to the machine level also means that managers can get insight into how each machine is operating, including quality, performance, and operational variances. This level of detailed insight is useful for streamlining workflows and ensuring compliance.
In order to streamline operations, reduce operating costs, and increase the quantity and quality of output, manufacturers have to design data centers that can handle advanced analytics. That means assessing the amount of data available for analysis, including IoT data, and determining the best way to store and process that information. For example, historical data can be stored in the cloud for access and analytics later, while data that relate to operations can be cached locally for immediate analysis. That type of architecture is going to require cloud data processing and archiving and local high-performance, solid-state data storage and processing.
As manufacturers demand more computing power, they are striving to cut operating and capital expenses through virtualization. Virtualization can maximize the hardware investment while providing a scalable platform that can accommodate changing computing needs, including adding more processing and storage resources in the cloud.
The new manufacturing data center has become a business enablement platform, providing end-to-end visibility and control over every aspect of production and delivery. Data center operations need to manage raw materials, factory processes, shipping, telecommunications, accounting, and a host of support services. And they have to be elastic enough to accommodate changing data demands and processes, expanding data storage as needed, and moving processes closer to operations for real-time monitoring and control.