Workloads in the cloud for industrial manufacturers
The cloud is of the big three technology innovations, with Big Data and machine learning (ML), which are powering a new generation of solution and plant infrastructure. The cloud is also a buzzword in modern consultant-speak: Industry 4.0, Smart Manufacturing, digital transformation, and the Industrial Internet of Things (IIoT). The cloud has become the backbone of our personal lives with online shopping on Amazon, applications like Microsoft Office 365, and services like banking.
At least we’ve agreed what to call on-demand computer rental since Amazon first offered virtual machine (VM) and storage services in 2006, followed by Microsoft and Google in 2008. Since then we’ve seen the hype and progress of automation company cloud-based services, and acquisition after acquisition as vendors shore up and expand their online offerings.
The cloud has produced a lot of talk, perhaps too much talk because while the cloud is everywhere in our personal lives, the details of where the cloud will matter first and best to manufacturing organizations is still an ongoing process. It’s more of a roadmap than reality.
Further, there are specific requirements of industrial organizations that preclude some cloud-based deployments. Unlike consumer and general information technology (IT), industrial cloud deployments have specific security requirements. They also need guaranteed availability and require software-as-a-service (SaaS) versions of important enterprise asset management (EAM), manufacturing execution systems (MES), and potentially supervisory control and data acquisition (SCADA) applications (see sidebar).
At the same time, despite the hesitancy and challenges of cloud deployments, there are five current and recognizable patterns of cloud use by industrial manufacturers, with each described in detail below:
- Born on the cloud: IIoT
- Lift and shift historians
- Data lakes
- Cloud reporting systems
- Quick start analytics.
These patterns, or use cases, are called “workloads” by the cloud platform vendors, as in “early cloud workloads for IT departments included low-cost, long-term storage, and on-demand computing resources.”
The list of cloud workloads will grow as issues such as the availability and acceptance of SaaS plant applications like MES are addressed. Using these as specific examples of the cloud in industrial use will help enable the transition from an amorphous cloud to a discussion of specific tradeoffs and benefits.
1. Born on the cloud: IIoT
The first cloud workload is the IIoT use case where new sensors are deployed on assets with the telemetry required to pipe their data to the cloud for storage, applications and analytics.
Alternatively, the assets are existing, but there are new sensors deployed, “lick and stick” as one vendor referred to them. This born on the cloud scenario is reminiscent of countries that skipped widespread phone use prior to the cell phone and went straight to a cell-based model vs. fully deploying a wired telephone network.
- Born on the cloud IIoT benefits: A fit for greenfield monitoring scenarios or expanding visibility in an existing facility to additional assets or resources.
- Downside: It’s still a work in process because end users must make decisions on each piece in the solution stack including sensor, device management, gateway, security, communications layer, cloud vendor, data storage, etc.
2. Lift and shift historians
“Lift and shift” is taking an IT workload and moving it out of the data center and into the cloud, running the application on a virtual machine (VM), which is an instance of an operating system decoupled from the underlying hardware. Virtual machines enable one physical server to host many virtual machines, which run many applications, improving hardware utilization versus a dedicated server for each application deployment.
The drive for moving application workloads out of on-premise data centers is lower costs: the fully burdened cost of a server in a data center can be 50 times the price of the server itself. Amazon’s AWS team, for example, advises companies to aim for a “zero square foot data center” with all applications running in the cloud (Figure 1). Companies have been moving IT applications to the cloud for years, including email, enterprise resource planning (ERP) and accounting systems. Historians, from a cost-savings perspective, are waiting their turn.
- Lift and shift historian benefits: If correctly executed, the users of the historian won’t see any impact on their use of historian data, and cloud-based historians are more accessible to IT.
- Downside: Internet bandwidth, historian read access, and security are issues to be addressed — but the cost advantages of cloud-based deployment make this more a “when” than an “if” discussion.
3. Data lakes
The big bang option for cloud workloads is building a data lake, which is an aggregation of unlike data types in one system, for example a pharmaceutical company combining sensor data from historians, laboratory information management systems data, batch data, quality data and other areas, to enable a global view into operations and business outcomes by data science and IT. The benefits of this approach may be considerable, but the list of challenges is just as long.
Data lakes are always bespoke in the details — schema, data requirements, use cases, and other details — and they are complex and expensive.
As one example, moving time series data doesn’t make it easier for analysis. From an end user data analytics perspective, moving historian data to a data lake, for example, doesn’t solve a problem; it just changes the location of the data.
The second biggest challenge with data lakes is the timing and infrastructure used to copy and update the data lake while the source data is constantly flowing from the system. Too slow of an update wastes insight opportunity; too fast can be very expensive to implement and still may not guarantee high-speed concurrency.
- Data lake benefits: The aggregation of disparate data types enables access by data science and other IT experts to find new insights to improve production and business outcomes via a view across the entire organization (Figure 2).
- Downside: The cost, time to insight, and company-specific challenges of data lake projects mean this is not an effort to be taken lightly: years and millions of dollars are the right denominations for implementation.
4. Cloud reporting systems
Three audiences need access to time series or production data for analytics and insights. First is data scientists, typically associated with data lakes and IT. The second is process engineers and plant employees, typically working with data in historians and manufacturing applications. Third is internal company business analysts who need reporting and “known” views on production processes.
For the third audience, visualizing data is done with business analytics products and SaaS applications. The challenge is giving these users access to the right data given the challenges visualization products have with time-series data.
The answer is creating tables of process data and storing them in a relational database service for easy access by business intelligence products. These data tables don’t enable all the flexibility required by process engineers, but for known questions like overall equipment effectiveness (OEE) or production accounting reporting, they provide access and flexibility.
- Cloud reporting system benefits: Using cloud-based relational database services to store data that answers known or defined questions for non-production specialists.
- Downside: The architecture creates an infinite loop of requests back to IT based on data structure and context to chase queries in new or unexpected directions.
Quick start analytics
The benefits of the SaaS application model are known to any consumer who has bought something online and any employee who has worked with online applications in their workplace: fast access, browser-based user experience, and little, if any, deployment overhead. This model is available for manufacturing applications including MES systems and will expand in the years to come.
The highest end-user priority, however, is improved analytics software to enable faster and deeper insights on expanding data volumes in manufacturing. End users need more and new types of insights, such as predictive analytics, and the kind of collaboration and knowledge capture features offered in modern workplace applications.
Advanced analytics applications leveraging machine learning to accelerate insights are in high demand. Applications offering this functionality as SaaS offering for fast deployment and low IT costs are of particular interest as first use cases for companies exploring cloud-based opportunities for innovation (Figure 3).
- Quick start analytics benefits: End users can implement and access advanced analytics for their data on premise or in the cloud with little to no IT touch. In particular, they can do so without first moving, copying, or changing their system of record (one or multiple historians).
- Downside: Like all cloud deployments, bandwidth and security are requirements for successful implementation, but there is limited downside with subscription models for software licensing. If it doesn’t work, the user can turn it off.
These five examples of using cloud platforms to benefit industrial manufacturers shows the cloud is more than its hype as the center and component of everything interesting. These workloads will expand as objections are overcome and new services are offered, but they already provide a short-term opportunity for manufacturers to realize immediate value.
SaaS manufacturing applications
As the quantity and quality of cloud applications for managing manufacturing operations increases, a new category of software has emerged: cloud industrial software. These are “cloud native” applications, meaning they have been designed from the ground up specifically for the cloud, as opposed to cloud versions of existing on-premise applications.
What they have in common are promises of lower cost, rapid deployment, and opportunities for return on investment (ROI) from the untapped value of the data in on-premise systems. In a typical application, data is stored in the cloud for access worldwide (sidebar photo below).
Here are a few examples – of the many – in several popular categories of SaaS offerings:
Manufacturing applications: 42Q – 42Q is a cloud-based MES and computerized maintenance management system (CMMS) that promises the benefits of cloud efficiencies and cost advantages with global visibility into production and maintenance, metrics and analytics.
Cloud data storage: OSIsoft Cloud Services (OCS) – OCS is intended to aggregate and augment operations data. The vision is for end users to combine and store multiple data types from multiple facilities in the cloud for real-time planning and monitoring that drives process improvement.
Manufacturing applications: Tulip –Tulip enables business users to build manufacturing apps to improve process outcomes without writing code while still taking advantage of back-end systems and data sources.
Machine vision: Spyglass Visual Inspection – An AI-powered platform that augments existing systems on the factory floor by delivering improved accuracy in detecting defects. Spyglass deploys in the cloud for a Lean approach to leveraging advances in emerging AI and machine vision technologies to drive continuous quality improvement.
Vibration analytics: Petasense – Makes industrial machines smarter by offering a stack of sensors, connectivity, and cloud solutions to enable web and mobile applications to provide acoustic-based insights for improving asset reliability and predictive maintenance. The benefit for end users is reduced downtime and lower repair costs on assets and production facilities.
A general benefit of SaaS-based manufacturing applications is a low-cost trial or proof-of-concept phase with connections to live operations data, along with low IT overhead. SaaS also allows prototyping and iteration without large capital investments in hardware or systems. With SaaS, how far or fast a manufacturer proceeds is a matter of choice without the constraints if there are significant sunk costs.
Keywords: Cloud, cloud software, Industrial Internet of Things (IIoT)
The cloud is powering manufacturing growth and improving connectivity.
Use cases involving the cloud include the Industrial Internet of Things (IIoT), data historians and reporting systems.
Software-as-a-service (SaaS) manufacturing applications are growing thanks to cloud industrial software.
What use case example has your plant used in relation to the cloud and what benefits did it provide to the facility?