Cloud-native zero-downtime manufacturing execution systems in the automotive industryCloud-native zero-downtime manufacturing execution systems in the automotive industry

In the first part of this MES blog series, Timo Busert describes how the potential of an MES is systematically determined and why the budget is the result of a potential analysis rather than its starting point. In the second part, Christian Neumüller examines SAP Digital Manufacturing Cloud (DMC) in the regulated life sciences environment and highlights where SAP DMC is currently GxP-compliant and where gaps remain. This final part takes us into a world where delivery dates, product variety and production cycles are crucial: automotive manufacturing.

Why automotive manufacturing ticks differently

Standing in front of a body press or on an OEM final assembly line, you can feel the rhythm of production: one vehicle every 60 seconds or so, cycle after cycle, variant after variant. Here, the MES is not a reporting tool, but the leading IT system and the backbone of the line. If it fails, the line stops, and the downtime costs of such a stoppage are discussed by the board, not by the controlling department.

At the same time, electrification, Gigafactory scaling, software-defined vehicles and new compliance requirements – from the Corporate Sustainability Reporting Directive to the Battery Passport and the Supply Chain Act – demand significantly more data intelligence than ten years ago. Since then, the expectations placed on production IT have been paradoxical: to become more stable whilst simultaneously rolling out new functions much more quickly. In this balancing act, traditional MES monoliths are reaching their limits.

Why the MES monolith has had its day in the automotive industry

Most OEMs’ production IT landscapes were designed in the 2000s and have since been overwritten on a plant-specific basis. Release capability is theoretically possible in such systems, but in practice it is tied to four plant holidays per year. Every major update means a planned weekend shutdown, a rollback plan in SharePoint and an escalation chain right up to the plant manager. Every new variant, every battery assembly line, every additional customer feature becomes a change request project. As long as the growth in variants remained linear, this was manageable. With software-defined vehicles and shorter model cycles, this becomes difficult.

The alternative is not another product, but an architecture: cloud-native, modular, API-first, with clear responsibilities per domain (e.g. order control, traceability, quality or worker guidance) and with edge components where latency and offline capability require them.

Implementing ‘cloud-native’ correctly

‘Cloud-native’ has become a term that is often used even when what is actually meant is ‘we’ve moved the VM’. In automotive production, however, the difference is crucial. Cloud-native means containerised, horizontally scalable, composed of distributed microservices, with automatic load balancing and self-healing. Redundancy is not the expensive hot-standby data centre that remains unused during normal operation, but active instances distributed across multiple zones that are utilised during regular operation and absorb the failure of a single zone.

However, this does not mean: everything into the public cloud. Signal processing at the machine is subject to latency and availability requirements that cannot be compromised when using the open internet. A viable architecture therefore makes a clear distinction. Edge components handle latency-critical, deterministic tasks directly on the shop floor and act as a buffer in the event of a connection loss. The cloud consolidates, analyses and scales. In this vision, hybrid is not a stopgap solution, but the target architecture.

“Zero downtime” for updates during ongoing production

The biggest cultural shift for many OEMs is the end of the maintenance window. With a cloud-native approach, updates become a non-event: rolling deployments, blue-green switches and feature toggles ensure that new functions are rolled out during live operations and can be quietly rolled back if necessary. This is not merely a theoretical convenience, but a prerequisite for ensuring that security patches are applied promptly without forcing production managers into a conflict of interest between operational reliability and IT security.

Added to this are automated backup and recovery mechanisms, continuously tested failover scenarios and active redundancy. These not only reduce the number of unplanned downtimes but also the total cost of ownership. Reliable studies indicate TCO savings of up to 30 per cent here – primarily through the elimination of duplicate infrastructures, lower operating costs and more efficient use of resources.

“Publish & Subscribe” for decoupled machine connectivity

Modern, cloud-enabled machine connectivity is the key to transforming heterogeneous brownfield environments into a future-proof zero-downtime architecture: Instead of extending a jumble of PLC protocols, fieldbuses and proprietary interfaces directly into every MES module, a dedicated integration layer bundles the communication, translates it into standardised formats (e.g. OPC UA, MQTT or REST) and makes it available as uniform, semantically enriched events.

Machines publish their signals, whilst MES functions, traceability, quality, maintenance or analytics specifically subscribe to the data streams they require. Machine connectivity and business logic are thus consistently decoupled. In combination with an edge layer on the shop floor, data can be buffered and pre-aggregated, ensuring that production lines continue to run smoothly even if central systems are temporarily unavailable. Zero downtime is thus not just a cloud feature, but is designed right down to the machine level. New lines and machine types are connected to this integration layer once and transferred to a unified namespace, which structures data hierarchically and makes it available as a central source under a uniform addressing scheme. This significantly reduces integration effort and time-to-market.

In this way, machine connectivity transforms from a tedious project side-issue into a strategic platform: capable of brownfield integration, capable of being released during live operation, and featuring a stable data foundation on which analytics and AI use cases can be implemented quickly and scalably.

‘Semantic Reasoning’ for data analysis

A cloud-native MES thus generates a continuous data flow with a structure that can actually be utilised for artificial intelligence. This sounds like slide 3 of any digitalisation strategy, but in practice it is the real lever. If a torque wrench at an assembly station shows a slight but statistically significant drift in its tightening values, maintenance should not receive an Excel list on Monday. Instead, the system suggests a tool change before the vehicle is flagged at the quality gate.

Particularly through the combination of various data sources in a structured RDF knowledge graph, AI-based, high-performance stream reasoning enables the implementation of versatile use cases that companies have been dreaming of for years: predictive maintenance, anomaly detection, root-cause reasoning for quality issues, dynamic bottleneck detection, adaptive material control and much more. The difference from traditional MES lies not in the volume of data collected, but in the fact that the loop of observation, evaluation and intervention closes within minutes, and that new AI models can be integrated into a cloud-native system without halting production.

“Custom Development” or Standard MES

In recent years, many German OEMs have opted for custom development when it comes to MES rather than another off-the-shelf product – but why is that? Our honest assessment: the best custom development projects are not about reinventing the wheel. They deliberately orchestrate standard building blocks such as IIoT platforms, workflow engines, traceability components or ML frameworks, and differentiate precisely where the company has a competitive advantage: typically in variant management, detailed planning, integration into the OEM-specific PLM and ERP landscape, and in shop floor guidance processes on the production line.

A pure product MES does not deliver this level of customisation, as it must, by definition, remain attractive to all potential customers simultaneously. A custom-developed MES on a cloud-native basis turns the equation on its head: standard where standard is sufficient; in-house development where the competitive advantage arises; and an architecture that remains deliberately interchangeable, so that today’s decisions do not become tomorrow’s lock-in.

Conclusion: The direction is set; implementation is key

Cloud-native MES systems deliver flexibility, reliability, efficiency and a robust foundation for AI-supported, data-driven decisions – in a market environment that tolerates neither stagnation nor long response times. For OEMs, the path is thus clear, and the strategic decision has long since been made in many corporations. What remains open is the implementation. It rarely fails due to the technology, but rather on three points: a lack of architectural governance, underestimated edge integration, and an acceptance culture that does not provide for continuous delivery. Those who tackle these issues consistently will gain an MES that not only survives the next SOP but also carries through several platform generations. The rest is discipline and choosing the right partner.

How adesso supports the journey to digital production

As a strategic partner, adesso supports manufacturing companies on this journey – from assessing the architecture of existing MES landscapes, through building cloud-native target architectures with a clear edge, container and API strategy, to our proprietary zero-downtime platform in series production. We combine MES process knowledge from industry-specific project work with robust cloud and AI expertise, delivering not just concepts but scalable production software right down to the shop floor. Our aim: production IT that follows the takt time, not the other way round.

This concludes our three-part MES blog series. However, even the most detailed online reports and websites can never replace a personal discussion about your situation and goals. So if you’d like to explore the topics from Part 1 (ROI and potential analysis), Part 2 (SAP DM in a regulated environment) or Part 3 (cloud-native zero-downtime MES in the automotive sector) in more detail for your company, please get in touch – ideally before the next release Friday rolls around.