Workload Consolidation

The ability to safely host functions with different safety integrity levels on a single platform. A Mixed-Criticality System enforces strict temporal and spatial separation.

• Temporal Isolation: A real-time hypervisor guarantees that a safety-certified control loop executes within its microsecond-level deadline regardless of the load on general-purpose operating systems.
• Spatial Isolation: Memory regions are strictly partitioned using hardware virtualization extensions to prevent a crash in a Linux-based analytics container from corrupting the memory space of a safety controller.
• Freedom from Interference: The architecture must demonstrably prove that non-critical functions cannot impact the performance or safety of critical functions, a key requirement for IEC 61508 certification.

Collapses multiple physical fieldbus networks into a single converged Ethernet fabric. Time-Sensitive Networking (TSN) is the essential enabler for this consolidation.

• Traffic Scheduling: TSN uses a gatekeeper mechanism (IEEE 802.1Qbv) to create protected time slots for cyclic control data, ensuring it bypasses queues filled with best-effort video or analytics traffic.
• Precision Synchronization: The Precision Time Protocol (PTP) synchronizes clocks across all consolidated nodes to sub-microsecond accuracy, allowing coordinated motion control over the same wire that carries HMI traffic.
• OPC UA Pub/Sub: This protocol leverages TSN to deliver a scalable, connectionless data bus where a single publisher can multicast sensor data to multiple consolidated subscriber functions without a central broker.

Treats the consolidated server as a single, version-controlled entity rather than a collection of individually managed components. This is the principle of Immutable Infrastructure.

• Golden Images: The entire software stack—hypervisor, real-time OS, soft PLC, and edge analytics—is pre-integrated and tested as a single bootable image.
• Atomic Updates: Patches are never applied in-place. Instead, the entire system is replaced with a new, validated image, eliminating configuration drift and ensuring every deployed unit is identical.
• Infrastructure as Code (IaC): The desired state of the consolidated workloads is defined in declarative configuration files, enabling automated, repeatable provisioning and disaster recovery.

Maintains continuous operation of all consolidated functions through automated failover mechanisms that preserve state across all virtualized workloads.

• Fault Tolerance (FT): A secondary host runs a shadow instance of the control VM in lockstep execution. If the primary fails, the secondary takes over with zero state loss and no process interruption.
• Live Migration: The hypervisor can move a running control workload from a failing hardware node to a healthy one without stopping the process, enabling proactive maintenance.
• Unified Namespace (UNS): A centralized data hub ensures that upon failover, the new active instance immediately has access to the complete, current state of all data sources, preventing information gaps.

Offloads infrastructure services from the host CPU to specialized hardware, preserving compute cycles for core control and analytics functions.

• Data Processing Unit (DPU): A DPU handles the entire virtual switch, network security, and storage virtualization, freeing the CPU to focus exclusively on executing control logic and machine learning inference.
• Hyperconverged Infrastructure (HCI): Abstracts and pools direct-attached storage from the server into a virtual SAN, eliminating the need for external storage arrays and consolidating the entire data path within the edge node.
• GPU/NPU Partitioning: A single physical GPU can be virtualized into multiple instances, simultaneously accelerating a computer vision quality inspection model and rendering a 3D HMI, all on the same consolidated hardware.

The foundational virtualization layer that makes workload consolidation possible. A bare-metal hypervisor partitions physical CPU cores, memory, and I/O to run a real-time operating system (RTOS) for control logic alongside a general-purpose OS (GPOS) for analytics and HMI on the same server. It guarantees microsecond-level determinism through hardware-assisted isolation, preventing non-critical workloads from starving safety-critical tasks of compute cycles.

The architectural pattern enabled by workload consolidation where functions of different safety importance coexist on one platform. A single edge server might simultaneously run:

• SIL-3 safety logic with strict temporal isolation
• Soft PLC runtime for non-safety sequential control
• Linux-based analytics container for OEE dashboards
• Protocol gateway translating Modbus TCP to OPC UA Spatial and temporal partitioning ensures a memory leak in the analytics container cannot corrupt the safety controller's execution.

A critical technique for consolidated architectures that binds a specific virtual machine vCPU or container process exclusively to a dedicated physical processor core. This eliminates the non-deterministic scheduler jitter caused by the hypervisor migrating threads between cores. In a typical consolidated setup:

• Cores 0-1: Pinned to the real-time control VM
• Cores 2-3: Pinned to the HMI and historian VM
• Cores 4-7: Shared pool for non-critical containers This guarantees the PLC runtime never waits for a cache line evicted by a database query.

The network infrastructure required when consolidated servers communicate with distributed I/O. IEEE 802.1 TSN standards provide deterministic, low-latency Ethernet delivery over converged networks. Key sub-standards include:

• 802.1AS: Precision time synchronization across all devices
• 802.1Qbv: Scheduled traffic gates that reserve time slots for control frames
• 802.1CB: Seamless redundancy via frame replication Without TSN, consolidated controllers cannot guarantee that a motion control packet arrives within the required cycle time when sharing a wire with video streams.

A PCI Express specification that allows a single physical network interface card (NIC) to present itself as multiple independent virtual functions (VFs). Each consolidated VM gets direct, hypervisor-bypass access to its own virtual NIC, eliminating the latency overhead of software-emulated networking. This is essential when a virtualized PLC must process EtherCAT or PROFINET IRT frames with sub-100-microsecond jitter on a shared server that also streams video to the cloud.

The data architecture that makes consolidated systems navigable. A UNS aggregates all data from the virtualized PLCs, HMI historians, and analytics engines into a single structured topic hierarchy—typically implemented via MQTT Sparkplug. Any application subscribes to Plant1/Line3/Cell2/Press/Temperature without knowing which physical or virtual controller generates it. This decouples data producers from consumers, allowing workloads to be migrated between hosts without reconfiguring every downstream dashboard.