Interfacing ATCA Hot-Swap with PCIe Hot-Plug for high-availability instrumentation in critical systems

Highlights: • IPFN developed control instrumentation for the ITER FPSC. • Both ATCA and PCIe were selected to fulfil ITER's HA requirements. • Hot-Plug for card add/removal of adapter cards while keeping the system running. • ATCA extension for PCIe on Fabric Interface does not specify PCIe Hot-Plug. • Availability may benefit from the coordination of ATCA Hot-Swap and PCIe Hot-Plug. - Abstract: High availability (HA) is a key requirement of next generation Fusion devices, targeting steady-state operation. HA is demanded in mission-critical systems, as is the case of experimental Fusion devices and future Fusion power plants, where safety of people, environment, infrastructure and investment is primordial. IPFN (Institute of Plasma and Nuclear Fusion, Lisbon − Portugal) developed control and data acquisition instrumentation based on the Advanced Telecommunications Computing Architecture (ATCA), composed of digitizing (nodes) and switching (hub) boards, which establish a PCIe data network through ATCA's backplane Fabric Interface, connecting to an external host computer. The ATCA standard was selected to achieve a high level of availability, as it is a modular platform of Field Replaceable Units (FRU) used with several redundancy resources. Its mandatory hardware platform management, includes a "Hot Swap" mechanism, which allows FRU insertion and extraction without having to power off the system. From the host/software perspective, ATCA nodes and hubs correspond to PCIe endpoint and switch devices. The PCIe standard specifies a Hot-Plug procedure for device add and removal for adapter cards only, whereas for other form-factors the Hot-Plug mechanism is implementation-dependent and should be defined by the form-factor itself. Although ATCA has an extension for PCIe on Fabric Interface, it does not define Hot-Plug procedures. From an initial Hot-Plug solution developed for IPFN's platform, the paper attempts to establish a relationship between these specifications, which could help to standardize PCIe Hot-Plug implementation for the ATCA form-factor, fostering the development of instrumentation with HA requirements.