eFPGAs Accelerate Data-Centric Processing

With the ever-increasing requirements to manage and process enterprise data, system architects are looking closer than ever at programmable logic, a technology to make computing much more efficient and secure.

While traditional processors force data into their pipelines through a complex hierarchy of caches, programmable logic makes it possible to construct data pipelines. Data can flow seamlessly from node to node with a combination of custom logic circuits and DSP engines manipulating data elements as they pass through. Each element is forwarded to the next node that requires it. As needs change, the fabric can be rewired with a new configuration, providing better support for data-centric applications than control-intensive code better suited for microprocessors.

While a viable solution, standalone FPGAs incur a power penalty because data is moved on and off chip to more specialized ASICs. A more effective approach is to embed FPGA technology inside the ASIC to satisfy the constraints of energy efficiency, size and performance in one package.

With eFPGA technology, functions that cannot be deployed in an ASIC because they require reconfiguration or are based on changing standards can be hosted in the programmable fabric embedded in the SoC. For machine-learning applications, those functions can be dedicated compute arrays for convolution kernels or max-pooling calculations. By embedding programmable logic in an SoC, large power savings can be made by avoiding the need to transfer data on and off chip.

Whereas containers and virtualization provide effective support for secure operation in the core cloud because these systems can take advantage of good physical security, eFPGAs offer another advantage that is apt for a remote cloudlet or edge-computing environments. Devices on the edge of the network require greater levels of hardware protection since it is easier for attackers to break into the enclosure and tamper with systems sitting in roadside cabinets or service rooms. Edge-computing systems have less support from administrators who monitor server behavior and track evidence of network-based intrusion, including malicious workload uploads to perform side-channel attacks.

Integrating security functions into the hardwired logic that surrounds eFPGA cores makes it possible to support encrypted uploads of virtual circuits into the fabric and continually monitor them for potential breaches. Logic in the host ASIC can ensure separation of programmable functions uploaded by different users and prevent them from spying on each other.

Having both security and programmable logic integrated on-chip makes it difficult for an attacker with physical access to the system to eavesdrop on communications. With the eFPGA integrated with the CPUs, the compute functionality of entire services can be isolated to the eFPGA, limiting the amount of information sent off chip. Communications with other services can be performed using strong encryption facilities in the hardwired logic. As a result, the eFPGA concept supports the security architecture suitable for the needs of edge computing.

The mix of hardware flexibility and reconfigurability of an eFPGA with the security and performance advantages of an ASIC makes eFPGAs vital technology for data-centric processing both at the edge and cloud.