The Impact of Terrestrial Radiation on FPGAs in Data Centers

Abstract

Field programmable gate arrays (FPGAs) are used in large numbers in data centers around the world. They are used for cloud computing and computer networking. The most common type of FPGA used in data centers are re-programmable SRAM-based FPGAs. These devices offer potential performance and power consumption savings. A single device also carries a small susceptibility to radiation-induced soft errors, which can lead to unexpected behavior. This article examines the impact of terrestrial radiation on FPGAs in data centers. Results from artificial fault injection and accelerated radiation testing on several data-center-like FPGA applications are compared. A new fault injection scheme provides results that are more similar to radiation testing. Silent data corruption (SDC) is the most commonly observed failure mode followed by FPGA unavailable and host unresponsive. A hypothetical deployment of 100,000 FPGAs in Denver, Colorado, will experience upsets in configuration memory every half-hour on average and SDC failures every 0.5–11 days on average.

Evaluation of the Xilinx Deep Learning Processing Unit under Neutron Irradiation

“…Fortunately, most MPSoC terrestrial applications would not experience failures due to atmospheric neutron radiation. The sensitivity per device to NSEUs is extremely low [2]. However, the radiation effects increase dramatically when MPSoCs are used on large-scale applications (e.g., data centres) or when operating in high-altitude (e.g., airliner's avionics).…”

“…In other words, if the FIT rate of one ICs is X, the overall FIT rate of a system incorporating N such ICs will be FIT overall = X ×N . In [2], the authors estimated that the MTTF due to neutron-induced errors on a hypothetical one-hundredthousand-node FPGA system in Denver, Colorado, would be 0.5 to 11 days depending on the workload. Indeed, projections from technology evolution roadmaps indicate that the MTTF of data centre computing systems may reach a few minutes [16].…”

“…Unfortunately, reprogramability comes at the expense of a large amount of SRAM cells to store the state and configuration of the device, making it vulnerable to Single Event Effects (SEEs) [2], [3]. Specifically, highly-energised particles like protons originating from deep space (i.e., cosmic rays) and our Sun (i.e., solar rays) collide with nitrogen and oxygen atoms of our Earth's upper atmosphere, producing secondary particles like neutrons and muons [4], [5].…”

“…In turn, neutrons with an energy ≥ 10 MeV interact with the atoms of the device's semiconductor material, causing SEEs, especially Neutron-induced Single Event Upsets (NSEUs) or otherwise soft errors. NSEUs can upset (i.e., corrupt) the on-chip SRAM memories of the PS, like the multiprocessor's register file and caches [6], as well as the application (e.g., flip-flops state) and configuration (e.g., programmable routing resources) memory of the PL [2]. NSEUs are not permanent, but their effects can threaten system dependability if not well understood and handled in an MPSoC.…”

Localized Stress Effects on the Single Event Effects Sensitivity of Microelectronics

“…[11][12][13] Another approach is to simulate fault injection to observe how a device responds when a certain, known component is corrupted. 14,15 These techniques significantly speed up the characterization process. However, that advantage is continually outpaced by the increase in size and complexity of modern electronics.…”