Networks-on-Chip constitute the interconnection architecture of future, massively parallel multiprocessors that assemble hundreds to thousands of processing cores on a single chip. Their integration is enabled by ongoing miniaturization of chip manufacturing technologies following Moore's Law. It comes with the downside of the circuit elements' increased susceptibility to failure. Research on fault-tolerant Networks-on-Chip tries to mitigate partial failure and its effect on network performance and reliability by exploiting various forms of redundancy at the suitable network layers. The article at hand reviews the failure mechanisms, fault models, diagnosis techniques, and fault-tolerance methods in on-chip networks, and surveys and summarizes the research of the last ten years. It is structured along three communication layers: the data link, the network, and the transport layers. The most important results are summarized and open research problems and challenges are highlighted to guide future research on this topic.

Methods for fault tolerance in networks-on-chip

The structural redundancy inherent to on-chip interconnection networks [networks on chip (NoC)] can be exploited by adaptive routing algorithms in order to provide connectivity even if network components are out of service due to faults, which will appear at an increasing rate with future chip technology nodes. This paper is based on a new, fine-grained functional fault model and a corresponding distributed fault diagnosis method that facilitate determining the fault status of individual NoC switches and their adjacent communication links. Whereas previous work on network fault-tolerance assume switches to be either available or fully out of service, we present a novel adaptive routing algorithm that employs the remaining functionality of partly defective switches. Using diagnostic information, transient faults are handled with a retransmission scheme that avoids the latency penalty of end-to-end repeat requests. Thereby, graceful degradation of NoC communication performance can be achieved even under high failure rates.

Fault Tolerant Network on Chip Switching With Graceful Performance Degradation

Continuing decrease in the feature size of integrated circuits leads to increases in susceptibility to transient and permanent faults. This paper proposes a fault-tolerant solution for a bufferless network-on-chip, including an on-line fault-diagnosis mechanism to detect both transient and permanent faults, a hybrid automatic repeat request, and forward error correction link-level error control scheme to handle transient faults and a reinforcement-learning-based fault-tolerant deflection routing (FTDR) algorithm to tolerate permanent faults without deadlock and livelock. A hierarchical-routing-table-based algorithm (FTDR-H) is also presented to reduce the area overhead of the FTDR router. Synthesized results show that, compared with the FTDR router, the FTDR-H router can reduce the area by 27% in an 88 network. Simulation results demonstrate that under synthetic workloads, in the presence of permanent link faults, the throughput of an 8 8 network with FTDR and FTDR-H algorithms are 14% and 23% higher on average than that with the fault-on-neighbor (FoN) aware deflection routing algorithm and the cost-based deflection routing algorithm, respectively. Under real application workloads, the FTDR-H algorithm achieves 20% less hop counts on average than that of the FoN algorithm. For transient faults, the performance of the FTDR router can achieve graceful degradation even at a high fault rate. We also implement the fault-tolerant deflection router which can achieve 400 MHz in TSMC 65-nm technology.

Addressing Transient and Permanent Faults in NoC With Efficient Fault-Tolerant Deflection Router

A novel strategy to detect interconnect faults between distinct channels in networks-on-chip is proposed. Short faults between distinct channels in the data, control and communication handshake lines are considered in a cost-effective test sequence for Mesh NoC topologies based on XY routing.

/pdf/a-high-fault-coverage-approach-for-the-test-of-data-control-2rqnbqs3t0.pdf

A High-Fault-Coverage Approach for the Test of Data, Control and Handshake Interconnects in Mesh Networks-on-Chip

Reliability, Availability and Serviceability of Networks-on-Chip

The paper proposes a new concept of diagnosing faulty links in network-on-a-chip (NoC) designs. The method is based on functional fault models and it implements packet address driven test configurations. As previous works have shown, such configurations can be applied for achieving near-100 per cent structural fault coverage for the network switches. The main novel contribution of this paper is to extend the use of test configurations for diagnosis purposes and to propose a method for locating faults in the NoC interconnection infrastructure. Additionally, a new concept of functional switch faults, called link faults, is introduced. The approach is well scalable (complexity is square root of the number of switches) and it is capable of unambiguously pinpointing the faulty links inside the switching network.

Test Configurations for Diagnosing Faulty Links in NoC Switches

Over the past few years, Network-on-a-Chip (NoC) has become increasingly popular as a scalable interconnect infrastructure for IP cores. Simultaneously to developing new design paradigms, testing strategies for such network architectures have to be considered. The previous works on testing NoCs have been mainly based on general purpose Designfor- Testability (DfT) approaches and there is a lack of test algorithms dedicated to on-chip networks. The main contribution of this paper is a well-scalable external test method, where insertion of wrappers and scan paths will not be required. The paper proposes an external test method for NoC based on functional fault models, which targets single stuck-at faults in the network switches. Furthermore, 100 per cent of delay faults, opens and shorts between adjacent interconnection lines are covered by the method. The approach allows to reach higher fault coverages in comparison to the recent DfT based solutions.

An External Test Approach for Network-on-a-Chip Switches

The study proposes a new concept of test and diagnosis in regular mesh-like network-on-a-chip (NoC) designs. The method is based on functional fault models and it implements packet address driven test configurations. As it will be shown, such configurations can be applied for achieving near-100% structural fault coverage for the network switches. Additionally, a concept of functional switch faults, called link faults, is introduced. The approach is scalable (complexity grows linearly with respect to the number of switches) and it is capable of unambigously pinpointing the faulty links inside the switching network. Current paper also presents a set of design-for-testability (DfT) techniques for the application of test patterns from the external boundary of a NoC. The authors have implemented a parametrisable switching network and developed a set of DfT structures to support testing of network switches using external test configurations. The proposed structures include resource loopback for testing the crossbar multiplexer of the resource connection, a modification to the control part to force YX routing and a compact logic built-in self test (BIST) for the control unit. Experiments show that the proposed structures allow near-100% test coverage at the expense of less than 4% of extra switch area.

Design-for-testability-based external test and diagnosis of mesh-like network-on-a-chips

The focus of the paper is detection of faults in NoC routers by combining concurrent checkers with embedded on-line test to enable cost-effective trade-offs between area-overhead and test coverage. First, we propose a framework of tools for formally evaluating the quality of the checkers and for optimizing the overhead area with given fault coverage constraints. The stress is in particular on the minimization of the error detection latency, which is a crucial aspect in order to eliminate (or limit) error propagation. Second, the concurrent checkers will be complemented by embedded on-line test packets which are to be applied as a periodic routine during the idle periods in router operation. The framework together with the corresponding methodology has been successfully applied to a realistic case-study of a fault tolerant NoC router design. The case study shows that combining concurrent routers with embedded test allows reducing the area overhead of the checkers from 31--35% down to 1.5--10% without sacrificing the fault coverage.

A Framework for Combining Concurrent Checking and On-Line Embedded Test for Low-Latency Fault Detection in NoC Routers

Current paper proposes a Built-In Self-Test (BIST) architecture for targeting the routing infrastructure of mesh-like NoCs from their boundaries. The architecture contains a counter and a Finite State Machine (FSM) implementing the test configurations. Test data is generated and test responses compacted by a dedicated hardware structure requiring very little silicon area. The advantages of this new boundary BIST concept with respect to existing methods is that costly data wrappers in the NoC network are unnecessary, and thus, area and performance penalties are avoided. We have also improved previously developed test configurations. Experiments show that up to two orders of magnitude gains in the speed of testing are achieved using the new method for large NoCs.

Vineeth Govind

Papers

Test Configurations for Diagnosing Faulty Links in NoC Switches

An External Test Approach for Network-on-a-Chip Switches

Design-for-testability-based external test and diagnosis of mesh-like network-on-a-chips

A Framework for Combining Concurrent Checking and On-Line Embedded Test for Low-Latency Fault Detection in NoC Routers

Low-area boundary BIST architecture for mesh-like network-on-chip