A routing methodology for achieving fault tolerance in direct networks
read more
Citations
Elevator-First: A Deadlock-Free Distributed Routing Algorithm for Vertically Partially Connected 3D-NoCs
Addressing Manufacturing Challenges with Cost-Efficient Fault Tolerant Routing
Understanding the interconnection network of SpiNNaker
Practical Deadlock-Free Fault-Tolerant Routing in Meshes Based on the Planar Network Fault Model
Cost-Efficient On-Chip Routing Implementations for CMP and MPSoC Systems
References
Deadlock-Free Message Routing in Multiprocessor Interconnection Networks
Deadlock-free message routing in multiprocessor interconnection networks
Myrinet: a gigabit-per-second local area network
Virtual cut-through: A new computer communication switching technique
Virtual-channel flow control
Related Papers (5)
Frequently Asked Questions (15)
Q2. How many virtual channels are used for a deterministic routing?
Each physical input port is split into several virtual channels (four or five), each providing buffering resources in order to store two packets.
Q3. How many faults were tolerated by I+D?
when using direction order (X þ Y þ Z þX Y Z ) routing instead of dimension order routing, only three faults were tolerated by I+D.
Q4. How many faults are tolerated in a packet basis?
packet basis (D), it is possible to guarantee an acceptablefault tolerance degree (up to five link faults are tolerated ina 3 3 3 torus) without significantly affecting the implementation cost or performance.
Q5. What are the popular multiprocessor routing algorithms?
A large number of fault-tolerant routing algorithms for multiprocessor systems have been proposed, especially for mesh and torus topologies.
Q6. What is the explanation for the failure tolerance of the methodology?
In light of the evaluation results, the authors observe that routing through an intermediate node is the fault-tolerant mechanism most widely used by the methodology.
Q7. What are some of the applications that require a huge amount of processing power?
THERE exist many compute-intensive applications thatrequire a huge amount of processing power (nuclear weapons simulations, protein folding, global climate modeling, galaxy interaction simulations, etc.).
Q8. How is the method used to avoid faulty components?
If faulty components can be encountered when routing packets between a source-destination pair, the methodology avoids these faults by using intermediate nodes for routing.
Q9. What are the two virtual channels used for the adaptive routing?
The first of these is used to implement the escape paths for the adaptive routing, whereas the second one is reserved for high priority packets.
Q10. What is the probability of failure of a massively parallel computer?
The huge number of processors and associated devices (memories, switches, and links, etc.) significantly affects the probability of failure.
Q11. What is the way to tolerate faults in a network?
When using reconfiguration, any number of faults is tolerated without requiring additional resources [38], as long as the network remains connected.
Q12. What are the three ways to tolerate faults in interconnection networks?
there are three ways to tolerate faults in interconnection networks: component redundancy, faulttolerant routing algorithms, and reconfiguration.
Q13. What is the purpose of using intermediate nodes for routing?
This way, intermediate nodes are used in order to obtain greater control over the paths followed by packets, thereby avoiding the faults.
Q14. Why is the performance of The author2+D not included in the results?
16 Furthermore, because most paths (all in this case) can be resolved using two intermediate nodes alone, the performance of The author2+D is very similar to that of The author2 and has not been included in the results.
Q15. How many KB of memory is required to store the routing info?
Notice that the amount of memory required to store the routing info is low, e.g., if the The author2 combination is used in a large system with 65,536 nodes, 256KB of memory would be required.