1. ECE 8813a (1) Non-minimal Routing Non-minimal routing  Wormhole degrades performance while VCT has less secondary effects  Fault tolerance is the main motivator Classes  Search-based algorithms  Virtual channel-based routing  Turn-based routing Non-Minimal Routing

2. ECE 8813a (2) Reading Section 4.7 and/or P.T. Gaughan, et al., “Distributed, deadlock-free routing in faulty, pipelined, direct interconnection networks,” IEEE Transactions on Computers, vol. 45, no. 6, pp.651-665, June 1996 A. Mejia, J. Flich, J. Duato, Sven-Arne Reinomo and Tor Skeie, “Segment Based Routing: An Efficient Fault-Tolerant Routing Algorithm for Meshes and Tori,” Proceedings of the International Parallel and Distributed Processing Symposium, April 2006 From J. Flich, A. Mejia, P. Lopez, and J. Duato, “Region-Based Routing: An Efficient Routing Mechanism to Tackle Unreliable Hardware in Networks on Chip,” Proceedings of the First International Symposium on Networks on Chip, May 2007

3. ECE 8813a (3) Backtracking Protocols Backtracking search + resource reservation Constrain the search  Minimal paths vs. #misroutes Non-Minimal Routing P.T. Gaughan, et al., “Distributed, deadlock-free routing in faulty, pipelined, direct interconnection networks,” IEEE Transactions on Computers, vol. 45, no. 6, pp.651-665, June 1996

4. ECE 8813a (4) Optimization Sensitive to choice of switching technique  Naturally suited to circuit switching and pipelined circuit switching  Overhead is large with SAF Deadlock is avoided by not blocking on busy channels Livelock is avoided by maintaining and using search history  In the header: large headers  In the routers: local state, headers comparable to e-cube Protocol variations  Multi-links  k-family  exhaustive: profitable and misrouting  limited misrouting  multi-phase Non-Minimal Routing

5. ECE 8813a (5) Topology Agnostic Routing Topology dependent vs. topology agnostic routing  Reliability  Increasingly important on-chip Approaches  Techniques based on virtual channels oExpensive on-chip oCompetes with QoS schemes  Techniques based on Turn restrictions oDifficult to ensure non-minimal paths Topology Agnostic Routing

6. ECE 8813a (6) Segment Based Routing Topology agnostic routing Restriction-based approach  Multiple restriction options oSelect restrictions based on performance goals  Source based routing oRouting table generation From A. Mejia, J. Flich, J. Duato, Sven-Arne Reinomo and Tor Skeie, “Segment Based Routing: An Efficient Fault-Tolerant Routing Algorithm for Meshes and Tori,” Proceedings of the International Parallel and Distributed Processing Symposium, April 2006. Topology Agnostic Routing

7. ECE 8813a (7) Key Idea: Segments & Subnets Partition topology into subnets and then segments in a subnet Goal: islands of regularity Topology Agnostic Routing

8. ECE 8813a (8) Key Idea: Optimization Placement of Turn restrictions in a segment  Placement for latency  shortest path  Placement for throughput  distribute traffic Topology segmentation  Can be optimized for regular topologies Topology Agnostic Routing

9. ECE 8813a (9) Requirements Avoid deadlock in a segment Avoid deadlock when traversing multiple segments Ensure routing connectivity when physical connectivity exists Avoiding congestion in path construction Topology Agnostic Routing

10. ECE 8813a (10) Construction of Segments Search for starting segment + “regular” segments  Unitary segments Add one bidirectional restriction in each segment Starting node Terminal node Unitary segment bridge segment Topology Agnostic Routing Failed links

11. ECE 8813a (11) Segment Types Starting segment Regular segment Unitary segment Starting node Terminal node Unitary segment bridge segment Topology Agnostic Routing Failed links

12. ECE 8813a (12) Deadlock Freedom One routing restriction per segment  No cycles in a segment Every cycle contains a segment  Hence cannot be “closed” to create deadlock No cycle from the start node back to itself  Cannot create cycles across subnets Think of a subnet as a union of 1-D segments Topology Agnostic Routing

13. ECE 8813a (13) Example From A. Mejia, J. Flich, J. Duato, “On The Potential of Segment Based Routing” Proceedings of the International Conference on Parallel Processing 2008

14. ECE 8813a (14) Routing From A. Mejia, J. Flich, J. Duato, “On The Potential of Segment Based Routing” Proceedings of the International Conference on Parallel Processing 2008 Segment routing is turn based and therefore partially adaptive Source routing can be layered on top of segments to balance traffic

15. ECE 8813a (15) Performance From A. Mejia, J. Flich, J. Duato, Sven-Arne Reinomo and Tor Skeie, “Segment Based Routing: An Efficient Fault-Tolerant Routing Algorithm for Meshes and Tori,” Proceedings of the International Parallel and Distributed Processing Symposium, April 2006. Topology Agnostic Routing

16. ECE 8813a (16) Region-Based Routing Recognize that routing decisions implicitly check for region membership  Think meshes Generalize the idea of regions  Can naturally be adapted for fault tolerant routing S D From J. Flich, A. Mejia, P. Lopez, and J. Duato, “Region-Based Routing: An Efficient Routing Mechanism to Tackle Unreliable Hardware in Networks on Chip,” Proceedings of the First International Symposium on Networks on Chip, May 2007 Topology Agnostic Routing

17. ECE 8813a (17) Example of Regions Static, off-line topology characterization Online querying of network structure  Built on segment-based routin { node set } From J. Flich, A. Mejia, P. Lopez, and J. Duato, “Region-Based Routing: An Efficient Routing Mechanism to Tackle Unreliable Hardware in Networks on Chip,” Proceedings of the First International Symposium on Networks on Chip, May 2007 Topology Agnostic Routing

18. ECE 8813a (18) Example of Regions What are the characteristics of these regions?  Note #regions = f(routing options) Note use of output port depends on input port  Check W output port from N input port { node set } From J. Flich, A. Mejia, P. Lopez, and J. Duato, “Region-Based Routing: An Efficient Routing Mechanism to Tackle Unreliable Hardware in Networks on Chip,” Proceedings of the First International Symposium on Networks on Chip, May 2007 Topology Agnostic Routing

19. ECE 8813a (19) Approach Observe that table-based routing is really region based  Each entry identifies a region Merge entries into compact region specifications at each switch Region construction is based on the paths  Any set of paths  fault tolerant routing  No virtual channels Topology Agnostic Routing

20. ECE 8813a (20) Key Idea Generate paths  All minimal  First non-minimal  Note: using SR routing Record paths at each router  Produce region representation for each output port  Record input port dependencies Program Routers From J. Flich, A. Mejia, P. Lopez, and J. Duato, “Region-Based Routing: An Efficient Routing Mechanism to Tackle Unreliable Hardware in Networks on Chip,” Proceedings of the First International Symposium on Networks on Chip, May 2007 Topology Agnostic Routing

21. ECE 8813a (21) Creating Regions Coalesce routing options based on inputs and outputs Represents a compact routing table From J. Flich, A. Mejia, P. Lopez, and J. Duato, “Region-Based Routing: An Efficient Routing Mechanism to Tackle Unreliable Hardware in Networks on Chip,” Proceedings of the First International Symposium on Networks on Chip, May 2007 Topology Agnostic Routing

22. ECE 8813a (22) Region Construction From J. Flich, A. Mejia, P. Lopez, and J. Duato, “Region-Based Routing: An Efficient Routing Mechanism to Tackle Unreliable Hardware in Networks on Chip,” Proceedings of the First International Symposium on Networks on Chip, May 2007 Topology Agnostic Routing @each node Segment Routing SearchCoalesce & Packing Region Formation Note this can be applied to any topology No virtual channels Offline optimization of latency vs. distance

23. ECE 8813a (23) Hardware Overheads Each region requires  Four registers that define the region  Mask registers to define input and output ports  Logic to determine routing options Hardware cost grows as the number of regions  Growth as f(network_size) is much slower From J. Flich, A. Mejia, P. Lopez, and J. Duato, “Region-Based Routing: An Efficient Routing Mechanism to Tackle Unreliable Hardware in Networks on Chip,” Proceedings of the First International Symposium on Networks on Chip, May 2007 Topology Agnostic Routing

24. ECE 8813a (24) Implementation Initialization of region registers and parallel evaluation of all regions From J. Flich, A. Mejia, P. Lopez, and J. Duato, “Region-Based Routing: An Efficient Routing Mechanism to Tackle Unreliable Hardware in Networks on Chip,” Proceedings of the First International Symposium on Networks on Chip, May 2007 Topology Agnostic Routing

25. ECE 8813a (25) Microarchitecture Issues Routing algorithm performance is sensitive to resource allocation schemes in the router Key resource management functions include  Routing function  Selection function  Arbitration/scheduling Mismatch can lead to poor performance

26. ECE 8813a (26) Resource Allocation: Selection Functions Selection function may be oblivious or informed  Common to favor minimal paths and lightly loaded links Examples:  Meshes: minimum congestion, maximum flexibility, straight lines Unlike routing functions, selection function must be serialized  Result updates the channel status – a centralized resource VC status/control VC buffer Selection function Routing function Input VC Output VCs

27. ECE 8813a (27) Selection Functions Favor adaptive channels  Improve probability of escape channel availability  Time dependent selection functions: give adaptivity a chance Selection functions for real time traffic  Separate best effort and guaranteed packets via VCs or virtual networks Note the impact on bisection utilization Selection functions for cache coherent systems?

28. ECE 8813a (28) Resource Allocation: Arbitration Tradeoffs: channel bandwidth vs. message sizes and types  Mix of buffering strategies across message types All three strategies must be co-designed for a tuned system arbitration Flow controlMessage size

29. ECE 8813a (29) Routing, Selection & Arbitration Input driven vs. output driven scheduling  Output driven scheduling requires replication of routers amongst inputs Lessons from the microprocessor world  Impact of complexity, workloads, and concurrency Impact of….  Symmetry of the topology  Locality of traffic  Packet size Locality, uniformity Irregular, hot spot deterministic routing adaptive routing

30. ECE 8813a (30) Characterization of Techniques Deadlock freedom achieved by  Path based techniques oRestrict paths  Buffer based techniques oStructured buffer pools  Channel based techniques o#VCs independent of the network

31. ECE 8813a (31) Summary Best routing algorithm driven by multiple considerations Deterministic vs. adaptive Uniform vs. non-uniform traffic Packet sizes Power envelope On-chip vs. off-chip Locality of traffic Hot spots Compatible micro-architecture Symmetric vs. asymmetric topology