Department of Computer Science Mixed Criticality Support on Networks-on-Chip Leandro Soares Indrusiak http://www-users.cs.york.ac.uk/lsi Dagstuhl Seminar 15121 – March 2015
Mixed Criticality Support on NoCs | L. S. Indrusiak Many-Core Systems Many-core systems present a shift towards communication-centric design abundant computation resources shared communication media Inter-core communication architectures point-to-point on-chip bus network-on-chip 2
Mixed Criticality Support on NoCs | L. S. Indrusiak Networks-on-Chip Communication infrastructure based on links and routers that interconnect cores providing packet-based data transfer R R R Low capacitive load and short wires Scalable throughput C C C Point to point connectivity R R R Communication parallelism Link C C C Shared media R R R • Reusability Router C C C Core 3
Mixed Criticality Support on NoCs | L. S. Indrusiak NoC parallelism and scalability CPU I/O CPU CPU Multiple connections simultaneously RAM CPU CPU CPU 4
Mixed Criticality Support on NoCs | L. S. Indrusiak NoC communication interference CPU I/O CPU CPU link contention leads to latency variability RAM CPU CPU CPU 5
Mixed Criticality Support on NoCs | L. S. Indrusiak Time predictability and isolation Full traffic separation (i.e. no packet blocking) deterministic routing, fully disjoint routes (e.g. Hermes) multiple overlay networks (e.g. Tilera), contention over NIs and memory still possible circuit switching (e.g. PNoC), unpredictable circuit setup time Virtual traffic separation fixed TDM traffic slotting (e.g. Aethereal, AElite) rate controlling (e.g. Nostrum, IDAMC) priority-arbitrated virtual channels (e.g. QNoC) 6
Mixed Criticality Support on NoCs | L. S. Indrusiak Priority preemptive virtual channels highest priority highest priority priority ID with remaining credit with remaining credit R R R data_out data_in C C C … … routing routing credit_out credit_in R & & R R transmission transmission control control C C C R R R … … C C C 7
Mixed Criticality Support on NoCs | L. S. Indrusiak Packet latency Packet flows suffer interference from other flows that have higher priority and share at least one link indirect interference also plays a role Worst case latency can be found by an application of Response Time Analysis Z. Shi, A. Burns: Real-Time Communication Analysis for On-Chip Networks with Wormhole Switching. NOCS 2008: 161-170 8
Mixed Criticality Support on NoCs | L. S. Indrusiak Mixed criticality packet flows Possible sources of uncertainty R R R packet length packet flow period C C C jitter R R R C C C All packet flows must be R R R schedulable under normal mode C C C Runtime monitoring detects when packets go “beyond network interface normal” 9
Mixed Criticality Support on NoCs | L. S. Indrusiak Mixed criticality packet flows Runtime monitoring detects when packets go “beyond R R R normal” C C C R R R if it is a LO-CRIT packet exceeding its C C C normal budget, reject it R R R if it is a HI-CRIT packet exceeding its normal budget, signalise a mode C C C change to the NoC, aiming to notify that a service degradation to LO-CRIT mode change notification packets is needed so that HI-CRIT packets can still be scheduled despite of potential increase of interference due to overbudget packets 10
Mixed Criticality Support on NoCs | L. S. Indrusiak Mixed criticality packet flows Two mode change R R R propagation protocols C C C WPMC: mode change flag R R R “piggybacked” on packets that pass through a router that has C C C changed mode R R R C C C WPMC-FLOOD: mode change is flooded to the entire NoC A. Burns, J. Harbin, L. S. Indrusiak: A Wormhole NoC Protocol for Mixed Criticality Systems. RTSS 2014: 184-195 11
Mixed Criticality Support on NoCs | L. S. Indrusiak Mixed criticality packet flows Two mode change R R R propagation protocols C C C WPMC: mode change flag R R R “piggybacked” on packets that pass through a router that has changed C C C mode R R R C C C WPMC-FLOOD: mode change is flooded to the entire NoC L. S. Indrusiak , J. Harbin, A. Burns: Average and Worst-Case Latency Improvements in Mixed-Criticality Wormhole Networks-on-Chip. ECRTS 2015 (submitted). 12
Mixed Criticality Support on NoCs | L. S. Indrusiak Mixed criticality packet flows Two HI-CRIT mode arbitration R R R schemes C C C routers that change mode ignore R R R arbitration requests of LO-CRIT packets C C C R R R routers that change mode arbitrate C C C links in criticality order (HI-CRIT then LO-CRIT), and in priority order within the same criticality A. Burns, J. Harbin, L. S. Indrusiak: A Wormhole NoC Protocol for Mixed Criticality Systems. RTSS 2014: 184-195 13
Mixed Criticality Support on NoCs | L. S. Indrusiak Mixed criticality packet flows Two HI-CRIT mode arbitration R R R schemes C C C routers that change mode ignore R R R arbitration requests of LO-CRIT packets C C C R R R routers that change mode arbitrate C C C links in criticality order (HI-CRIT then LO-CRIT), and in priority order within the same criticality L. S. Indrusiak , J. Harbin, A. Burns: Average and Worst-Case Latency Improvements in Mixed-Criticality Wormhole Networks-on-Chip. ECRTS 2015 (submitted). 14
Mixed Criticality Support on NoCs | L. S. Indrusiak Mixed criticality packet flows Response Time Analysis formulations for each of the protocols were developed Evaluation with synthetic flowsets (against no criticality awareness and criticality-monotonic arbitration) and cycle-accurate NoC simulation WPMC-FLOOD slightly better in general, significantly better in stress scenarios Less restrictive arbitration allows LO-CRIT packets to flow when there are no HI-CRIT packets or when they are blocked due to interferences 15
Mixed Criticality Support on NoCs | L. S. Indrusiak Mixed criticality packet flows 16
Mixed Criticality Support on NoCs | L. S. Indrusiak Mixed criticality packet flows 17
Mixed Criticality Support on NoCs | L. S. Indrusiak Open issues Handle recovery how to detect that there are no further overbudget packets in the network? how to make sure their impact on the network (i.e. additional interference) is no longer there? how to notify all routers to return to normal mode? Explore optimisations on task allocation and packet routing Improved experimental work how many packet flows are HI-CRIT and how many are LO-CRIT? how much overbudget can HI-CRIT reasonably be? 18
Department of Computer Science Mixed Criticality Support on Networks-on-Chip Leandro Soares Indrusiak Alan Burns James Harbin Dagstuhl Seminar 15121 – March 2015
Recommend
More recommend