Latency Optimization for Coded Computation Straggled by Wireless Transmission

Hierarchical Coded Matrix Multiplication in Heterogeneous Multihop Networks

“…1: procedure 2: For each node in worker layer, solve equation (19) or (21) according to whether µ t L,i is equal to µ c i , obtain parameter λ L,i for worker node N L,i and parameter s L−1,j for group G L−1,j ; 3: For each node in relay layer, solve equation ( 28), obtain parameter λ l,i for relay node N l,i and parameter s l−1,j for group G l−1,j layer by layer; 4: For master node, obtain parameter s 0,1 ; 5: Obtain…”

“…For the task allocation strategy given by algorithm 1, we know that λ L,i is a constant from ( 19) and (21). Then, we have s L−1,j = Θ (K L−1,j ).…”

“…When applying coding to wireless distributed network, the impact of transmission latency was further studied in [18]- [21]. The authors of [18] proposed a unified coding framework for distributed computing with straggling servers by introducing a tradeoff between latency of computation and load of communication.…”

“…In coded edge computing systems, a subtask assignment method based on edge node selection was proposed to mitigate both straggling computing nodes and straggling receivers [20]. Assuming different computation and transmission capacities of worker nodes, the authors of [21] extended HCMM to a wireless coded computation algorithm to deal with both computation and transmission stragglers in a heterogeneous wireless computing network.…”

Reliable Distributed Computing for Metaverse: A Hierarchical Game-Theoretic Approach

“…In [12], the authors extend the problem of distributed matrix multiplication in [11] to heterogeneous environments, and propose a coding framework for speeding up distributed matrix multiplication in heterogeneous clusters with straggling servers. In [13], the authors consider that the performance of wireless distributed computing networks is straggled by not only local computation but also wireless transmission. In [14], a coded computing framework for federated learning is proposed, which uses structured coding redundancy to mitigate the straggler effect and speed up the training procedure.…”