… selective replication technique for HPC applications for both fail-stop errors and SDCs. Since
complete replication of applications … a runtime-based technique for selective replication. …

… In this research, our key findings were: First, complete replication of HPC applications is
not required to mitigate the foreseen exascale error rates while achieving the same reliability …

… challenge for high-performance computing (HPC) applications as … able to detect SDC in HPC
applications to a certain level by … far from fully protecting applications to a level comparable …

… applications to mitigate silent and fail-stop errors. To avoid the prohibitive costs of complete
replication, we introduce a lightweight selective replication … and selectively replicate only the …

… In this section, we show how the properties of the application are leveraged by selective
replication in order to identify what to replicate in the application. This analysis consists of three …

… not been widely adopted in HPC environments, due to its dependency on applications [15],
[… for HPC environment [24]. In this work, we study a practical selective replication strategy for …

… selectively replicating tasks based on criticality[22][23][24]. These works replicate tasks from
an application … The idea of criticality is orthogonal to our task of selectively replicating nodes …

… needed for allocating replicated arrays across processors. CON… (rep), replicated buffer with
links (ll), selective privatization (sel)… Write method because iteration replication is very difficult …

… replication vary from 0.1% (0.11% in total, Pingpong) to 7% (15% in total, FFT) showing that
replication … For more discussion and techniques on replication and selective replication, we …

… Since we focus on evaluating selective instruction duplication for application code, we do
not inject instructions within libraries (eg, libc and libm). Note that checking instructions are …