Dependable Systems | Amir Taherin

Goal: computing systems that keep their promises under constraints — the research line where I started, and the lens my edge-AI work inherits.

Supercomputer failure analysis

Multi-year failure and repair logs from the Tsubame supercomputers, characterizing how GPU, node, and interconnect failures evolve across generations of multi-GPU compute nodes and how large systems actually recover. Published at DSN 2021 (Taherin et al., 2021).

Reliability-aware energy management

Energy management for mixed-criticality embedded systems where saving power must never compromise a safety guarantee: exploiting slack and controlled service-level degradation of low-criticality tasks while preserving reliability for critical ones (Taherin et al., 2018) (Taherin et al., 2015).

References

2021

DSN

Examining Failures and Repairs on Supercomputers with Multi-GPU Compute Nodes

Amir Taherin, Tirthak Patel, Giorgis Georgakoudis, and 2 more authors

In 51st Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN), 2021

DOI Bib PDF Code

@inproceedings{taherin2021examining,
  title = {Examining Failures and Repairs on Supercomputers with Multi-GPU Compute Nodes},
  author = {Taherin, Amir and Patel, Tirthak and Georgakoudis, Giorgis and Laguna, Ignacio and Tiwari, Devesh},
  booktitle = {51st Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN)},
  year = {2021},
  doi = {10.1109/DSN48987.2021.00043},
}

2018

TSUSC

Reliability-Aware Energy Management in Mixed-Criticality Systems

Amir Taherin, Mohammad Salehi, and Alireza Ejlali

IEEE Transactions on Sustainable Computing, 2018

DOI Bib

@article{taherin2018reliability,
  title = {Reliability-Aware Energy Management in Mixed-Criticality Systems},
  author = {Taherin, Amir and Salehi, Mohammad and Ejlali, Alireza},
  journal = {IEEE Transactions on Sustainable Computing},
  year = {2018},
  doi = {10.1109/TSUSC.2018.2801123}
}

2015

RTEST

Stretch: Exploiting Service Level Degradation for Energy Management in Mixed-Criticality Systems

Amir Taherin, Mohammad Salehi, and Alireza Ejlali

In CSI Symposium on Real-Time and Embedded Systems and Technologies (RTEST), 2015

DOI Bib

@inproceedings{taherin2015stretch,
  title = {Stretch: Exploiting Service Level Degradation for Energy Management in Mixed-Criticality Systems},
  author = {Taherin, Amir and Salehi, Mohammad and Ejlali, Alireza},
  booktitle = {CSI Symposium on Real-Time and Embedded Systems and Technologies (RTEST)},
  year = {2015},
  doi = {10.1109/RTEST.2015.7369846}
}