3 ms·

Yeah, I only posted two links from my notes, from when I was looking at this a few months ago. Here's one on MIG. https://arxiv.org/abs/2207.11428 https://arxi

Yeah, I only posted two links from my notes, from when I was looking at this a few months ago. Here's one on MIG. https://arxiv.org/abs/2207.11428 https://arxiv.org/abs/2207.11428

stygiansonic 1y ago
That paper doesn’t seem to be about security vulnerabilities in MiG but rather using it to improve workload efficiency
throw0101d 1y ago
As per sibling comment, this is about utilization efficiency and not breaking isolation (between MIG instances). The conclusion: > In this paper, we presented MISO, a technique to leverage the MIG functionality on NVIDIA A100 GPUs to dynamically partition GPU resources among co-located jobs. MISO deploys a learning-based method to quickly find the optimal MIG partition for a given job mix running in MPS. MISO is evaluated using a variety of deep learning workloads and achieves an average job completion time that is lower than the unpartitioned GPU scheme by 49% and is within 10% of the Oracle technique.