News for hackers. Stuff that matters.
Re:Ask HN: Who operates at scale without containers? by bogomipz on Tuesday March 22, 2022 @01:13PM (#30768755)
Can you elaborate on what makes containerized CUDA so difficult?
Re:Ask HN: Who operates at scale without containers? by jcadam on Tuesday March 22, 2022 @09:04PM (#30773643)
My last project where we needed CUDA, we ended running our GPU workloads directly on a dedicated machine loaded up with GPUs. The rest of the system was containerized. Though, we were just a small team of software engineers working on an R&D project - we didn't have any k8s wizards on our team.
Re:Ask HN: Who operates at scale without containers? by Melatonic on Tuesday March 22, 2022 @01:29PM (#30768949)
I would guess the fact that its a highly proprietary design that is designed to specifically scale to massive amounts of parallel threads on specific physical hardware devices
Re:Ask HN: Who operates at scale without containers? by nancarrow on Tuesday March 22, 2022 @02:21PM (#30769551)
the CUDA library versions are very coupled to the hosts' CUDA driver versions and the container system itself (Docker) needs special code to link the two
Re:Ask HN: Who operates at scale without containers? by bogomipz on Wednesday March 23, 2022 @08:08AM (#30777595)
Can't you just do a host mount/bind mount of those specific libs though?