Currently I’m acting as a PI in a research institute. I’m still baffled by the amount of people we hire that have a postdoc in advanced computational topics and still struggle to do the most basic things such as sshing into a computational cluster to run their calculations.
I hired you because you’re more expert than me on the topic, and I always end up studying what you’re supposedly expert about so that I can explain you what to do…
because they have never done it before, or to a very limited degree. why would you assume anyone knows how to ssh from a terminal? that’s a very specific skillset. nobody is learning that at school.
using linux/unix has nothing to do with computer science especially today, vs 20 years ago.
maybe you use ssh daily your entire life, but that’s not how other people do things. my research org has moved off linux/unix platforms to the point that our researchers only use ssh for legacy/archival projects. nobody is using it for anything after 2014. yeah, 10-20 years ago it was a ssh was a daily necessity, now it’s like once a month or less.
all of our new stuff is web based. you login to a website and you access resources through the web, that way everyone can work remote and the infrastructure for those at home and in office and across the country, is unified and simplified and easy to access.
Sure that’s great and all, but I’d imagine university software programs frequently still require ssh in the intro courses. In 2018 for our intro to programming the requirement was for the code to compile and work on the Uni cluster. Sure you can make a web infrastructure for everything, but teaching ML, Simulation science and computer science students some high performance computing and IT obviously still makes sense in 2026.
Ssh is at the very start of all of this stuff so …
SSH is still a daily tool of sysadmins and security people. It’s just maybe not a tool for the pure machine learning people with their Jupiter notebooks who have only ever used cloud or organization resources managed by somebody else. The people managing these things and building these clusters almost certainly still use SSH and other basic Unix tools as all modern large scale machine learning like those used to train ChatGPT or DeepSeek still happens on Linux systems. Although to be honest SSH works equally well from Mac and Windows these days, and has always worked from FreeBSD. Whatever platform you choose has SSH capabilities.
Currently I’m acting as a PI in a research institute. I’m still baffled by the amount of people we hire that have a postdoc in advanced computational topics and still struggle to do the most basic things such as sshing into a computational cluster to run their calculations.
I hired you because you’re more expert than me on the topic, and I always end up studying what you’re supposedly expert about so that I can explain you what to do…
because they have never done it before, or to a very limited degree. why would you assume anyone knows how to ssh from a terminal? that’s a very specific skillset. nobody is learning that at school.
using linux/unix has nothing to do with computer science especially today, vs 20 years ago.
maybe you use ssh daily your entire life, but that’s not how other people do things. my research org has moved off linux/unix platforms to the point that our researchers only use ssh for legacy/archival projects. nobody is using it for anything after 2014. yeah, 10-20 years ago it was a ssh was a daily necessity, now it’s like once a month or less.
all of our new stuff is web based. you login to a website and you access resources through the web, that way everyone can work remote and the infrastructure for those at home and in office and across the country, is unified and simplified and easy to access.
We are talking about people who daily run calculations with tens of GPUs that have to run for weeks.
The only way to run those calculations is in a computing cluster. Knowing how to use Linux and ssh is the most basic requirement for the job.
I’d assume someone who has a postdoc in the field would know how to connect to a server.
Sure that’s great and all, but I’d imagine university software programs frequently still require ssh in the intro courses. In 2018 for our intro to programming the requirement was for the code to compile and work on the Uni cluster. Sure you can make a web infrastructure for everything, but teaching ML, Simulation science and computer science students some high performance computing and IT obviously still makes sense in 2026.
Ssh is at the very start of all of this stuff so …
SSH is still a daily tool of sysadmins and security people. It’s just maybe not a tool for the pure machine learning people with their Jupiter notebooks who have only ever used cloud or organization resources managed by somebody else. The people managing these things and building these clusters almost certainly still use SSH and other basic Unix tools as all modern large scale machine learning like those used to train ChatGPT or DeepSeek still happens on Linux systems. Although to be honest SSH works equally well from Mac and Windows these days, and has always worked from FreeBSD. Whatever platform you choose has SSH capabilities.