But I find it hard to believe that HPC in the cloud is cheaper than HPC on bare metal, assuming high load most of the time. For computing weather models, that should he the case.
At peak capacity, bare metal pays off soon. Even when they run at near 100% of CPUs, weather models never run 24/7 but only fraction of the time.
Examples: A national weather service (any country) may run at 20-50% capacity. In a research setting, 5-10% is more common (personal experience). Consulting businesses are somewhere in between. The lower the fraction of capacity used, the longer the time after which bare metal investment pays off.
For businesses that need to do a one-off feasibility study or to work on a few projects/events, upfront investment of > $100K for bare metal is prohibitive. And this is before hiring a sysadmin and a weather modeler to run the thing.
I appreciate your comment, it motivates me to take some time soon to write up the unit economics in more detail.
Ive been reading some books about innovations and it often seems like this. A lot of random encounters (is word serendepity correct?) coming together to create new idea.
Now if only there was more options for this random encounters on the Internet. For me only HN and reading nonfiction helps in this
On one hand it is dismissive to say that any innovation occurred due to the laws of physics and the starting condition of the universe at the Big Bang. On the other hand it makes your insight that most innovations were chance happenings less surprising.
When I applied to a major university at age 17, I put down two potential majors. One was Geology.
I didn't attend that university. I went for two years to a local college and didn't bother to declare a major the first year. The second year, I declared as a History major because I was taking a lot of history classes and my counselor was in the history department. It let me keep the same counselor.
I dropped out in part because I knew two people with substantial amounts of college who were delivering newspapers. One was living with his mother. One was financially dependent upon his wife.
I figured I could deliver newspapers for a living without a pile of student loans and my quality of life would be higher without the student loans. I decided I would return to college if I ever had any idea what I wanted to do with a degree.
I eventually decided I wanted to be an urban planner, which was a good reason to go back to school. Most urban planning programs are Masters degrees. You typically have a Bachelor's in something else.
I decided to pursue a Bachelor's in Environmental Resource Management as background. I felt I needed to understand the underlying natural environment in order to do good things with the built environment.
Most Environmental Studies programs began in Geology departments. It's possible that I might have ended up sort of where I am currently even without the long, winding detour in between.
Very very similar to my path. Graduated with a geography degree. Got into a planning job but left after 6 months because I didn't want to spend my life arguing about people's fence heights. Ended up becoming a CPA instead.
10 comments
[ 4.2 ms ] story [ 20.9 ms ] threadBut I find it hard to believe that HPC in the cloud is cheaper than HPC on bare metal, assuming high load most of the time. For computing weather models, that should he the case.
Examples: A national weather service (any country) may run at 20-50% capacity. In a research setting, 5-10% is more common (personal experience). Consulting businesses are somewhere in between. The lower the fraction of capacity used, the longer the time after which bare metal investment pays off.
For businesses that need to do a one-off feasibility study or to work on a few projects/events, upfront investment of > $100K for bare metal is prohibitive. And this is before hiring a sysadmin and a weather modeler to run the thing.
I appreciate your comment, it motivates me to take some time soon to write up the unit economics in more detail.
If the rule you followed brought you to this - of what use was the rule?
https://youtube.com/watch?v=_fv7XSrrI9M
Now if only there was more options for this random encounters on the Internet. For me only HN and reading nonfiction helps in this
Huh. Google Cloud Platform has a product with the same name now.
https://cloud.google.com/run/
I didn't attend that university. I went for two years to a local college and didn't bother to declare a major the first year. The second year, I declared as a History major because I was taking a lot of history classes and my counselor was in the history department. It let me keep the same counselor.
I dropped out in part because I knew two people with substantial amounts of college who were delivering newspapers. One was living with his mother. One was financially dependent upon his wife.
I figured I could deliver newspapers for a living without a pile of student loans and my quality of life would be higher without the student loans. I decided I would return to college if I ever had any idea what I wanted to do with a degree.
I eventually decided I wanted to be an urban planner, which was a good reason to go back to school. Most urban planning programs are Masters degrees. You typically have a Bachelor's in something else.
I decided to pursue a Bachelor's in Environmental Resource Management as background. I felt I needed to understand the underlying natural environment in order to do good things with the built environment.
Most Environmental Studies programs began in Geology departments. It's possible that I might have ended up sort of where I am currently even without the long, winding detour in between.