From both a hardware and software perspective those are very different types of parallelism that Nvidia's architects and the architects of its predecessors at Sun/SGI/Cray/elsewhere were intimately familiar with. See: https://en.wikipedia.org/wiki/Flynn%27s_taxonomy
From the outside I dont perceive the promotion dynamics for engineers working on old services is that different at AWS than Google or Azure. Finance and engineering do conspire to kill off things despite customer annoyance, perhaps as they should.
There is also a constant churn of retiring EC2 instance types, security certs, old database versions, EMR versions, OS versions, security practices, cost optimizations, etc to suck up your time with minimal business value. This is not different fundamentally than on prem, you just have slightly less control and an outside party forcing your hand to do the right thing isnt purely a bad thing...
AWS is pretty good about giving you a grace period to migrate off and informal warnings if they strategically want you to move to a different service of theirs before stuff gets killed if you are looped in with a TAM (e.g Data Pipelines vs more-expensive Glue). They seem to have recently migrated to a strategy where they disable services for new customers but actually dont kill them completely off but keep them on life support.
The only product I know well in your list is pre-VPC ec2 instances, which (to be fair) are a terrible product : it is cross tenant (as in : you can impact other customers). Good riddance.
I think many AWS services are well designed : isolated software-only components. They build a lot on top of a very stable infrastructure (VPC / s3 / ec2 / IAM), which means supporting a service is really cheap : they are just a couple of containers running somewhere
i thought Hezbollah had their own broadcast pager network for their pagers for security reasons. a trigger sent out on that frequency would only affect Hezbollah, no?
We don't know how exactly how the trigger mechanism worked. It was an import of pager batteries that were intercepted and replaced, it's plausible that someone else using a pager in the country would be affected if they had received a pager battery from the same batch
Don't store database credentials at all. Ensure your product and recommended database configuration supports SSO/SAML/etc with credentials managed through Okta or Active Directory. You'll need that if you go up-market into an enterprise.
You can go to the Youtube page for the video:
https://www.youtube.com/watch?v=si9iqF5uTFk ( or https://www.youtube.com/watch?v=AW7ZHpKuqZg for part 2)
click on the "...more" link in the video description,
then click on the "Show Transcript" button further down that appears and you will see a transcript in the upper right of your screen which you can scroll through and read.
A bunch of threads in parallel implies MIMD parallelism- multiple instructions multiple data.
A warp implies SIMD parallelism - single instruction multiple data (although technically SIMT, single instructions multiple threads https://en.wikipedia.org/wiki/Single_instruction,_multiple_t...).
From both a hardware and software perspective those are very different types of parallelism that Nvidia's architects and the architects of its predecessors at Sun/SGI/Cray/elsewhere were intimately familiar with. See: https://en.wikipedia.org/wiki/Flynn%27s_taxonomy
reply