,
9 tweets,
3 min read
Read on Twitter
Excellent post, but breakthrough does feel somewhat ahead of schedule. I'm more anticipating 2024-2029 as the timeframe for serverless to become the new norm in IT -> “2019 could be serverless’ breakthrough year” by @erezberkner jaxenter.com/serverless-lum…
i.e. roughly on the same track as IaaS which I estimated all those years ago to become the new nom around 2016-2021 and it certainly feels that way now. Of course the battle for dominance is decided earlier in 5-8 yrs i.e. IaaS that was 2011 to 2014 and Serverless 2019-2022 ...
i.e. we're at last orders in the last chance saloon at the end of line railway for serverless. Winners are ready to be announced for the future domination of IT. Front runner- Amazon, 2nd tier MSFT then ... well, some are still pondering whether to get out of bed, 500 miles away.
X : Serverless is still new.
Me : 5 yrs old. Ok, it normally takes around 30-50 yrs (some evidence that's accelerated to 20-30 yrs) to evolve to point of commodity, 10-15 years to become new norm and in the first 5-8 years the winners are decided. Best explained with a map.
Me : 5 yrs old. Ok, it normally takes around 30-50 yrs (some evidence that's accelerated to 20-30 yrs) to evolve to point of commodity, 10-15 years to become new norm and in the first 5-8 years the winners are decided. Best explained with a map.
X : How do containers fit into that picture?
Me : They are increasingly more invisible subsystems if used at all. Eventually Lambda will be written on chip.
Me : They are increasingly more invisible subsystems if used at all. Eventually Lambda will be written on chip.
X : What about conversational programming then?
Me : UP the stack. You need to focus UP the stack. Repositories of functions, IDE, management tools covering debugging and capital flow, new security tools etc etc. Things below the line become less and less visible over time e.g.
Me : UP the stack. You need to focus UP the stack. Repositories of functions, IDE, management tools covering debugging and capital flow, new security tools etc etc. Things below the line become less and less visible over time e.g.
In the past, if I wanted 100 servers up and running, I'd care about aircon, airflow, network cabling, power, racks. These things still exist when I create 100 instances on EC2, I just don't care. Serverless is an abstraction above this, I don't care about servers, OS etc.
In 2005/06, one of my devs Tom built and released a new form of wiki (using Zimki, an early serverless environment) in less time than it took to get one of the sysadmins to get to the data centre, let alone install a new box, let alone unpack it, order it or wait for delivery.
Of course, I use this as an example. We had virtualised our data centre sometime around 2003-2005. It was all the rage. We moved on. But serverless isn't even about costs or speed. It's the billing per function that really changes the game on how you invest, operate and manage.
