Microservices were sold as the panacea for better TTM, but it's clear that they are only sometimes the right choice. It also highlights the importance of critically evaluating our choices regarding software development. This shift from time-to-market first to a cost-effective first approach means the next level in adopting distributed systems architectures and a push towards developing better practices and tools for monolithic architectures. It's essential that this is only one of many services, and I'm sure that at Amazon scale, there's a lot where microservices have advantages. And it's more than just opinion pieces - this recent Amazon case study showed that they struggled with the high cost and scalability inefficiencies in that particular service. It also increased our scaling capabilities." Piece by Marcin Kolny.Īre monolithic architectures making a comeback? While microservices have been all the rage in recent years, I've been reading more and more articles that challenge the popular microservices and even the cloud approach. Lately, I read an Amazon Prime Video team case study that states: "Moving our service to a monolith reduced our infrastructure cost by over 90%. That medieval farmer, would not understand that there will be a need for management consultants, flight attendants and software developers just as today we can’t imagine the new needs for software that will be there in 20, 30 or 100 years. Likewise they ignore the concept that richer, more productive humans invent all kinds of new “needs” that they want met. Underlying both these fallacies is the same wrong idea that demand is fixed and so that increased productivity negates the need for workers. “Every worker knows how to read?!! That’s crazy! Why would we train everyone to read? Who is going to pay for that? Why do you need to know how to read to run a farm? Am I going to read Thomas Aquinas to my sheep? Ha, ha, ha. Wow, everyone doing some programming! That sounds crazy right? This is just as crazy sounding as it would be to a medieval farmer if you told them that in 2019 almost every working individual knows how to read and uses it every day in their profession. And that will be fine because we are outputting 100 times as much code. And then there are people that still act more or less like today’s software developers except that they are 100 times more productive. Maybe it’s done though natural language processing and allows people to specify only the creative part that humans do well and let the language or framework do the rest. I mean working with a computer through some kind of interface to put together working programs that solve problems. In fact, it’s not inconceivable that almost every working professional will include some software development as part of their job. If I had to guess, in 20–30 years, there will be 10 times as many software developers. It’s simply a productivity enhancing tool. But this is really the same thing that people are saying today who worry that the software profession will be made obsolete by AI. So, I bet in 2019 there are fewer than 50 software developers in the whole world.” My productivity is easily 100 times higher. Someone plays with it a bit and says “This is fantastic, incredible. You show them Python and SQL and a modern software stack. Imagine going back to 1960 where there probably about 5000 programmers in the world. Can AI replace software engineers in the next 20 years?
0 Comments
Leave a Reply. |