Cloud Architectures: A View from the Top
// Virtualization Management Software & Data Center Control | VMTurbo » VMTurbo Blog
In my daily interactions with IT professionals, I've noticed that conversations around next-generation infrastructure delivery and cloud strategy can really be summed up with one word: uncertainty.
We're in the midst of a major shift in the enterprise technology landscape – the giants of eras past are attempting to transform themselves, while newer technologies seem to generate enormous buzz overnight. To make matters even more complicated, some of those newer technologies, that were supposed to be the standard of the future, disappear as quickly as they emerged.
Vendors, technology stacks, data locality, clouds, different operations strategies; they're all getting more confusing, and our options are only growing. To some extent, this is fantastic for customers. They now have a broader variety of technologies to match their business goals. But where do you begin when you're looking at architecting your next generation infrastructure?
My answer is simple: you start at the top.
What's "Stack Fallacy"?
What really got me thinking about how to answer some of these questions was an interesting article describing the concept of "stack fallacy" that I stumbled upon on twitter. I'd highly recommend checking out the article. If you're looking for a quick summary, though, the essential idea is that technology companies struggle when they try to innovate up the stack, building on top of the technology that is their core competency.
The reason for this is that they have tremendous institutional knowledge regarding the building blocks, and mistakenly consider the development and design of a product that meets customer use cases to be trivial. One example the article gives is Oracle's difficulty building a successful CRM, despite the fact that Salesforce runs on Oracle databases.
What Really Matters When Architecting Your Cloud?
Because trying to build up is so difficult, the design of any environment should focus on what's most critical, and why we build these infrastructures in the first place – the application and its performance.
Looking around for a real world example of this theory, I found a blog post from Nanit, an early-stage startup focusing on building parenting software. They had chosen Docker containers as the core of their infrastructure design, and were evaluating different orchestration solutions. Specifically, they were running their infrastructure in Amazon's EC2, and were comparing ECS, the native Amazon Docker management system, to Kubernetes, an open source project. The blog post outlines the key differentiators between the two tools, and what it ultimately came down to, time and time again, was that Kubernetes was better at serving their application's needs.
For instance, ECS requires each container running inside an instance to use a unique port. Kubernetes, on the other hand, uses service discovery and natively handles the port routing. Because Kubernetes was built to solve the problems of people trying to deliver applications via containers, the solution better fit this specific user's needs.
More Technologies, More Silos…Or Not
It's entirely possible that I could be wrong and some company out there could deliver a full-stack solution that works for everyone. But the logical result of the shift we're seeing is that organizations will end up with a broader variety of technologies specifically designed for their unique portion of the stack.
A lot of these stack specific technologies will contain automation, making each piece of the puzzle easier to solve if you're looking at it in isolation. At the end of the day though, your applications will only be as fast as the slowest component in its stack. So how do you ensure that each one of those pieces is working together? How do you make decisions around application performance as the amount of knobs and levers that need to be turned drastically increases?
Great thing is, we've got you covered.
The post Cloud Architectures: A View from the Top appeared first on Virtualization Management Software & Data Center Control | VMTurbo.
Shared via my feedly reader
Sent from my iPad