DZone Research Report: A look at our developer audience, their tech stacks, and topics and tools they're exploring.
Getting Started With Large Language Models: A guide for both novices and seasoned practitioners to unlock the power of language models.
Global Leader, API Strategy at MuleSoft, a Salesforce company
Vancouver, CA
Joined Oct 2017
Matt McLarty is MuleSoft's Global Leader for API Strategy. In this role, Matt helps our customers take maximum advantage of their API opportunities through strategic guidance, sharing organizational practices, and the development of API ecosystems. He’s based in Vancouver, BC, Canada. twitter: @MattMcLartyBC
Stats
| Reputation: | 521 |
| Pageviews: | 130.0K |
| Articles: | 2 |
| Comments: | 4 |
API Management
DZone Trend Reports zoom in on the content from our Research Guides that readers have found most useful. The API Management Trend Report features in-depth original research on developer preferences surrounding API programs and integrated systems. Experts in the field also share their perspectives on the importance of developer experience to API strategy, the business value of various API types, and the challenges of API-driven digital transformation.
Comments
Dec 10, 2021 · Arran Glen
Yes, definitely logical. Great questions!
Dec 10, 2021 · Arran Glen
I think it depends on how you define "single microservice"... Depending on platform/paradigm (e.g. Kubernetes vs. Lambda) you might have different runtime components... Depending on volumes you might split or not... Whether you split or not, I agree with your main point: you need to factor non-functionals into your ultimate implementation architecture. That's why the "Qualities" section exists on the canvas. I get into an example in this workshop that shows a service being split in exactly the scenario you describe: https://www.oreilly.com/library/view/oreilly-software-architecture/9781492050506/video324023.html
Dec 15, 2017 · Arran Glen
Exactly. Always a matter of perspective :-)
Dec 15, 2017 · Arran Glen
Fair point, and is certainly more technically correct. The reason I originally had it in the interface is to catch as many unnecessary request/reply interactions (that should actually be modelled as events) early. Many distributed application designers default to request/reply when they could be using events (as you know), and this tool is trying to help break that thinking.