Mauro Servienti - Talk Session: All Our Aggregates Are Wrong 

Thanks

I fully agree

Thank you

Thank you

Jordan first of all, thanks! There are a few resources out there and we have a few customers using this approach. It’s important to keep in mind that in my experience such an approach makes things more complex for developers while the system is built. It’s easier to start with a monolithic UI and split it up later into a composed UI/ViewModel. This is because while the system is developed from scratch there is a high frequency of changes to the UI adding unnecessary friction to the process. On the other hand as the system is stable enough, it’s simpler to start thinking about the composition part; in most cases it’ll also be clear that service boundaries are correctly identified and stable. Udi Dahan (that by the way is the CEO of the company I work for) talks about UI composition for the first time in 2012 udidahan.com/2012/06/23/ui-composition-techniques-for-correct-service-boundaries/, udidahan.com/2012/07/09/ui-composition-vs-server-side-orchestration/. In 2014 there is a more formal definition udidahan.com/2014/07/30/service-oriented-composition-with-video/. I then started discussing the topic particular.net/blog/secret-of-better-ui-composition still using the client side composition concept. I then started transitioning from UI Composition to ViewModel Composition, nowadays UI Composition identifies a very different thing. Recently I wrote a series of articles on the topic milestone.topics.it/categories/view-model-composition, Jimmy Bogard discusses the topic as well jimmybogard.com/composite-uis-for-microservices-a-primer/. Finally there are a few stack overflow discussions such as stackoverflow.com/questions/8916379/how-do-you-aggregate-data-in-a-udi-style-soa-architecture or stackoverflow.com/questions/29224572/please-explain-me-the-service-oriented-composition-architecture-described-by-mr I recently found also this article www.ashrafmageed.com/Scalability3/ (2015). Finally, a very valuable resource is the ADSD workshop by Udi Dahan. Feel free to reach out to me in case you have any further questions.

Thanks for the thoughtful response! I’ll take a look at these resources when I get some free time :)

@@mauroservienti this comment is gold

Hi Matthew, thanks for the comment and more importantly for the patience. Unfortunately, there is no way to get notifications for comments on RU-vid videos not uploaded by me. Back to your questions: > does the information in each boundary transition to its own order model? Correct. > I could assume sales would be the one where this operation takes place, but how does validation in other services regarding the placement of this work? e.g. you could call sales to say place this order/order this cart, but what is there is a business rule for an order can only be placed if all of the items have stock at that time? Sales can have a saga that waits for confirmation events from other services before proceeding with the order. The most common use case being waiting for Payments to raise an event confirming the authorization on the user's credit card, for example. I wrote about some of these topics on my blog: - milestone.topics.it/view-model-composition/2019/04/18/the-fine-art-of-dismantling.html - milestone.topics.it/2019/05/02/safety-first.html Let me know if that answers your question. Do not hesitate to get in touch, the best way is to use the contact form on the blog. Cheers.

thanks for the feedback, much appreciated. Regarding the price update, out of the top of my head my answer would be yes. However it's really a business requirement, so it's a question that should be asked to the business. let me use a sample in which that price change should not be reflected: Hotel booking. The hotel room is basically the stock item, from the catalog, and room availability comes from "warehouse". In the hotel booking domain prices are frozen at search time. The query is performed, results are returned to the user along with prices. Those prices cannot change while I'm selecting a room.

It doesn't have; As suggested, when a command to update quantity is fired from client, and view-model decomposition occur, warehouse service will pick that command and behave according to it, so sales remain out of picture here.

:)

Didn't get what you mean, didn't he denormalized the data by creating redundancy in each micro-service? It seems like it's the payoff you have to make in order to gain the benefits of dependent services.

Why event modeling exists

@@AdamDymitruk Services have to communicate with each other in one way or another. Events are interfaces which services communicate through. So we have to model them I guess.

@@brainoverflow98 Yes sure - I agree that this is the trade off to make. I am just saying that data modelers/architects have know this for some time. This presentation seems to be saying "here is this great new way to handle this difficult problem" but data people have done this for years. So I am disappointed that data people are not more active in the microservice discussion - I think they could contribute a lot more or even lead much of it. And I am disappointed that programmers seem to be still quite uneducated about some database concepts that could provide the answers they are looking for.

@@timhosking ok got your point 👍 But what he has done in this video was denormalization, not normalization, right?

It's the defined convention for this programming language

@@coding-esmaster3259 stupid idea, but ok

@@cya3mdirl158 I think using Capitalizion convention for methos is really elegant :)

@@coding-esmaster3259 No it is only c# conventtion... Look at Kotlin, Groovy, Scala, Java, Python, C++ etc....

With all of the amazing info in this video, that’s what you focus on?