I have a question for the microservices community. I'll give an example from the educational field but it applies to every microservices architecture.
Let's say I have student-service and licensing-service with a business requirement that the number of students is limited by a license. So every time a student is created a licensing check has to be made. There are multiple types of licenses so the type of the license would have to be included in the operation.
My question is which approach have you found is better in practice:
Build a composite service that calls the 2 services
Coupling student-service to licensing-service so that when createStudent is called the student-service makes a call to licensing-service and only when that completes will the student be created
Use an event-based architecture
People talk about microservice architectures being more like a graph than a hierarchy and option 1 kinda turns this into a hierarchy where you get increasingly coarse composites. Other downsides is it creates confusion as to what service clients should actually use and there's some duplication going on because the composites API would have to include all of the parameters that are needed to call the downstream services.
It does have a big benefit because it gives you a natural place to do failure handling, choreography and handle consistency.
Option 2 seems like it has disadvantages too:
the API of licensing would have to leak into the student API so that you can specify licensing restrictions.
it puts a lot of burden on the student-service because it has to handle consistency across all of the dependent services
as more services need to react when a student is created I could see the dependency graph quickly getting out of control and the service would have to handle that complexity in addition to the one from its own logic for managing students.
Option 3 While being decoupling heaven, I don't really think would work because this is all triggered from an UI and people aren't really used to "go do something else until this new student shows up" approach.
Thank you
Option 1 and 2 creates tight coupling which should be avoided as much as possible because you would want to have your services to be independent. So the question becomes:
How do we do this with an event-based architecture?
Use events to keep track of licensing information from license service in student service, practically a data duplication. Drawbacks here are: you only have eventual consistency as the data duplication is asynchronous.
Use asynchronous events to trigger event chain which ultimately trigger a student creation. From your question, it looks like you already got the idea, but have an issue dealing with UI. You have two possible options here: wait for the student creation (or failure) event with a small amount of timeout, or (event better), make you system completely reactive (use server-client push mechanism for the UI).
Application licensing and creating students are orthogonal so option 2 doesn't make sense.
Option 1 is more sensible but I would try not to build another service. Instead I would try to "filter" calls to student service through licensing middleware.
This way you could use this middleware for other service calls (e.g. classes service) and changes in API of both licensing and students can be done independently as those things are really independent. It just happens that licensing is using number of students but this could easily change.
I'm not sure how option 3, an event-based approach can help here. It can solve other problems though.
IMHO, I would go with option 2. A couple of things to consider. If you are buying complete into SOA and furthermore microservices, you can't flinch everytime a service needs to contact another service. Get comfortable with that.... remember that's the point. What I really like about option 2 is that a successful student-service response is not sent until the license-service request succeeds. Treat the license-service as any other external service, where you might wrap the license-service in a client object that can be published by the license-service JAR.
the API of licensing would have to leak into the student API so that you can specify licensing restrictions.
Yes the license-service API will be used. You can call it leakage (someone has to use it) or encapsulation so that the client requesting the student-service need not worry about licensing.
it puts a lot of burden on the student-service because it has to handle consistency across all of the dependent services
Some service has to take on this burden. But I would manage it organically. We are talking about 1 service needing another one. If this grows and becomes concretely troublesome then a refactoring can be done. If the number of services that student-service requires grows, I think it can be elegantly refactored and maybe the student-service becomes the composite service and groups of independently used services maybe be consolidated into new services if required. But if the list of dependency services that student-service uses is only used by student-service, then I do not know if its worth grouping them off into their own service. I think instead of burden and leakage you can look at it as encapsulation and ownership.... where student-service is the owner of that burden so it need not leak to other clients/services.
as more services need to react when a student is created I could see the dependency graph quickly getting out of control and the service would have to handle that complexity in addition to the one from its own logic for managing students.
The alternative would be various composite services. Like my response for the previous bullet point, this can be tackled elegantly if it surfaces as a real problem.
If forced each of your options can be turned into viable solution. I am making an opinionated case for option 2.
I recommend option 3. You have to choose between availability and consistency - and availability is most often desired in microservices architecture.
Your 'Student' aggregate should have a 'LicenseStatus' attribute. When a student is created, its license status is set to 'Unverfied', and publishes an event 'StudentCreated'. The LicenseService should then react to this event and attempt to reserve a license for this student. It would then publish a 'Reserved' or 'Rejected' event accordingly. The student service would update the student's status by subscribing to these events.
When the UI calls your API gateway to create a student, the gateway would simply call the Student service for creation and return a 202 Accepted or 200 OK response without having to wait for the student to be properly licensed. The UI can notify the user when the student is licensed through asynchronous communication (e.g. via long-polling or web sockets).
In case the license service is down or slow, only licensing would be affected. The student service would still be available and would continue to handle requests successfully. Once the license service is healthy again, the service bus will push any pending 'StudentCreated' events from the queue (Eventual consistency).
This approach also encourages expansion. A new microservice added in the future can subscribe to these events without having to make any changes to the student or license microservices (Decoupling).
With option 1 or option 2, you do not get any of these benefits and many of your microservices would stop working due to one unhealthy microservice.
I know the question has been asked a while ago, but I think I have something to say that might be of value here.
First of all, your approach will depend on the overall size of your final product. I tend to go with a rule of thumb: if I would have too many dependencies between individual micro-services, I tend to use something that would simplify and possibly remove these dependencies. I don't want to end up with a spider-web of services! A good thing to look at here are Message queues, like RabbitMQ for example.
However, if I have just a few services that talk to each other, I will just make them call each other directly, as any alternative solutions whilst simplifying the architecture, add some computing and infrastructure overhead.
Whatever approach you will decide to go with, design your services in a Hexagonal architecture in mind! This will save you trouble when you decide to migrate from one solution to another. What I tend to do is design my DAOs as "adapters", so a DAO that calls Service A will either call it directly or via message queue, independent of the business logic. When I need to change it, I can just change this DAO for another one, without having to touch any of the business logic (at the end of the day business logic doesn't care how it gets the data). Hexagonal architecture fits really well with micro-service, TDD and black-box testing.
Related
I'm an architect in a large scale financial company and we are in the beginning of implementing a new business oriented infosystem across our different countries.
From the very early on the core idea has been to follow microservice oriented principles as much as possible (and making sure engineers have read Building Microservices book by Sam Newman).
By now I've come to crossroads. Our services are primarily JSON REST services using Swagger for automated documentation, but in order to use these services in our business processes and making sure not to write business logic into services outside the domain of those services, we've been using Camunda as an orchestration tool. And Camunda is fine (though some have considered Corezoid as an alternative), but somewhat clumsy in what is an otherwise an elegant set of services.
Now service orchestration is a concept pretty familiar to most engineers. But it is one that I am not entirely happy with due to still having a central engine that drives everything. It is incredibly expensive to replace later down the road (though still cheaper to replace than a monolith). And even if this central engine is split into multiple engines (which is actually the case today), it does not necessarily make it much better.
In recent years there has been a movement with microservices towards choreographed (close to event-driven) architecture. It is at this point where I am looking for advice from engineers and architects who have faced similar crossroad decision points.
I absolutely love the idea of decoupled architecture and despite feeling good about killing monoliths and having elegant independent services, I still detect a lot of dependencies in business process as a whole in current orchestrated solution in where it should not actually exist.
And it's not like we are avoiding events. We have actually implemented events on our architecture as well in order to decouple many processes with the core principle that if you don't need a synchronized response and just need to notify of something happening to initiate another process an event is put up that may be caught by another process that starts executing. And orchestration is easier to explain and visualize, it is easier to tweak and modify by more technical minded business users. And I think it is easier to test and validate from business perspective. Orchestrated architecture like this also (usually) expects a good service discovery and quality automated documentation and non-functional requirements which are all things I value greatly.
All of those things that are a question to me in choreographed approach since I don't have first-hand experience in running this in large scale - just some local test prototypes.
But I think you see where I am coming from. I'm trying to consider alternatives without having to regret driving the company all the other way in the end.
Perhaps you can share your own experience with a similar situation or share an interesting link or two? Or am I looking for a silver bullet that doesn't exist yet?
Services need to interact - services that don't interact are not part of the same system. The search needs access to the catalog, the cart doesn't get the price info from the page, the account needs the purchase history, the recommender needs purchase history, the cart needs to verify the currently available coupons, the inventory needs to know something was purchased etc.
Service boundaries are set to minimize the needed interactions. It can make sense to cut a service to smaller components but if they share a database (internal structure) they are different aspects of the service.
When services interact it creates a level of coupling - at the least, this coupling is some API (JSON or otherwise) that the service has to "maintain" for so other services can interact with it.
Another coupling type is temporal coupling - which is what you get in request-reply situations (and you can eliminate in event driven systems) However, Orchestration vs. Choreography is not about these differences (even though orchestration is mostly associated with request/reply) - it is about central control and governance vs.flexibility and serendipity.
Orchestration has risks like migrating business logic out of services into the orchestration while choreography runs the risk of chaos. By the way, direct request/reply integration has the worst of both worlds but wins on simplicity when systems are small enough.
Choosing between the two is a balancing act (like most architectural decisions) for instance, Netflix built on choreography for a lot of time but then found they need some control back and introduced an orchestration engine. Nothing is a silver bullet :)
Personally, I like choreography better because of the reduced coupling and flexibility and favor tools like open Zipkin to bring some order into the chaos.
You can see a partial example for an orchestration based arch in slides 10-22 of a presentation I did about microservices
I think I understand where you're coming from, having recently redesigned a system to a "microservices" architecture. I like (and use) the approach by these guys: http://scs-architecture.org/
The main point is, that you try to avoid cross-dependencies between you "services", which basically makes choreography obsolete. The hard part is decomposing your problem domain into chunks which do not need eachother for any of the executed business-cases. They may need different kinds of data that may or may not be "shared", as in present in multiple systems, but they don't need synchronous calls between them for any given business case.
This is quite different from what Netflix is doing for example. Those guys/gals are doing chain-calling through different layers of services, each adding its logic to the "process". This model might fit in some cases, and probably fits in Netflix's case. But it may not be necessary for you.
The ideal Self-Contained System would be completely independent of other Self-Contained Systems, would cover one or more highly cohesive business functions (in full depth from the UI to Persistence!), and would be not calling any other system synchronously. The ideal system would let the client "orchestrate", by just offering links through its Web (HTML) interface.
Think more like Amazon. The "Landing Page" is a different application than the "Search", which is still different from the "Checkout". They are completely different, sometimes even look a bit different! Integrated by links and forms in HTML, not explicitly orchestrated.
This might be what you are looking for.
Some warnings: First instinct of some people is to have "Customer" microservice, or "Product Repository" microservice, and similar. This will not lead to Self-Contained Systems, as you will need synchronous calls to these things, making them essentially "central" components. The key is to split the business domain, so bounded contexts a la Eric Evans.
When building a microservice oriented application, i wonder what could be the appropriate microservice granularity.
Let's image an application consisting of:
A set of various resources types where each resource map a given business model. (ex: In a todo app resources could be User, TodoList and TodoItem...)
Each of those resources are saved within a NoSQL database that could be replicated.
Each of those resources are exposed through a REST Api
The application manage an internal chat room.
An Api gateway for gathering chat room and REST api interaction.
The application front end: an SPA application connected to the API Gateway
The first (and naive) approach when thinking about how microservices could match the need of this application would be:
One monolith service for managing EVERY resources and business logic:
By managing i mean providing the REST API for all of those resources and handling the persistance of those resources within the database.
One service for each Database replica.
One service providing the internal chat room using websocket or whatever.
One service for Authentification.
One service for the api gateway.
One service serving the static assets for the SPA front end.
An other approach could be to split service 1 into as many service as business models exist in the system. (let's call those services resource services)
I wonder what are the benefit of this second approach.
In fact i see a lot of downsides with this approach:
Need to setup an inter service communication process.
When requesting a service representing resource X that have a relation with resource Y, a lot more work are needed (i.e: interservice request)
More devops work.
More difficulty to share common code between resource services.
Where to put business logic ?
When starting a fresh project this second approach seams to me a bit of an over engineered work.
I feel like starting with the first approach and THEN split the monolith resource service into several specific services depending on the observed needs will minimize the complexity and risks.
What's your opinions regarding that ?
Is there any best practices ?
Thanks a lot !
The first approach is not microservice way, by definition.
And yes, idea is to split - each service for Bounded Context - One for Users, one for Inventory, Todo things etc etc.
The idea of microservices, at very simple, assumes:
You want to pay extra dev-ops work for modularity, and complete/as much as possible removal of dependencies between different bounded contexts (see dev/product/pjm teams).
It's idea lies around ownership, modularity, allowing separate teams develop their own piece of code, without requirement from them to know the rest of the system . As long as there is Umbiqutious Language (common set of conventions/communication protocols/terminology/documentation) they can work in completeley isolated, autotonmous fashion.
Maintaining, managing, testing, and develpoing become much faster - in cost of initial dev-ops and sophisticated architecture engeneering investment.
Sharing code should be minimal, and if required, could be done to represent the Umbiqutious Language (common communication interface/set of conventions). Sharing well-documented code, which acts as integration/infrastructure mini-framework, and have special dev/dev-ops/team attached to it ccould be easy business, as long as it, as i said, well-documented, and threated as separate architecture-related sub-project.
Properly engeneered Microservice architecture could lessen maintenance and development times by huge margin, but it requires quite serious reason to use it (there lot of reasons, and lots of articles on that, I wont start it here) and quite serious engeneering investment at start.
It brings modularity, concept of ownership, de-coupling of different contexts of your app.
My personal advise check if you really need MS architecture. If you can not invest engenerring though and dev-ops effort at start and do not have proper reasons for such system - why bother?
If you do need MS, i would really advise against the first method. You will develop wrong thing's, will miss the true challenges of MS, and could end with huge refactor, which could take more work than engeneering MS system from start properly. It's like to make square to make it fit into round bucket later.
Now answering your question title: granularity. (your question body bit different from your post title).
Attach it to Domain Model / Bounded Context. You can make meaty services at start, in order to avoid complex distributed transactions.
First just answer question if you need them in your design/architecture?
If not, probably you did a good design.
Passing reference ids between models from different microservices should suffice, and if not, try to rethink if more of complex transactions could be avoided.
If your system have unavoidable amount of distributed trasnactions, perhaps look towards using/making some CQRS mini-framework as your "shared code infrastructure component" / communication protocol.
It is the key problem of the microservices or any other SOA approach. It is where the theory meets the reality. In general you should not force the microservices architecture for the sake of it. This should rather naturally come from functional decomposition (top-down) and operational, technological, dev-ops needs (bottom-up). First approach is closer to what you would need to do, however at the first step do not focus so much on the technology aspect. Ask yourself why would you need to implement a separate service for particular business function. Treat it as a micro-application with all its technical resources. Ask yourself if there is reason to implement particular function as a full-stack app.
Some, of the functionalities you have mentioned in scenario 1) are naturally ok, such as 'authentication' service - this is probably good candidate.
For the business functions decomposition into separate service, focus on the 'dependencies' problem, if there are too many dependencies and you see that you have to implement bigger chunk of data mode - naturally this is not a micro service any more.
Try to put litmus test , if you can 'turn off' particular functionality and the system still makes sense - it is the candidate for service or further decomposition
When establishing several modular and independent services, I am challenged with dependencies / stored relationships between entities. Consider Job Position and Employee. In my system, the Employee's Assignment is linked (URI) to the Job Position.
For our application, the Job Positions would be managed by a separate service than the Employee service, which leads to the challenge of constraints to prevent inadvertent removal of a Job Position if an employee is already matched to that position.
I've designed a custom solution leveraging a Registry (which should have dependency details, etc.) and enforce a paradigm across the inter-dependent services, however it is complex. In the SOA environment, how could one manage these inter-dependencies?
Many thanks in advance!
In some ways your question could be rephrased as "How to enforce referential integrity in SOA environment". Well the answer is you can't. That's kind of a by-product of the Autonomous in the tenets of SOA.
So almost by definition, the Job Position in the Employee service is not the same thing as the Job Position in the Job Position service. This is actually a good thing. Even though both services define Job Position, they do so from two different capabilities, and are free to develop and evolve their capability as needs arise.
So, hard constraints on the removal of data within one service boundary based on the existence of similar data inside another service boundary are just not possible (or even desirable).
This is all very well, but then how do you avoid the situation where Employees may be "matched" to a Job Position which has changed in some way, either via removal or update?
Well, services can be interested in changes to other services. And in these situations, services can become consumers of each other. It's fairly obvious the Employee capability would be interested in changes to the Job Position capability.
Events are actually a fairly well used design pattern for this scenario. If a business action results in a change the data of a service, that service can publish an event message which describes the change. Other services can become consumers of this type of event and can handle it in their own fashion. Because eventing is usually implemented with a pub-sub semantic, any service capability which so desires can subscribe to the event.
In your example, the event which could be published if a job position was deleted could be defined as (using C#):
class JobPositionRemoved
{
int JobPositionId { get; set; }
string JobPositionName { get; set; }
...
}
How a consumer of this event actually handles it (what action would be taken by the consumer) is another question and would depend on the capability of the consumer. As an example, your Employee service could gather a list of the Employees with this job position and flag them for review, or add them to a queue for "job position reassignment".
Your event could even include a field called int ReplacedByJobPosition which would enable consumers to automatically update any capability that depended on the removed job position.
As long as your event is delivered across a fault-tolerant transport (such as message queuing), you can be fairly confident that while you won't have referential integrity between your service capabilities, your system as a whole should become consistent eventually.
By using events in this way, you also avoid the need for a centralized registry of inter-dependencies (which sounds like a nasty idea). Each service is responsible for publishing events about changes to it's own data, and dependencies are defined by services consuming events from each other.
Hope this is helpful.
EDIT
In answer to your comment - while I can see the benefit of having another service taking care of the position:reassignment problem and I don't see any massive problems with this, there are a few considerations.
One of the reasons why service boundaries and business capability boundaries are a natural fit is that when you change a business capability (eg a change in Billing procedure) it does not generally impact other business capabilities (CRM/Finance/etc). By introducing shared services you're coupled to more than one capability, your service doesn't have well defined boundaries, and as a result has a higher cost of ownership as it will need to be changed a lot.
Additionally you could argue that the consumer of a business event (eg, JobPositionRemoved) should take responsibility for the entire handling of that event.
The handling of the event may well trigger a subsequent event to be published (such as ReviewTaskCreatedForEmployeeChange) which can then be handled by another consumer (eg a workflow tool) if desired.
To what degree should web service providers limit implementation changes without creating a new service version? One view is that as long as the contract is upheld, the service owner should be free to update the implementation as needed. Schemas are not always air tight and it is foreseeable that changes within the service implementation affect the service output while still upholding the contract.
To what degree should consumers be notified of implementation changes? Its one thing to notify consumers of updates to your own web service implementation. How feasible is it to track implementation changes to all downstream dependencies? Should service owners create a new version when they know that a change may affect consumers? And try to be a good citizen and notify consumers of all other changes?
Lots of questions and I doubt there is one size fits all answer. It could just depend on the situation. Maybe this is what SLAs are for.
Good questions, and I think you've already answered it. Yes, these details would be in an SLA and I think that if the contract/WSDL is the same that why would the service need to notify its' consumers? Unless of course changes to the service impact response times and performance. Maybe the service would notify consumers when another contract is introduced (in addition to the original). Consumers become aware of any new capabilities and can adjust their clients accordingly if desired.
I'm in an environment where SLAs don't exist for internal clients, so absent an SLA, the following are some common sense guidelines
Attempt to limit number of modifications to services
Communicate service implementation releases so consumers can plan test cycles
Provide consumers with the list of direct downstream dependencies and location to find their schedules and release notes
Consider a new version if an implementation change will semantically affect consumer
A lot depends on your specific circumstances. Speaking generally, here are a few top considerations.
The service contract and schema are all that a service and client share in common. A service implementation change that does not change the contract or schema (e.g., fixing a bug in the implementation logic) should not necessitate notifying the clients, nor should it be considered a new version.
OTOH, if you have a poorly constructed, overly-loose contract, such as passing all of the data as one big string, where the client had to do extensive interpretation to consume the service, and now you're looking to exploit that overly-loose contract in a way that would likely break the client, you owe it to all parties to change the contract (and improve it!) and publish that as a new version of the service.
Since services are often used to enable loose coupling between services, it is sometimes not practical or even possible to identify all of the clients of a service. Producing a new version of a service in these situations often entails maintaining multiple versions of a service for some period of time, often as directed by some governance body.
Providing details about service implementations, implementation dependencies, etc., encourages creating tight coupling by disclosing non-contract related details that the client may then take a dependency on. That can limit the ability of the service to change independently of the client.
The book Web Service Contract Design and Versioning for SOA
by Thomas Erl is a good resource on the topic, and details several common scenarios.
We have a website, where transactions are entered in and put through a workflow. We are going to follow the standard BLL(Business Logic Layer), DTO(Data Transfer Object), DAL(Data Access Layer) etc. for a tiered application. We have the need to separate everything out because some transactions will cross multiple applications with different business logic.
We also have a backend processor. It handles our transactions once the workflow has been completed. It works with various third party systems, some of which are unstable, or the interface to them is unstable, and then reports the status of the transaction. Each website will have its own version of the backend processor.
Now the question, with N-Tier, they suggest a new BLL for each application. With the layout of the application above, it can be argued that the backend processor and website is one application acting in unison, or two applications with different business logic. What would be the ideal way to handle this? Have it act like one system, or two?
One thing that I picked up on while learning MVC over the last couple years is the difference between what I call application logic and domain logic. I don't like the term business logic anymore, because it has too much baggage from all the conflicting theories and practices that have used that term too loosely.
Domain logic is the "traditional" business logic, how things are supposed to act, what they require (validation), etc. Application logic is anything that is specific to a given presentation of your domain, IE when the user clicks this submit button in your web app then they are directed to this web page over here (note that this has nothing to do with how a WinForms app or a background processor would work). Application logic should live in your application. Domain logic should live in your BLL and lower, and be reusable across the different applications that may use your common "business logic".
Kind of a general answer, but I hope that helps.
You might consider partitioning the functionality to reflect the organization of the stakeholders. Usually if you have two distinct organizational groups, then development and administration requirements are easier to manage if the functionality is similarly partioned. And vise versa.
Most of us don't spend that much time writing applications that explore the outer boundaries of hardware and software capabilities.
If you separate your concerns well then I think that you will be able to view them as the same application with a single business logic layer, there is no point writing the same code twice. The trick will be forcing the separation of concerns between the user interface portions of the website and the business logic in your BLL library.
Performance is going to be an issue as well, you have to ensure that your batch processing doesn't block your website from performing tasks that it needs to perform due to your resources. This may be an argument to keep them more separate, however as they're likely sharing a database anyway (or some other file based resource) then that may be an issue regardless.
I would keep a common business logic library programmed to interfaces and fully separated from your other concerns.
The "Ideal" way to do this depends on the project at hand and the various requirements of the system.
My default design is to have it act as one app. But if there are more heavyweight processes taking place, I like to create a batching process where the parameters of the requested job are stored and acted upon by a seperate process.