Service orchestration on Google Cloud
Staff Developer Advocate, Google Cloud
Going from a monolithic architecture to microservices has clear benefits, including reusability, scalability, and ease of change. Most of the time, business problems are solved by coordinating multiple microservices. This coordination is based on event-driven architectures, which can be implemented via two approaches: choreography and orchestration.
Service choreography and service orchestration
Service Choreography - With service choreography, each service works independently and interacts with other services in a loosely coupled way through events. Loosely coupled events can be changed and scaled independently, which means there is no single point of failure. But, so many events flying around between services makes it quite hard to monitor. Business logic is distributed and spans across multiple services, so there is no single, central place to go for troubleshooting. There's no central source of truth to understand the system. Understanding, updating and troubleshooting are all distributed
Service Orchestration - To handle the monitoring challenges of choreography, developers need to bring structure to the flow of events, while retaining the loosely coupled nature of event-driven services. Using service orchestration, the services interact with each other via a central orchestrator that controls all interactions between the services. This orchestrator provides a high-level view of the business processes to track execution and troubleshoot issues. In Google Cloud, Workflows handles service orchestration.
Once you have decided between the two approaches for your application, design questions are largely about the characteristics of the services and the use case. You should prefer orchestration within the bounded context of a microservice, but prefer choreography between bounded contexts. That is, you'll likely have choreography at a higher level, with orchestration at lower levels, both in the same system.
Google Cloud provides services supporting both orchestration and choreography approaches. Pub/Sub and Eventarc are both suited for choreography of event-driven services, whereas Workflows is suited for centrally orchestrated services.
Google Cloud support for service orchestration
Workflows is a service to orchestrate and automate Google Cloud and HTTP-based API services with serverless workflows. It is a fully managed, scalable, and observable way to define a business process and orchestrate calls to several services. Workflow calls those services as simple web APIs. Using Workflows you can define the flow of your business logic in a YAML-based workflow definition language and use the UI or API to trigger the workflow. You can use Workflows to automate complex processes including event-driven and batch jobs, error handling logic, sequences of operations, and more. Workflows is particularly helpful with Google Cloud services that perform long-running operations, as Workflows will wait for them to complete, even if they take hours. With callbacks, Workflows can wait for external events for days or months.
Google Cloud support for service choreography
Pub/Sub enables services to communicate asynchronously, with latencies on the order of 100 milliseconds. Pub/Sub is used for messaging-oriented middleware for service integration or as a queue to parallelize tasks. Publishers send events to the Pub/Sub service, without regard to how or when these events will be processed. Pub/Sub then delivers events to all services that need to react to them (Subscribers). Pub/Sub is also used for streaming analytics and data integration pipelines to ingest and distribute data (as covered in the Data Analytics post).
Eventarc enables you to build event-driven architectures without having to implement, customize, or maintain the underlying infrastructure. It offers a standardized solution to manage the flow of state changes, also known as events, between decoupled microservices. Eventarc routes these events to Cloud Run while managing delivery, security, authorization, observability, and error-handling for you. Eventarc provides an easy way to receive events not only from Pub/Sub topics but from a number of Google Cloud sources with its Audit Log and Pub/Sub integration. Any service with Audit Log integration or any application that can send a message to a Pub/Sub topic can be event sources for Eventarc.
Additional services that help with both choreography and orchestration
Cloud Tasks lets you separate out pieces of work that can be performed independently, outside of your main application flow, and send them off to be processed asynchronously using handlers that you create. These independent pieces of work are called tasks. Cloud Tasks helps speed user response times by delegating potentially slow background operations like database updates to a worker. It can also help smooth traffic spikes by removing non-user-facing tasks from the main user flow.
Difference between Pub/Sub and Cloud Tasks. Pub/Sub supports implicit invocation: a publisher implicitly causes the subscribers to execute by publishing an event. Cloud Tasks is aimed at explicit invocation where the publisher retains full control of execution including specifying an endpoint where each message is to be delivered. Unlike Pub/Sub, Cloud Tasks provides tools for queue and task management including scheduling specific delivery times, rate controls, retries, and deduplication.
With Cloud Scheduler, you set up scheduled units of work to be executed at defined times or regular intervals, commonly known as cron jobs. Cloud Scheduler can trigger a workflow (orchestration) or generate a Pub/Sub message (choreography). Typical use cases include sending out a report email on a daily basis, updating some cached data every x minutes, or updating summary information once an hour.
For a more in-depth look into the services covered in this post, check out the documentation.
Choosing the right orchestrator in Google Cloud
There are a few tools available for orchestration in Google Cloud—some better suited for microservices and API calls, others for ETL workflows. It’s not always clear which tool to use when. This post covers these products with examples of when to use which product.
By Rachael Deacon-Smith • 4-minute read