Learn what to consider when migrating to a cloud graph
GraphOS supports self-hosted and cloud graphs, each recommended for different use cases. You may want to migrate to a cloud graph in these scenarios:
- You have multiple GraphQL APIs you want to compose into a supergraph.
- You have one or more self-hosted GraphQL APIs and want to use GraphOS features beyond graph composition, for example,
@defersupport or metrics reporting.
- You have a self-hosted supergraph and want to offload the management of your router service to Apollo while retaining or gaining access to premium router features like subscription support, authentication, and more. Your router service may be either the Apollo Router or the
Dedicated cloud routers currently support all premium router features except for safelisting with persisted queries, automatic persisted queries, and offline licenses. Support for both persisted queries features is on the roadmap.
Cloud routers use the same Apollo Router binary that you can self-host. Therefore, migrating from
@apollo/gateway entails migrating to the Apollo Router. Refer to the Gateway migration guide for tips.
The Apollo Router supports a few avenues for customization:
- Custom router binaries
- Rhai scripting
- External coprocessing
As a managed service, cloud routers don't support running custom binaries. Cloud routers don't currently support Rhai scripts, though support is on the roadmap. Therefore, you must migrate any customizations to external coprocessors or built-in router features to use cloud routing.
Built-in router features that you may have previously supported with customizations include:
See the Apollo Router documentation for a full list of features.
Once your implementation is ready to run on the Apollo Router, including customizations, you can follow the Dedicated quickstart to get started.
Cloud Dedicated pricing depends on throughput instead of operation volume. Refer to the Throughput guide to learn more.
When you no longer host your router, router-to-subgraph communication may be inter-region, inter-Availablity Zone (AZ), or egress traffic. Be aware of how your new network topology may affect your cloud costs.