PaaS from Google, give it some code and it will run it and scale it automatically
Overall Satisfaction with Google App Engine
It serves all our traffic to end users, which is basically one of the most important things for our organization.
Pros
- Serving traffic to end users. It can scale automatically when traffic spikes.
- The standard environment has some limitations, but it encourages you to write "scalable" code.
- With Flexible Environment, you can serve any Docker container you want, still taking advantage of auto scaling.
- Easy integration with other Google Cloud products, e.g. Datastore, Pub/Sub, Cloud Storage, etc.
Cons
- Flexible environment needs scaling to zero and support for all APIs available in Standard Environment like ndb for Python and Task Queue.
- Standard Environment needs to update some outdated libraries like lxml for Python.
- Instance pricing of Standard Environment could be lowered, since it wasn't updated for many years.
- It serves traffic very well with almost zero down time – it's always positive impact.
You can spawn up your own cluster using Kubernetes or Container Engine which will scale automatically when configured properly, but you have to keep an eye on that cluster. In App Engine you don't have to worry about it at all, just ship your code and it will run.
Google App Engine Feature Ratings
Using Google App Engine
Pros | Cons |
---|---|
Like to use Relatively simple Easy to use Technical support not required Well integrated Consistent Quick to learn Convenient Feel confident using | None |
- Scaling -- you just focus on writing your code without worrying about how to scale it.
- Versioning -- you can deploy multiple versions of your app and run them at the same time to perform QA, internal testing, public testing, a/b testing, etc.
- Datastore -- very scalable as long as you understand limitations, can write tens of thousands of entities from e.g. Dataflow/Apache Beam without any issues, and they are immediately ready to be served to end users.
- Migration to 2nd gen runtimes is hard, but they do provide a lot of benefit as well (ability to run any compiled code).
- No (good) full text search solution. While you can scale the datastore as much as you want, you have to go for third party solutions (e.g. elasticsearch) for full text search. Built-in Search API (which is not available on second-gen runtimes) is very limited and does not perform well.
Yes - It's OK, they provide some basic monitoring tools.
Comments
Please log in to join the conversation