Azure API Management vs. CloudFoundry

APIM is useful for the standard scenarios:

1) Securing your back-end APIs - If you have a legacy back-end web service that has a basic authentication scheme, you can add some additional security by placing APIM in front, and requiring subscription keys. Leverage your existing firewall to ensure only your APIM instance can communicate with your back-end API, and you've basically added a layer of protection.

2) Lift and shift - there are always going to be clients that don't want to update their clients to use a newer API; in some cases you can make a newer API look like an older one by implementing some complex policies in APIM. You can also do the opposite, making older APIs look new, such as making an XML back-end accept both JSON and XML.

3) Centralizing your APIs - if you've acquired another company and want to make their API set look as if it's a part of the larger whole, APIM is an easy way to provide a consistent front-end interface for developers.

Incentivized

It's well suited if:

• The organization has large number of applications that needs to be deployed frequently.
• The organization is tied to the DevOps mindset.
• The organization has programs in different languages.
• The applications does not need EJB's support that servers like web logic provide.

It's less suited if:

• The applications needs security configuration within the same CloudFoundry instance.
• The organization, for whatever reason does not want developers to manage the instances.

Incentivized

• Easy commissioning of APIs.
• Great policies to control access.
• Easy mock services for testing.

Incentivized

• Support for Orgs and Spaces that allow for managing users and deployables within a large organization.
• Easy deployment, deploying code is as simple as executing single line from CLI, thanks to build-packs.
• Solid and rich CLI, that allows for various operations on the instance.
• Isolated Virtual Machines called Droplets, that provide clean run time environment for the code. This used to be a problem with Weblogic and other application servers, where multiple applications are run on the same cluster and they share resources.
• SSH capability for the droplet (isolated VM's are called droplets), that allows for real time viewing of the App code while the application is running.
• Support for multiple languages, thanks to build-packs.
• Support for horizontal scaling, scaling an instance horizontally is a breeze.
• Support for configuring environment variable using the service bindings.
• Supports memory and disk space limit allocation for individual applications.
• Supports API's as well as workers (processes without endpoints)
• Supports blue-green deployment with minimal down time

Incentivized

• Lack of robustness is a bit of an issue. Several other providers offer more options and capabilities, but then, they are lacking in interface ease.
• As with anything Azure, pricing is really hard to stay on top of. I always find that you really don’t know what you’re paying for until you get the bill. Having an excellent Azure Administrator can help resolve that.
• Integrating with app services outside of Azure can be a challenge, or at least much more challenging than just using Azure App Services.

Incentivized

• Does not support stateful containers and that would be a nice to have.
• Supports showing logs, but does not persist the logs anywhere. This makes relying on Cloud Foundry's logs very unreliable. The logs have to be persisted using other third party tools like Elk and Kibana.

Incentivized

Azure APIM vs Amazon API Gateway:

1) Azure APIM was a complete package that included a developer portal.

2) We are very Microsoft centric - so the Microsoft product suite aligned very well with our business needs.

3) It was faster and easier to stand up Azure APIM for testing than it was for the Amazon API Gateway.

Incentivized

While Docker shines in providing support for volumes and stateful instances, Cloud foundry shines in providing support for deploying stateless services. Heroku shines in integrating with Git and using commits to git as hooks to trigger deployments right from the command line. But it does not provide on-premise solution that Cloud foundry provides.

Incentivized

• We can always think of positive ROI impact on business
• It helps to easily facilitate the design, deployment, and maintenance of our APIs

• Positive impact, since it simplifies the deployment time by a huge margin. Without cloud foundry, deploying a code needs coordination with infrastructure teams, while with cloud foundry, its a simple one line command. This reduces the deployment time from at least few hours to few minutes. Faster deployments promote faster dev cycle iterations.
• Code maintenance such as upgrading a Node or Java version is as simple as updating the build-pack. Without cloud foundry, using web logic, the specific version only supports a specific version of Java. So updating the version involves upgrading the version of web logic that needs to involve few teams. So without cloud foundry, it takes at least few days, with cloud foundry, its a matter of few mins.
• Overall, happier Developers and thats harder to quantify.

Incentivized