High number of Gateways hosting Few APIs

[rw]

             We are developing towards an architecture of having Multiple Gateways clusters ( 50 to 100) that will each host a small number of apis (<10).  We would like all of the gateways to communicate back to a centralized API Portal to enable registry, discovery and policy management of all APIs.  Is this a pattern anyone has experience with?

[MattM]

Hi rw, I definitely have experience with this.  I sent you a separate note.

Thanks, m@

[Chris Mullins]

I suspect all of the big players in the space do what you've just described. 

There are dragons here. The problem is frightfully complex at scale and deployed worldwide. Managing the dependency chain is key, as the number of failure modes is very high, and the method of failure is (seemingly) different every time. 

Make sure you've got folks who are distributed systems experts in place. Dealing with unexpected latency issues, network partitions, cascading failures, and accidental call rate amplification attacks should all be front of mind. 

With that said, it obviously works. It just takes engineering effort, and operational excellence. 

Cheers,

Chris