What can startups learn from Pokémon GO’s server scalability engineering?

[Quora Feeds]

Pierre Chapuis


“Spinning up more instances” is a gross oversimplification of the issue.

Remember, from what we know:

• Niantic is full of Xooglers (it used to belong to Google);
• they built this on the Google Cloud platform;
• they had years to prepare with Ingress.

What can we learn from that?

• The issue is probably a software design problem;
• geo-located, write-heavy loads are hard;
• even the best teams will have scalability issues in the face of crazy growth like this.

Also:

• A CDN won’t solve this. They probably already use one for assets but almost everything the app does is dynamic requests and a lot of those are writes.
• They can’t rely on the client to do much for them because it would make it too easy to cheat.
• The client deals with outages poorly. This means its release has probably been rushed. I would guess they had a hard deadline to launch, agreed on with Nintendo, and they made compromises.

Now they need to work on both the server and client side to solve this. Progress in the weeks / months to come will let us judge how good their team really is.



See Questions On Quora

Continue reading...