Hacker News Reader

Comment by mpdehaan2 | Hacker News Reader

Programming architecture/problems in SC, yes.

I think Pokemon scaled remarkably better than any app many of us have ever dealt with, seeing the crazy demand spikes. You would run into a lot of app issues, data choices, a lot of things that don't just get solved from simple uses of autoscaling.

View on HN

It scaled better than Twitter, it hit the same active user count (or close to it) in a month that took Twitter years to reach, and Twitter would sh*t the bed all the time with the infamous "fail whale" popping up for users constantly even with its steady increase in growth over time and time to adapt. Twitter was pioneering at reaching that scale, that many now use as guide/lessons learned now, so its forgiven, but how many companies see that much growth, that fast?

And for just a little extra context, Twitter launched the same year as AWS.

To be fair -- Twitter is a much harder architectural problem (significant fan-out and fan-in problems to solve).

Pokemon Go is less complicated in terms of the conceptual model. There are only a few pretty simple ways that different players can interact (gyms, lures on pokestops, that's about it).

On the other hand, Twitter is dealing with strings < 140 character. Pokemon Go is tracking all the trainers and Pokemon at all times, not to mention the gyms and lures.

Sure, but you can partition based on country, city and even block, it's clear how this should work.

None of Twitter can be partitioned like that because every user follows different people.

Of course this makes you wonder if this is at all necessary and the right design decision.

Reply on news.ycombinator.com