Hacker News new | comments | show | ask | jobs | submit login

PaaS is a tough thing to do. You have to automatically deploy new instances of your applications, distribute those instances according to the load on individual nodes of its clusters. Also, they have to do more or less the same with the datastore. If you have loads that experience sharp peaks, new instances will have to be spun up and that may take some time. Some seemingly random variations on response time are to be expected when you consider it's not only your application that's running there. You can somehow mitigate that by setting up warmup requests (https://developers.google.com/appengine/docs/python/config/a...) and tuning the minimum pending latency (https://developers.google.com/appengine/docs/adminconsole/pe...). If you build a fixed size cluster to run a single application, you'll experience much more predictable response times.

As for the price, you should consider what would it cost running your application on IaaS, rented or co-lo hardware. As always, YMMV.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact