Our legacy LAMP stack would have required a lot of knobs within AWS to scale the entire system (the ELB, application layer, caching layer and MySQL). All of these would have to be coordinated together and likely some of them would take at least 30 minutes to add and provision a new VM. Beyond that, adding new MySQL replicas on the fly with the amount of data we have was simply not possible to do reliably.
Considering the same ideas in GAE, the only layer we have to worry about scaling is the application layer. App Engine not only makes provisioning applications much easier but it can also add new instances in less than 30 seconds sometimes.
If you read the blog above, I also cover several other reasons beyond scaling that were factors in making the move (local dev, monitoring & alerting, etc.)
