Aye. The images seem to be static: you always get the same one for any given size. So there isn't much work being done to serve them most of the time. I expect they have nginx configured to serve a file statically if found, and on error go to the script that will pic one, resize as needed, serve it, and also drop the result into place so next time that size it requested the image can be sent as a simple static file.
CPU time is often a limiting factor with shared or VPS hosting, this will most likely keep that from being where the bottleneck actually is. Or the hit of DB access, which this scheme would not need (at least not for most requests, perhaps there is a DB of images used when a request for a new size comes in).
CPU time is often a limiting factor with shared or VPS hosting, this will most likely keep that from being where the bottleneck actually is. Or the hit of DB access, which this scheme would not need (at least not for most requests, perhaps there is a DB of images used when a request for a new size comes in).