Yes, I know that first (and a few in case of firing them one after
another in short delays like 200 ms like in my benchmark) request is
slow, but I care about the next requests, that take ~50 ms without any
distance between server and client and ~200 ms from
appspot.com, 150
ms more then by pinging it. Well, sometimes I get a few requests in
middle with ~600 ms lag - can it be because requesting my app every
200 ms still isn't enough to keep it warm?? I rather suspect that too
big frequency of requests is making a bottleneck on client side.
I deleted all Aral Balkan's stuff and made my own code:
http://2.latest.lagstest.appspot.com
and this time tiny sourcecode:
http://2.latest.lagstest.appspot.com/lagtest.zip
and used URLLoaded instead of AMF - no difference.
I tested it on a cheap PHP server I use and it was better, but when I
test it know to make sure, it's much worse then GAE.
Probably You can't achieve better (smaller) lags with this technology.
But that is an answer I want to get from some guru, so I don't need to
bother doing various benchmarks.
Because lags increase when user makes client CPU busy, I'm pretty sure
the 50 ms local lag is a flash issue. I will repost it on some flash
forum.
Still, I'm VERY suprised that noone before benchmarked GAE's lag...
On 15 Lip, 01:38, Nick Joyce <
n...@boxdesign.co.uk> wrote:
> Sorry for the double post, couldn't see my first answer :)
>
> Appspot likes to consume instances of your application fairly quickly
> if you don't get a lot of traffic. Initialising a cold one (for the
> first request) can take quite a bit longer than a warm instance, due
> to all the code loading you have to do (python imports etc.). You may
> be hitting into this.
>
> I would suggest removing your dependancy on Aral's code and build an
> app as described inhttp://
pyamf.org/tutorials/gateways/appengine.html.