Even with threading I don't seem to be able to get an F8 backend to do more
than 1.9 CPU Seconds per second. Which seems to be the same number I get on
an F4 Backend.

Am I doing something wrong? Or is the Python Gil not capable of using all
the CPU on this system? Or is an F8 CPU Second twice as much as an F4 and
this is as expected.  If the latter why is the app only 15% faster on an F8
(I know you can't answer that with out the code)


If the gil is limited case why not allow backends to have more than one
request at a time?  If the Gil is not limited still why can I only have 1
request per backend?

 

Why can't I configure how many requests  Frontends handle? F1's serving 8
requests is silly. F4's handling 8 requests is often equally silly in the
opposite direction.

Why can't I configure how many requests a backend can handle? Or have a back
end act like a front end and handle 8?

 

Why can't I specify a specific version of an App in Apps For domains?  It
would be very useful to have  User.Somedomian.com be a version of an app
and Affiliates.somedomian.com be an app and they would share a database and
nothing else.

 

Public backends can run longer than 60s but only handle one request.  But
they 503 if the request before them in line takes more than 10s to complete.
Why is this not configurable?  Especially since dynamic Public frontends
seem to only spin up a new instances if there are 8 requests in the queue.
This means you get a lot of 503's 

 

Defer task buckets don't have a mechanism for manual token addition.  I'd
like this. It would let me read the queue and decide how many backends to
spin up, and when tasks completed in 1 to 500s I would put a token in the
bucket. (that's not a question so much as a Does this make sense?)

Memcache gets slow sometimes. Really slow. Slower than datastore. Is there a
"memcache health status" that we can hit to identify when we should stop
using it for a while? MasterSlave had a Readonly we could detect and change
behavior. I can't find any such thing for HR.

 

Requests through edge cache get edge cached. Why do outgoing requests not do
the same?  There seems to be a lot of people "scraping" content via
Appengine. Why are those requests not "edge" cached so that a. it would be
harder to build appengine DOS attacks. B. fetches would take less time.

 

Because I am threaded and use 200% more CPU seconds than Realtime seconds I
suspect I'm a bad neighbor to share an instance with.  Is this true? Am I
clobbering the neighbor?

 

Do we always have to go through Apps For Domains? How much money do I have
to send you to never have to talk to their tech support again? Similar
question Naked Domians. I want one. Building a URL Shortener that has to
have www. On the front is breaking my heart.

 


Brandon Wirtz 
Stremor.com Chief Technology Officer
BlackWaterOps: President / Lead Mercenary 

Description: http://www.linkedin.com/img/signature/bg_slate_385x42.jpg



Work: 510-992-6548 
Toll Free: 866-400-4536 

IM: [email protected] (Google Talk) 
Skype: drakegreene 
YouTube:  <http://www.youtube.com/blackwateropsdotcom> BlackWaterOpsDotCom 



 <http://www.blackwaterops.com/> BlackWater Ops 

 <http://www.cloudonastring.com/> Cloud On A String Mastermind Group



                

 

 

 

 

 

 

 

-- 
You received this message because you are subscribed to the Google Groups 
"Google App Engine" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/google-appengine?hl=en.

<<image001.jpg>>

Reply via email to