On Saturday, February 22, 2014 2:22:56 AM UTC+8, Rafael Sanches wrote:
>
> The scheduler logic doesn't make any sense for java apps. For frontend 
> serving I would rather just use a scheduler that boots one instance at time 
> and would spread traffic linearly through all instances. This would be more 
> reliable due to the issues with warm up. 
>
> I have to keep 6 useless instances running, just because when a instance 
> is idle for a long time its very probable its in a zombie state, so even if 
> you are paying high costs for resident instances there's a high probability 
> it won't be able to serve requests when you need it.
>
> The real and only problem is that we are stuck into this senseless 
> scheduler due to the fact that its the only one who have discounted hours. 
> It almost seems that google uses the resident instances to serve other 
> traffic when they become idle (even if I'm paying for it)
>
I want the scheduler always use the only resident instance to handle 
requests. :)

Yes, we need more settings for the scheduler to define the beviour of the 
scheduler. 
Otherwise, it make you feel your fate is controlled by scheduler instead of 
yourself. :D
 

> On Feb 21, 2014 12:47 AM, "Tapir" <[email protected] <javascript:>> wrote:
>
>>
>>
>> On Friday, February 21, 2014 3:29:37 PM UTC+8, Vinny P wrote:
>>>
>>> On Fri, Feb 21, 2014 at 1:23 AM, Tapir <[email protected]> wrote:
>>>
>>>> So for the app with warmup problems, to avoid the warmup the problem, 
>>>> the real free hours would be only a little more than 4 hours, right?
>>>> And for the 15 minutes tax design, the "a little more than 4 hours" is 
>>>> about several minutes in fact, right?
>>>>
>>>> Ok, I see why one hour compute hour is enough for my app but my app is 
>>>> still often counted more than 28 hours now.
>>>> So you help me confirm again using Java
>>>>
>>>
>>>
>>>
>>> The 15 minute rule is not confined only to the Java runtime, it exists 
>>> for all runtimes. It's not a tax; it's there so you avoid the overhead of a 
>>> second warmup request if another request comes in soon after the first. 
>>>
>>
>> I really admire the design of GAE instance scheduler and billing. 
>> According what you say, at an extreme case
>> 1. to avoid warmup request, a resident instance is created, 
>> 2. the website get a request every 15 minutes, 96 request a day.
>> So the scheduler doesn't like let the resident instance to handle the 
>> requests and let another dynamic instance to handler the requests.
>> By what you say above, the total counted hours for the dynamic instance 
>> is 24 hours!
>> The total front end hours is 24 dynamic instance hours + 24 resident 
>> instance hours = 48 hours!
>> So cool! The cooler thing is someone like it and think it is great!
>>  
>>
>>>
>>> If you wanted to avoid keeping instances up for the extra 15 minutes, 
>>> you could always cause the instance to terminate; for example, exceed the 
>>> memory allocated to your instance.
>>>   
>>>  
>>> -----------------
>>> -Vinny P
>>> Technology & Media Advisor
>>> Chicago, IL
>>>
>>> App Engine Code Samples: http://www.learntogoogleit.com
>>>  
>>>  -- 
>> You received this message because you are subscribed to the Google Groups 
>> "Google App Engine" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to [email protected] <javascript:>.
>> To post to this group, send email to 
>> [email protected]<javascript:>
>> .
>> Visit this group at http://groups.google.com/group/google-appengine.
>> For more options, visit https://groups.google.com/groups/opt_out.
>>
>

-- 
You received this message because you are subscribed to the Google Groups 
"Google App Engine" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/google-appengine.
For more options, visit https://groups.google.com/groups/opt_out.

Reply via email to