Thanks for replying...

So in my example of rate=50/s and bucket-size=5:
When no tasks are in the queue, clearly the bucket fills up and has 5
tokens.
If I add 100 tasks, the first 5 will execute immediately and take all
5 tokens... but now the bucket is empty.

I have specified a rate of 50 tokens to be added to the bucket every
second... but there are two ways that could happen:
1) 50 tokens are deposited every 1 second.
2) 1 token is deposited every 1/50th of a second.

If the first is true, then by using a bucket of size 5, no more than 5
tasks will ever be able to execute in a second because tokens are only
being added 50 at a time to a bucket that can only hold 5 tokens.
If the second is true, 5 will burst initially, but then there will be
a steady stream of 1 task every 50th of a second.

>From what you said in your post: "the token-adding resolution is the
unit of the rate you set..." it sounds like the first is true... which
would mean that the task queue is essentially ALWAYS bursting.  I
guess that's not a huge deal, but means that to truly execute 50 tasks
a second, I have to bump my bucket-size to 50 as well, in which case
the task queue essentially starts 50 simultaneous tasks instead of
staging them across the entire second...  the behavior I would prefer
is the second...

On Feb 1, 1:44 pm, Robert Kluin <[email protected]> wrote:
> Hi,
>   The rate is like an average upper bound;  in other words, the
> token-adding resolution is the unit of the rate you set.  And, it is
> 'bursty,' so if you set a rate of 50/M you might very well have 50
> tasks execute in the first second then the queue will not execute any
> tasks for about 59 seconds.  Setting a higher bucket_size will let the
> queue 'burst' at a higher rate, I usually see that after a queue has
> emptied.
>
>   At least that's how I've seen the task-queue behave.
>
> Robert
>
>
>
> On Mon, Jan 31, 2011 at 14:57, HalcyonDays <[email protected]> wrote:
> > So I've read through a couple of things about the Task Queue's Token-
> > Bucket system, and I've taken a look at the wikipedia article to no
> > avail, so forgive me if this question has been answered elsewhere:
>
> > What is the resolution of the clock that deposits tokens into the
> > bucket for each queue?  Is its minimum resolution "per second?"
> > i.e.  If I say, "I want my rate to be 10/s," are 10 tokens deposited
> > into the bucket every second, or is one token added every 10th of a
> > second?
>
> > It seems like if the resolution is only per second, the rate is
> > limited not only by the user's specification, but also by the bucket
> > size...
>
> > For example: If I have a specified rate of 50/s but a bucket size of 5
> > and the system deposits 50 tokens into the bucket every second, are
> > the additional 45 just thrown away?  Does my rate actually become 5/s?
>
> > I'm specifically using the Java app engine, if there's a difference
> > between the Task Queues in Python/Java.
>
> > --
> > 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 
> > athttp://groups.google.com/group/google-appengine?hl=en.

-- 
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.

Reply via email to