Hi Fred,

You may also know it as consistent hashing, it always sticks in my
head as clock hashing for some reason. Here is article about it:
http://www.spiteful.com/2008/03/17/programmers-toolbox-part-3-consistent-hashing/

Regards,
Erich

On May 16, 8:49 pm, Fred Wulff <[email protected]> wrote:
> Hi Erich,
>
> Could you elaborate on what you mean by clock-hash style resource
> management? I don't think I'm familiar with the concept.
>
> Thanks
> -Fred
>
> On Wed, May 13, 2009 at 8:52 AM, Erich <[email protected]> wrote:
>
> > I think a unique instance id, (sha-1 or whatever) is a good idea also.
> > There are weird edge cases otherwise. For instance if there is a
> > beanstalkd on 2 subnets for whatever reason, what is the convention?
>
> > I suspect however that I am just excited about he idea that this makes
> > clock-hash style resource management way trivial, and that I'm looking
> > for other excuses to  back me up.
>
> > Either way, the globally unique id saves code on the client side as
> > well, by not needing to maintain references to connections in jobs.
> > This allows for easier fun like passing jobs off to internal sub
> > processes and whatnot.
>
> > Regards,
> > Erich
>
> > On May 13, 4:30 am, Keith Rarick <[email protected]> wrote:
> >> On Tue, May 12, 2009 at 6:18 PM, Fred Wulff <[email protected]> wrote:
> >> > At the moment, there doesn't appear to be a way to easily and globally
> >> > identify a beanstalk job.
> >> > ...
>
> >> I like this idea. I have a couple of thoughts about the details.
>
> >>  * I think it should appear as a field (or fields) in the stats response.
>
> >>  * I'd prefer if it were transparent. For example, if it is based on
> >> the host name and a time stamp, those should be plainly visible. In
> >> other words, I don't want it to be just a SHA-1 hash or something.
>
> >>  * Perhaps the actual constuction can be left to the clients, who
> >> already have almost enough information. They already know (or can get)
> >> the IP address and port number of the service, the pid of the daemon,
> >> and the job id. All they need in addition is the time the server was
> >> started. Perhaps a random number is also necessary. How likely is a
> >> server OS to reuse a pid within one second? These things could be
> >> added as two separate fields in the stats output.
>
> >> kr
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"beanstalk-talk" 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/beanstalk-talk?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to