[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gearman
Subject:    Gearman & TheSchwartz (was: Cache miss stampedes)
From:       ask () develooper ! com (=?ISO-8859-1?Q?Ask_Bj=F8rn_Hansen?=)
Date:       2007-07-30 7:55:38
Message-ID: CBE5870B-D40D-4139-9801-30E51B903A6E () develooper ! com
[Download RAW message or body]


On Jul 27, 2007, at 16:41, Brad Fitzpatrick wrote:

> This is totally the wrong list, but...

Moving it to the right list.  :-)

> Gearman:  low latency, failures/retries done by client (who's waiting,
> perhaps on dozens of things in parallel).  No disks involved.  Just  
> load
> balancing function calls inside your network.
>
> TheSchwartz:  high latency.  retries done by system.  database(s)
> involved, thus disks.  use for long/slow tasks like sending email  
> or SMS
> or video transcoding, etc... gearman is for calling functions  
> elsewhere on
> your network while you wait.
>
> Both complement each other wonderfully.  Neither is going away nor  
> being
> folded into the other.

Why can't TheSchwartz just be a kind of backend to gearman / use the  
gearman protocol and services?

Something like:

Queue A Job:
    Client -> gearmand -> TheSchwartzWorker (as a sub-process or 3  
off the gearmand)

Get A Job Done:
    TheSchwartzScheduler (again process from gearmand) -> gearmand ->  
JobWorker -> ...



    - ask

-- 
http://develooper.com/ - http://askask.com/


[prev in list] [next in list] [prev in thread] [next in thread]