Mosh, On 5/31/2011 1:42 PM, Mosh Teitelbaum wrote: > > All: > > I'm working on a system that will have a large number of timed, > one-time tasks that need to be executed on a schedule. I'm trying to > figure out the best way to accomplish this with cfschedule. The two > options I see are as follows: > > 1. A single, recurring task that runs every 5 minutes and queries the > database for any tasks that need to be executed. The benefit of this > approach is that there is only a single task to create. The > negatives of this approach are (a) potentially querying the database > when there are no tasks to execute (waste of resources) and (b) risk > of too many tasks at any given time causing the scheduled execution > to time out or cause later tasks in the run to be executed so late as > to be off schedule.
(a) The overhead of running a script that doesn't find anything to run should be extremely minimal, so I wouldn't even factor this in as a waste of resources. (b) You can always limit the number of tasks you're going to execute in one pass--that way you'll never run too many requests in a single batch. Also, depending on your needs, if you have some requests that you know are long running, you could build into your scheduling queue when a valid window for that task is to run. So, if you know a task might take 5 minutes to run (because it does a ton of warehousing, etc) you could make sure those type of events only run during non-peak hours. > 2. Create a new cfschedule task for each task that needs to be run. > The benefits of this approach are that it solves all of the negatives > of the other approach. The negatives are that it would necessitate > the creation of a very large number of cfschedule events and I don't > know if this could negatively affect the scheduler or CF > performance. > > Either approach would include code to do cleanup (e.g., delete the > task from the CF scheduler after it has been executed, etc.). > > Any thoughts as to which approach (or a different one altogether) is > better? I've always preferred having a single background task to manage a queue, rather than trying to schedule everything via <cfschedule /> and clean up afterwards. For one, it's much easier to catch errors. It also gives you a single point of entry/failure. You can easily build reports on your executed tasks--because you're already storing that stuff in the database. If it's important enough, you can find out how long each task is taking to execute, find which generated errors, re-queue events, etc. You can also really easily just fire off stuff in queue manually. -Dan -- Dan G. Switzer, II dswit...@pengoworks.com http://blog.pengoworks.com/ âââââââ ââ âââââ âââââââ â âââ â âââââ ââ â âââ â â âââ â ââ â âââ â âââ â âââââââ âââ â â â âââââââ ââââââââ âââââ âââââ âââ ââââââââââââââ âââââ ââ ââââââââââââââ âââââ ââââ ââ ââ â ââââââââ âââââââ ââ ââ â ââ âââ ââââââââ â âââââââ âââ âââââ â âââââ â âââ â âââ ââ ââââââ ââ â âââ â â ââ ââ âââ âââ â âââââââ ââ ââ ââ â ââ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~| Order the Adobe Coldfusion Anthology now! http://www.amazon.com/Adobe-Coldfusion-Anthology/dp/1430272155/?tag=houseoffusion Archive: http://www.houseoffusion.com/groups/cf-talk/message.cfm/messageid:345015 Subscription: http://www.houseoffusion.com/groups/cf-talk/subscribe.cfm Unsubscribe: http://www.houseoffusion.com/groups/cf-talk/unsubscribe.cfm
