Projecting Server Scheduling Use

Question

I'm familiar with how to pull data from the MongoDB, the basics of its layout, the Server Use macro, etc.

But what I think is missing from the Server Use macro is a process to project forward the expected use in order to identify the best times for new schedules to be added.

I get the idea that this information can somehow be pulled from AS_Schedules, by converting the text in IterationFormula into something that can be used by the Generate Rows tool.  Whereby one row is added for each next date in the iteration sequence, as long as the date is less than (set variable date here), for example Current Date + 28 to see what the busy times will be over the next 28 days)

Has anyone done this or know how to do this?

ScheduleIteration.png

patrick_mcauliffe · Answer

@Matt_Curtis

Predicting the on-demand runs would probably be much further down the road.  That's a whole different concept, and probably easier (to me) since you can make projections based on history.

Yes, I would use the DateTimeAdd(); but keep in mind, that only produces one single answer per row and we need to know every instance that will run over the next 28 days of those currently scheduled.

I believe the DateTimeAdd() would have to be a function in a Generate Rows tool.

The formula itself is in AS_Schedules, but I can't get Alteryx to treat it as a formula.  If I use the field as a formula in a tool, it still only sees it as text.

See the second row of "NextRunDateTime":

Matt_Curtis · Answer

That's a really good idea!

I'm not familiar with pulling the Mongo data myself yet, but I understand what you mean.  You must be able to see when things are scheduled, their frequency, etc. from the AS_Schedules table (?). Why not use a datetimeadd() function to project out the schedules?  Your projection would be based on your reporting time frame (say, 28 days) and the unit of measure on the job frequency (weekly, daily, hourly).

The other issue would be to predict the 'on demand' reports that are run, if any.  We currently have Gallery users (non-Artisans) using Apps that have been developed which get used sparingly.  I suppose you could pull frequency stats for those and project forward based on averages.  For instance, one of our Apps is used much more frequently than the others, usually pulled around 9 or 10 am CST.

I think to cap it off it would be great to produce a .tde and visualize via a time-based grid heat map similar to the viz in this blog post that displays Trulia's site traffic by day & time of day.