Log in

No account? Create an account

prev | next

datastore/business flow question.

so tulane does all of their financial hoopla through oracle 11i. i haven't messed around with it too much because i don't have proper access and most of our budgeting happens via an external database through zoho.com that i set up.

something that comes up with the person who actually deals with the oracle system is this statement: "finance sends us an email around the beginning of the month when the monthly datastore reports are ready." the implication is that if we run a report for, say, february and datastore isn't "ready", then we lose data. Typically we get email notification about a week or a week point five into the new month.

on the one hand, this sounds fishy to me. ODS's are typically realtime data, yes? Yet there's an implication in the process that what people run reports from are monthly snapshots that act more like a datawarehouse rather than a datastore, and there's a grace period in which the data needs to be re-goobledeegooked into datawarehouse tables that the business then runs reports from.

on the other hand, it could be that we only have visibility to reports post-AR as opposed to what i'm used to which is transactional, so any $$ needs to close with an invoice date before we would actually see it on our report. If finance has some sort of end of month process that has a 7-day grace period before they hit zero backlog, that could also explain why reports wouldn't be ready until well into the next month.

so who can give me a more intelligent conjecture about this (if possible based on my limited information)? the main reason i ask is because i seem to recall that ODS date stamps retain hh:mm:ss and datawarehouse date stamps drop hh:mm:ss which makes start/end date parameters for report runs very different.

tag cloud:


( read spoken (4) — speak )
Robert Thatcher
Jan. 2nd, 2011 10:23 pm (UTC)
Sounds like they are using views for reporting and not live tables, which would require some sort of DTS job or whatever Oracle's equivalent is to populate the view with table data.

If you were running reports against live table data, and you knew the fields, you could construct whatever report you want regardless of monthend processing.

Your best bet is to talk to an actual DBA and not a front office person if you want the real scoop - you certainly won't get it from the front of the house.
Jan. 3rd, 2011 01:21 am (UTC)
yeah, i sent an email to the datastore admin a couple of weeks back but i never heard anything back from her. it's not a *huge* deal or anything, but i was curious about it and it also does potentially affect how we run the reports, although once we pull a couple of months of data we'll have it figured out through trial and error, so whatever.
Jan. 3rd, 2011 12:45 am (UTC)
Another possibility is that reports are run against aggregates that are cached manually over specific spans, and there's a job that periodically fills in those caches, in order to allow reporting them.

That's pretty much the opposite of real-time, but something you run into quite often.
Jan. 3rd, 2011 01:23 am (UTC)
ah, that makes sense too. cool, thanks.
( read spoken (4) — speak )


welcome to the lifeofmendel

you can also find me here:

meSubscribe to me on YouTube


March 2017