How recent can be the data in TM1

[aagrawal12]

Hello Experts,
Can anycody tell me how recent can be the data in the TM1.
I mean to say can i schedule the chores to run in 1 min or will it impact my performance of Server.

Thanks,
Abhishek

[Alan Kirk]

Hello Experts,
Can anycody tell me how recent can be the data in the TM1.
I mean to say can i schedule the chores to run in 1 min or will it impact my performance of Server.

Of course it will. Every refresh that you do will lock up your server for the amount of time that it takes to run, as well as blowing away the relevant calculations that may have been stored in memory which would then need to be recalculated when the users retrieve them. Unfortunately there ain't no such thing as a free lunch, even with something as fast as TM1.

We update our GL cube every hour normally, every 15 minutes on the last day of month end but we limit it to current month data only. That means that it takes 20 to 30 seconds. We have to wear the price of losing the stored calculated values since with journals flying through the GL every couple of minutes it's more important to have the updated numbers.

At a Cognos User Group I once saw a presentation by a group that did something similar every 5 to 10 minutes with a very fast connection, and were able to do it in a few seconds, but I think that even with a lightning fast connection doing it every minute would be pushing it. Even if it took only 10 seconds to run, your users would still be losing 1/6th of your working day just because of updates, not taking into account consolidation recalculation time.

[mattgoff]

The limiting factor for us is the performance of the source database. Although TM1 has made strides with locking granularity, it still seems that TM1 will lock up while it's waiting for the first row of data, even though I'm storing to a cube with no rules/feeders to or from it. This is with 9.1.4-- anyone have a different experience with 9.4 or 9.5?

During close, our Discoverer performance is pitiful so a 15 second process becomes a 2-3 minute process. I'm toying with writing a prefetch process in perl to trigger via TI to avoid this locking.

Matt

[lotsaram]

If your source supports polling for changes in tables you can also dramatically cut down on data volume and load time with a "trickle-feed" approach and do a cumulative load to the cube without a zero out. Locking with an SQLFetch is always problematic as this holds TM1 hostage to load on the source system, speed of driver, etc. Flat file output is an easy way around this issue.

The trickle-feed approach can reduce load time down down to seconds and make something like every 15min or even every 5min loads feasible. But typically there isn't a requirement for data to be "real time" to this degree. Of course the big watch out with this kind of approach is much more process control is needed to ensure that batches are not skipped or loaded more then once otherwise the data in TM1 cubes will no longer be accurate!