Page 1 of 1

Organization of objects?

Posted: Thu Jun 30, 2011 7:21 pm
by sonofodin
Hi all,

Have been developing in TM1 for about 7 months now. Although I am the lead, there are 3-4 other developers right now, and a bunch more projects in the pipeline.

We currently just have one instance in each environment (Dev, UAT, Prod).

I've noticed things getting out of hand in Architect. There are 20+ cubes, 50+ dimensions, 70+ processes. My initial thought was to create a new instance in each environment for different business groups, but I wanted some feedback before doing this (coming from the Cognos BI side of the house, all I really need is folders, but they don't exist).

Pros:
-Better organization
-Names of objects can be shorter (currently they include business group)
-Security could be simpler, given most users only see one business group and have blanket access to it
-Instances not in use could be shut down to preserve resources

Cons:
-Increased time to set up initially, and when new business groups come aboard
-If a user is in more than one business group, security can take longer to set up
-Overhead of service itself (my admin quoted 10mb, but I sense this was rather unscientific number)
-Inability to link data sources across instances
-Inability to share dimensions/processes

Does that about cover it? Any other thoughts? How do you guys manage a large number of objects (aside from deleting them :) )?

Re: Organization of objects?

Posted: Thu Jun 30, 2011 8:13 pm
by tomok
Determing how many envronments you were going to need should have been part of the original project scoping. That being said, the decision really depends on how the different cubes are going to share dimensons, attributes, and data. If there is going to be a lot of sharing going on then you should really leave it in the same instance and come up with a better naming scheme to keep the different developers from bumping in to each other. TM1 is a tool that when you have multiple developers and no traffic cop, then it can get messy REAL quick, with a lot of duplicaton and such. If each of the developers is really working in a silo then it may make sense to have separate instances. There is no one right answer, it's mostly a judgement call. I normally put functional systems in the same instance. For example, I would normally create a planning system in a single instance. If I was also creating a reporting system and/or consolidation system, that would normally be another instance.

Re: Organization of objects?

Posted: Thu Jun 30, 2011 11:50 pm
by Martin Ryan
I agree with all tomok said.

If you feel you are a strong enough lead (or have a strong enough project manager) who knows TM1 well then I'd recommend using one instance and enforcing naming conventions.

When I was at a large organisation a while ago we had about half a dozen developers, all with very healthy opinions of their own abilities and methods/conventions of doing things. The PM wasn't especially skilled in TM1. What ended up evolving was separate instances of TM1 and any shared dimensions had an agreed master dimension that was then propogated out to the other instances via csv files.

In fact even this became a bit unweildy so we implemented a "metadata" instance that was tightly controlled. It maintained version control by keeping datestamped copies of dimensions. No elements were ever deleted to avoid loss of data on any instance, they were only removed from hierarchies and made orphans. This system did a pretty good job of keeping the common dimensions in sync but leaving developers to do other things their way.

Martin

Re: Organization of objects?

Posted: Wed Jul 06, 2011 1:59 pm
by sonofodin
Thanks for the replies, more or less what I figured.

Has there ever been talk of implementing a folder structure? I figure it has to have been talked about before, and is something the veterans are divided on.

Re: Organization of objects?

Posted: Wed Jul 06, 2011 2:13 pm
by tomok
sonofodin wrote:Has there ever been talk of implementing a folder structure? I figure it has to have been talked about before, and is something the veterans are divided on.
What does it matter? I'm sure it has been talked about on this forum but there is no correlation between the wishes on this forum and what Applix/Cognos/IBM develop. I also doubt there has ever been any division on the subject, who wouldn't want to see the capability to organize cubes and dimensions under folders if they so choose?

You are aware of the Applications folder aren't you? You can organize cube views, Excel/Web reports, and TI processes into your own oranization structure using this main folder. If you do it properly, a user would never have to look anywhere else. Admins still need access to the full tree and that's where the folders concept for the other objects would come in handy. So, in other words, all they are really missing is a folder scheme for administration of TM1.

Re: Organization of objects?

Posted: Wed Jul 06, 2011 11:21 pm
by Alan Kirk
tomok wrote:
sonofodin wrote:Has there ever been talk of implementing a folder structure? I figure it has to have been talked about before, and is something the veterans are divided on.
What does it matter? I'm sure it has been talked about on this forum but there is no correlation between the wishes on this forum and what Applix/Cognos/IBM develop.
I wouldn't have expressed it so narrowly. I might have said it as:
There is often little correlation between what end users as a whole want and/or need and/or would make life easier for them, and what Iboglix develops unless it looks good in a flashy multimedia presentation.
The last time I recall them being truly responsive to user needs was at a user conference in the late 90's when people were complaining about the hassle of having to export and re-import data when element names changed in version 6. "Is that really a problem for many of you?" the Applixian asked. Most hands in the room shot into the air. "Well that's surprising, we didn't think it would be that much of an issue." Then in the very next version, version 7 we got aliases, yay.

On the other hand I can recall people asking for a "Who Am I" worksheet function. Yes, we eventually got it, but how many years did it take, and how much fluffing around in the meantime in irrelevant contemplation about whether it should be a worksheet function or a macro (what the...??)?

Icons to show which chores are active? Finally got it in 9.1 but they couldn't be bothered retrofitting it to the then-still "supported" 9.0 which was probably the most widely used version at the time.

Undo spread? An obvious and huge need despite some users being told that it was irrelevant. Yes, again we finally got it but they had to be dragged kicking and screaming into implementing it over many years and many versions. And many years after it had been implemented in some competing products.

But an ElIsAnc worksheet function? Local time scheduling of chores to prevent the "6 month shuffle"? Being able to separate user views from TI source views? The ability to have Websheet hyperlinks to non-Excel documents in Applications folders so that you can restrict access based on TM1 security and don't have to fluff around hosting them on another site? Variable AND LARGER fonts in the TI editor, or even something as basic as find and replace? (You can even change the frappin' font in NOTEPAD for Sagan's sake, but in the TI editor? That'd be way, way too 21st century. And nobody can tell me that those things haven't been asked for, loudly and repeatedly, by many, many users over the years.) Neh. We don't see why you need those things, therefore you don't.

Then there are the ones which are half-a**ed jobs which allow "completed" boxes to be ticked but seem to display limited grasp of real world needs. F'r instance? The ability to turn off some of those bloody annoying dialogs like "You don't have access to all elements". Yes, you can finally, FINALLY check "Don't put this waste of time in my face again" but if you ever want to see it again, say to check security access for a group, you have to go in and hack an .ini file to turn it back on again. (Still using an .ini file for anything other than default initialisation in 2011. Cheeezes.... :roll: ) Gods forbid that ALL options (including whether or not the toolbars load, especially that waste-of-space, no error handling included TM1 Servers one) could be set through an options form.

Or the implementation of a "safety lock" on the Clear spread command, a clearly overwhelming need. I was originally told that it was going to be implemented as a multi-key command. "Sounds good", thought I, "after all, the modifier keys on a keyboard are there for a purpose". What did we get instead? A stupid bleeding confirmation dialog which (a) Ties one hand (or at least some fingers) behind the back of anyone who prefers to use a keyboard over a mouse for speed, (b) Requires two actions to do the clear instead of one, and the cherry on the cake (c) Prevents you from entering the letter "C" into a string cube without resorting to ridiculous workarounds like entering "cx" then editing out the x. ("Aw gee, we didn't think of that!") All of which would have been avoided had they listened to the user base in the first place and implemented what had been asked for, and what was at one point planned, which was a multi-key solution. You'd think that a company with IBM's resources could afford to kidnap someone from Adobe's Photoshop GUI development team to teach them how to design a GUI right. (With the sole exception of the [F12] key; I have no idea what on Earth Adobe was thinking in assigning that to "Revert" in PS and AI given the wide use of that key as "Save As". The fact that it's undoable in PS saved my bacon more than once before I customised it out. Yes, PS lets you completely customise your keyboard shortcuts and menus. That's because the last time Adobe worked on its GUI (aside from changing a bunch of long-familiar icons in Subset Editor), the year had a "2" in front of it. )

So no, it would be rather optimistic to expect Iboglix to listen to what's being said in this Forum or anywhere else. Unless of course you can do a video mock-up of the proposed feature, overlaying it onto shots of a sports car driving quickly down night time roads with a gift-wrapped box which has semi-transparent brightly-coloured web dashboards flying out of it sitting in the passenger seat. If you make such a video, load the video up onto You Tube and point the Iboglix marketing team to it, then they won't be able to implement it fast enough.

Which leads us to Applications folders. I can imagine the scene when those, which are in themselves a very good idea, were developed.

"OK, so users can click and drag views into the Applications folder! Look, it's really easy!" {Demonstrates on a copy of Planning Sample.}
"Thumbs high! We've done it again!"

Except... most real world models tend to have rather more cubes than Planning Sample does. So the reality is that if the cube or process is down near the bottom of your list, you click on the view/process and drag.

And drag. And drag. And drag. And drag. And drag. And drag. And drag. And drag.

And drag.

And hope that the mouse button doesn't lose its contact while you're dragging.

Want to drag multiple views into an application folder? Nope, one at a time. You can select multiple views in the Properties window, but can you either drag them or copy and paste them into an Applications folder? Nope. Better still, do you have a right click "Send To Applications Folder" option? Nope. Can you apply wildcard filters (more than one, the way the subset editor doesn't allow you to) to the lists of objects shown in Server Explorer so that you can limit it to only the objects that you specify, which would make this process easier? Nope. Do I expect to see such advances in the current millennium? Nope.

Hmm, probably best not to get me started on that topic.
tomok wrote:I also doubt there has ever been any division on the subject, who wouldn't want to see the capability to organize cubes and dimensions under folders if they so choose?
I can't recall any division on the subject either, aside from whether they should have done it 5 years ago or 10 years ago. However I suspect that they're probably of the view that Application folders serve this role adequately, notwithstanding the many, many limitations of setting them up. Once they are set up though, I'd agree that it could become the primary interface for many.