TM1 Forum

We had a sticky thread for New Features and Issues of the 9.5.2 release to prevent the discussions fragmenting across an infinite number of threads, so we may as well do the same for 10.0.

I'll update the FAQ with the new documentation links shortly. {Edit: Since done.}

For anyone who hasn't got it here's the link to the 10.1 docs

http://www-01.ibm.com/support/docview.w ... RXT&mync=E

(Note the English Language docs are half way down the page and you may find yourself continually having to rescroll, but hey ho that's the IBM website (did I miss an 'h' out of 'website'?) )

Now I dont know where I got this impression but I was under the impression that 10.1 was going to be an absolutely ground breaking quantum leap ahead of 9.5.2 - the Sgt Peppers release we were all waiting for.

However:

Contributor has been renamed Insight (Well done marketing you really earned your money during this development cycle).

There seems to be a new development shell: Cognos TM1 Performance Modeler which might address a lot of the shortcomings of the existing development environment. But there is no manual which seems to lead me to the conclusion they haven't finished this bit yet.

In TM1 the new features seem to be

Everything is now parallel by default but you can make it serialisable if you want.
You can run TI 's from the command line (can do this already with freeware on this site and the implementation seems a lot cleaner the old way)
They've revamped TM1TOP and called it TM1 Operations Console (as above see Kyro's freeware on this site)
You can run a chore on server startup - quite handy
You can read data from C8 packages - conceivably useful

And err... that's it.


Cognos TM1 Performance Modeler seems interesting though - anyone got an insider track as to whether it's worth the wait and what it looks like, because otherwise there seems to be precious little over 9.5.2.

John Hammond wrote: Cognos TM1 Performance Modeler seems interesting though - anyone got an insider track as to whether it's worth the wait and what it looks like, because otherwise there seems to be precious little over 9.5.2.

It's a new model development tool (bye Architect?), web-based, with all the common Architect functionality + the ability to create "links" between cubes with a simple GUI (dlinks, anyone?), including matching up picklist items to real dimensions (virtual dimensions in EP). These "links" can then by saved either as rules (added to target cube, with feeders in source), or as a TI process. And there are some dev-prod migration features as well.
It's a totally new product and looks fancy, but one wonders how will it perform formatting of hundred pages rule-files and how interesting it'll be to debug those auto-generated feeders.

ykud wrote:It's a new model development tool (bye Architect?), web-based, with all the common Architect functionality + the ability to create "links" between cubes with a simple GUI (dlinks, anyone?), including matching up picklist items to real dimensions (virtual dimensions in EP). These "links" can then by saved either as rules (added to target cube, with feeders in source), or as a TI process. And there are some dev-prod migration features as well.
It's a totally new product and looks fancy, but one wonders how will it perform formatting of hundred pages rule-files and how interesting it'll be to debug those auto-generated feeders.

The concept of dlinks for TM1 is great. Hopefully it works! No more needing to manually enter the same rule in every cube for things like % of sales or growth % v LY woudl be great! I'm a lot more suspicious about bifs for TM1 and automated feeders, yes it might do the job but will it do it correctly and anywhere near as efficiently in terms making calculations optimally performant? I have my doubts, it could just be a revenue generator for consultants to clean up a whole lot of badly written auto generated code.

Great for presales but will it add any value to a customer once purchased? I'll reserve judgement until I can get my hands on it but I have some doubts.

Well so far it has been...painful.

Download was OK, unpacking less so. My server doesn't handle .tar archives so i had to extract it from my laptop across the network. Wasn't quick. Then i realised winzip wasn't playing ball and screwing up directory structures and had to extract it a second time. The multi-lingual package links me to install docs that are not locally held but on IBM webservers. Thanks, but for really fecking obvious reasons none of our servers are allowed to connect to the outside world, so none of that is any help. They are also in Mandarin or some similar Asian language....so much for multi-lingual hay?

The installer itself has changed significantly from 9.5.2. For the better too - its far better at guiding you thru the install process. Found that it "required" extra Microsoft junk before installing (Microsoft Chart Controls for .NET 3.5 & Visual C++ 2005 SP1). I'd like to say the rest has been a breeze but nearly 1hr 30mins later it is still on 167 of 187 components. No idea what that is but looking at task mangler i can see our virus software is chomping over half the CPU so my guess is that is slowing the installer considerably. Why is beyond me - no other previous installer has been anything like this bad. I also can not stop the AV, so i'm stuck with watching a non-moving installation progress bar while it does it's thang. The server i'm using is fairly puny as far as CPU goes but 9.5.2 only took 5 mins at most to install....

To Steve's point the installer seems much better. I installed the x64 server on my laptop without an issue. Though I have not used the new configuration manager to get all the other pieces going. For the core Tm1 components the installer got things up and running quickly.

I will see how much fun the it is to configure the new pieces tonight.

You have got to be **[Expletive Removed]** kidding me

You know that change people had been requesting for years where all we wanted was an easy way to see what chores were active? The one where in 9.5-ish they did it by greying out those not active?

Some bright spark has decided all the icons needed changing. Again. And in doing so you can no longer tell what chores are active. Seriously WTF is wrong with people at IBM?

John Hammond wrote:For anyone who hasn't got it here's the link to the 10.1 docs



Contributor has been renamed Insight (Well done marketing you really earned your money during this development cycle).

Cognos Insight is not a renamed Contributor - Contributor is now called Cognos Applications (or similar).

Cognos Insight is a new personal analysis tool and fat TM1 client - fairly cool.

J

Kyro's version of TM1top hasn't worked since around 9.0 - i asked a while ago about updating it and got no reply...

I currently can't get their new fangled web version to work. Seems to be installed and configured as per instructions but it refuses to let me log in (failed authorisation). No trace of any way to debug that, no clue in the guide either.

Steve Vincent wrote:Kyro's version of TM1top hasn't worked since around 9.0 - i asked a while ago about updating it and got no reply....

I think it's had a name change to cubespy and not sure if it's still free but the link is here http://www.tm1forum.com/viewtopic.php?p=27495

Interesting, but unsurprising, to see two things on the depreciated environments list a.k.a. "The Deathwatch":
- 10.1 is the last hurrah for Windows Server 2003 (no surprise there); but also...
- for TM1 Workflow.

I haven't had the pleasure of using Workflow but can't recall any posts singing its praises.

I was a bit confused by that as they have created a version 10.1 of TM1 Workflow and made it available for download.
Could be meant to be version specific or front end specific or something.

Gregor Koch wrote:I was a bit confused by that as they have created a version 10.1 of TM1 Workflow and made it available for download.
Could be meant to be version specific or front end specific or something.

They haven't pulled it as of this release; had they done that they would have had some hacked off users out there. Maybe not a lot, but enough. The EOL statement just sends a message... 'find another solution, because this one won't be here come version 11".

It's a common complaint, not least by me, that Iboglix fixates too much on brightly coloured semi-transparent eye candy which is of questionable real world value.

It therefore comes as a surprise bordering on shock to see something as well thought out (from the point of view of those of us on the pointy end of the legionary cohort) as a new TI function that I spied in the reference guide:

CellIncrementN
This is a TM1 TurboIntegrator function, valid only in TurboIntegrator processes.
This function increments an existing numeric cell value by a specified value.

When aggregating, this means half the amount of code; no more CellGetN/CellPutN, you can write the data source value straight into the cell using CellIncrementN.

We still don't (as far as I can see; I have yet to install it) have a decent TI editor in which to write the code, but this one statement does provide us with the chance to write less of it. Well done Iboglix!

(Now about an AsciiAppend function...)

Well, y'all should have known that me saying nice things about Iboglix couldn't last without a rip in the space time continuum occurring.

And so it came to pass.

Now, I'll confess that I got sick of waiting for the .pdf of the installation guide to show up, and trying to read sequentially through a web site is not my thing, so I missed this bit:

Upgrading to Cognos TM1® 10.1 from version 9.0.x, 9.1.x, or 9.5.x requires a full manual uninstall, and then a full installation of 10.1.

Silly me, thinking that the install package might take care of that the way many other apps do. So I'm not going to hold Iboglix to account for anything stemming from that one, though I'll mention in passing that Perspectives doesn't appear to care for having both versions installed. When I unloaded the 9.5.2 add-in and loaded the 10.1 one (which is a different location, of course, since we now have to brand the path with the IBM name) it bitched about not being able to find a .dll file. I was able to fire up Architect, though, but we're getting ahead of ourselves. I installed it on a 32 bit Win 7 Ultimate Core i3 notebook under an admin account but didn't set the install package to run as Administrator.

I'm not crazy about the new install package. Yes, the old one was a train wreck when it came to trying to install multiple databases (with the path in particular popping back to the default one unless you selected the order of the controls just so), but at least you had control over it. In this one, you don't seem to (unless I missed something) so the sample databases, once again, go in a sub-folder under the Program Files path. Now as I learnt from bitter experience the Vista and onward generation of Windows does not like data files being mixed with application files (something that purists claim is long overdue) so although I've not been able to test this yet I'm laying odds that I'll have the same happy experiences there unless I move the databases elsewhere.

But other than that it ran pretty smoothly. I installed the whole smash, and it took about 26 minutes all up. I had the same experience as Steve with item 167; whatever it is it took the bulk of the time though mine was 167 out of... 288 I think it was. The end dialog reported that the Configuration Tool completed with warnings and errors but there was nothing in the log about what those might have been.

As I mentioned, I couldn't get Perspectives up but that's probably down to the issue of both versions being installed. The Configuration Manager allowed me to launch one of the sample databases though I had to tell it to ignore the configuration changes (whatever it thinks those were) otherwise it would take forever to generate some encryption; I need to read up on what that's about.

The first thing that I noticed was that the 7 separate non-searchable help files are still on the Help menu despite this being griped about frequently and by many, so the "Muppet Of The Month" award is still in safe hands.

But Steve Vincent, HOW could you have possibly thought that the wonderful GUI design team would leave us without a way of telling whether a chore is active?? Why, if you put your glasses on... no, not those glasses, the ones that I sent you for Christmas to work with the TI editor (which is unchanged, by the way); yes, that's right, the ones from the Hubble Telescope Optometry Company, and look way, way down in the bottom right corner. When a chore is inactive, there's a little square with a diagonal bar through it. Well, no, maybe you can't see it but if you do a screen dump, paste it into PhotoShop and pump that sucker up to 300% magnification, then it'll be quite clear. I make it 7 pixels by 7 pixels. On a 1200 * 800 notebook screen. You do the maths and work out the clarity on that. This would seem to fail "Screen Cues 101". The old system and its colour coding system took only a glance, it was immediately obvious whether it was stop or go, on and off. If you have to look, really look at something and then work it out, that is A Design Fail.

But it gets even worse. Look at the icons for elements in subset editor. Yes, the consolidations are now a chunky orange sigma to help them stand out. I have no issue with this. But the N and S are no more. Instead, they've been replaced by a blue horizontal bar (oh, but a blue horizontal bar with a kinda-sorta gradient fill, don'cha know, which makes them all new wave and edgy and Web 2.5 and stuff to the extent that you can see it at normal resolution.) Underneath the bar is a microdot sized pictogram which seemed to be slightly longer for the S elements than the N ones. Again, copy it to PhotoShop, pump the h3ll out of the magnification, and the mystery is revealed; the thing under the N elements is a hash mark (sometimes incorrectly called a pound sign (#)), and the one under the string elements is a teeny, tiny a and b. To the extent that you can draw letters at that scale, that is.

So once again, real world, front line usability loses out to... oh, I don't {expletive deleted} know, somebody's idea of style or something? Honestly this kind of absolutely absurd design, with no regard for function and barely any for form, has just left me exasperated beyond the speaking of it.

I do not expect that to be the end of the fun, though.

I'm hoping that I'll have further positive things to say but it seems to be 1 step forward, 5 steps back.

But you never know, maybe replication works now. If not in this universe, maybe in a parallel one.

Edit: For those who have yet to have the pleasure of meeting 10.1, the new element icons are shown below. Tell me this isn't some bad joke.

Disclaimer, the rant below is based on the Beta codebases, I haven't had an opportunity to download the GA code yet - so I'm sure all my concerns have been addressed!

John Hammond wrote:Cognos TM1 Performance Modeler seems interesting though - anyone got an insider track as to whether it's worth the wait and what it looks like, because otherwise there seems to be precious little over 9.5.2.

You mean quite interesting in an 'OK, so I've gone to a web page and I've clicked an obscure button that launches a download that takes so long to load the splash screen let alone the full application that I might as well go and make a cup of tea' sort of way? Or the 'I've opened a rule and a cube view so I can see how my rule changes affect my cube, but now the panes of my single window are so small as to make them useless' way? Or the 'I want to see what the difference is between the rule in my Dev, UAT and Prod servers, what? I've got to go back to that web page and open up two more copies of NonPerformance Modeler as I can only see one server in an instance at a time?' way?

I'm sure the original intention was for PM to be a complete replacement for Architect, but there are so many things that you will find yourself having to revert to Architect to do. But to make things really annoying a lot of the new features you can't get to in Architect and have to use PM. I'm going to be sick of tea given the amount of time I'm going to have to spend waiting for PM to load. To make things really, really annoying, some of the new features will create objects that you can't even edit in Architect, even though they should be nothing more than good old fashioned TM1 objects.

Either way, IBM have said that PM is only aimed at people who are TM1 Administrators - in an access rights way, not necessarily in an 'I know what I'm doing building this system' way, so why the ridiculously convoluted method of launching it? Why not allow it to be installed locally as part of the standard install routine? And who on earth thought that the Cognos installer was a perfomance enhancement compared to the 9.5.2 installer? Yes the old one had its quirks, but at least they were (normally) quick quirks. Which brings me to the subject of Cognos Configuration. Who in their right mind thought that the world needed more instances of this snail of an application. If I am expected to use Cognos Configuration (and if I manage to get the correct instance, as opposed to the other instances, all called Cognos Configuration, that I have for Business Viewpoint Server, Business Viewpoint Client and Cognos BI) to create my TM1 Services, why can't I delete them using the same tool?

ykud wrote:and how interesting it'll be to debug those auto-generated feeders

Best part about the auto-generated feeders is that you can turn them off and write your own.

lotsaram wrote:it could just be a revenue generator for consultants to clean up a whole lot of badly written auto generated code

The thing that worries me is that the new type of customers that IBM are now also aiming for will be sucked in by the bells and whistles and smoke and mirrors and the promise of dragging and dropping an excel workbook to build a TM1 server. Then they will try it, realise that it doesn't work out-of-the-box like that nice man in the sharp suit said it would - unless you spend time rejigging the excel workbook to a format that TM1 can cope with - then, rather than calling in people who actually know what they are doing and who can show them how to write the system properly, they will put the box back on the shelf, well, if we still had something to put in a box that could then be put on a shelf, and ask Bill Gates' forgiveness for ever having doubted that excel was the be all and end all of business computing. Sure some IBM salesman makes his target, but the brand is soiled.

Steve Vincent wrote:I currently can't get their new fangled web version to work. Seems to be installed and configured as per instructions but it refuses to let me log in (failed authorisation). No trace of any way to debug that, no clue in the guide either.

I remember there being some definite strangeness with this thing. After you have set up the group and modified the xml file to give yourself access to the Operations Console I seem to recall I could access more servers than I was expecting to.

Alan Kirk wrote:It's a common complaint, not least by me, that Iboglix fixates too much on brightly coloured semi-transparent eye candy which is of questionable real world value.

Prepare yourself for a whole new world of flim-flam.

Alan Kirk wrote:We still don't (as far as I can see; I have yet to install it) have a decent TI editor in which to write the code

Depends if you like the 'Advanced' rules editor and whether you like 'Performance' Modeler. I like neither. You may have guessed that already...

As for Workflow, I was under the impression that it was lucky to make it to 9.5.2. It will probably live on in the same fashion as the Lazarus baiting ISB.

Alan Kirk wrote:Yes, the old one was a train wreck when it came to trying to install multiple databases

Hmm, yes, tried that once. Reverted to tm1sd -install.

But as I say, I'm sure they have fixed all of this and you are all wondering what I am blathering on about...

Roger.

About to test this upgrade ... do you folks have any show-stopper concerns with this release?

Going from 9.5.1 (64 bit - windows 2008 R2 install + cognos 10.1 / optional tm1web)

From reading documentation it sounds like a complete uninstall / re-install (much like the 10.1.1 release pack for 10.1 fp1 customers)
Are there any issues upgrading the cubes / models from previous versions?
do rules/feeders/cubes just work when added to this new platform?

you'll see from most of the posts so far that no one has really had a chance to do much with it. it's only just come out and getting hold of a 2GB installer has been hard enough. all i can tell you so far is it's different - all the icons have changed but it ran a test server from my 9.5.2 just fine. i really haven't done anything with it tho so who knows what may lurk when you start poking it with the test stick

Alan Kirk wrote:It's a common complaint, not least by me, that Iboglix fixates too much on brightly coloured semi-transparent eye candy which is of questionable real world value.

It therefore comes as a surprise bordering on shock to see something as well thought out (from the point of view of those of us on the pointy end of the legionary cohort) as a new TI function that I spied in the reference guide:

CellIncrementN
This is a TM1 TurboIntegrator function, valid only in TurboIntegrator processes.
This function increments an existing numeric cell value by a specified value.

When aggregating, this means half the amount of code; no more CellGetN/CellPutN, you can write the data source value straight into the cell using CellIncrementN.

We still don't (as far as I can see; I have yet to install it) have a decent TI editor in which to write the code, but this one statement does provide us with the chance to write less of it. Well done Iboglix!

I agree, this is great, ....BUT why for the love of <insert deity of choice> cant we have a TI function that returns a numeric value of the server build? Every other software on the planet has a function to do this and it would be EXTREMELY useful to have such a function in TI to allow for code branching to handle running the same TI under different TM1 versions where newer functions may or may not be supported. Quite a few come to mind beyond this new CellIncrementN function; GetProcessName, CubeClearData, ViewExists, SubsetExists, BatchUpdate, LockOn .....

There seems to be an assumption on the part of IBM that each and every TM1 model is 100% bespoke but clearly this isn't the case (nor is it desirable really). Even outside of a partner/implementer paradigm where a standard toolkit (like bedrock) clearly makes sense within the confines of a large corporate there may be multiple TM1 applications running on multiple versions (no IBM it is not always possible to "upgrade to the latest version" and harmonize environments at the drop of a hat) and the lack of such a function really impairs the ability to at least standardize TI processes (unless they are all written to the lowest version denominator) which in turn means more maintenance and more risk.

lotsaram wrote:BUT why for the love of <insert deity of choice> cant we have a TI function that returns a numeric value of the server build? Every other software on the planet has a function to do this and it would be EXTREMELY useful to have such a function in TI to allow for code branching to handle running the same TI under different TM1 versions where newer functions may or may not be supported. Quite a few come to mind beyond this new CellIncrementN function; GetProcessName, CubeClearData, ViewExists, SubsetExists, BatchUpdate, LockOn .....

Agreed. However given that they can't even get this right in the API (I haven't checked whether it's still a bug in 10.1 yet but I'm betting it will be) I hold out little hope. It is, however, something that may need to be added to the "If you could ask the development team a question" thread, if you know what I mean. Given the changes that have occurred over the last few major versions and the number of older versions still in use, I think it's important enough to be.