TM1 Forum

Hello all. Just a quick question...

Is it possible to run a 64bit version of TM1 and a 32bit version on the same physical server?

mmsantiago wrote:Hello all. Just a quick question...

Is it possible to run a 64bit version of TM1 and a 32bit version on the same physical server?

Someone else can correct me if I'm wrong about this, but AFAIK you can't run two different versions of TM1 at all on the one box. You used to be able to do it with version 7 and early incarnations of version 8, but at some point... and I don't recall the exact version... you stopped being able to run the install program without uninstalling the earlier version first. Yes, it's possible to hack your way around that restriction (I think that Mr. Hobson's done it once or twice, unless I was dreaming that)... but probably not advisable unless you really, really need to since there's no guarantee that you'll get the same type of performance and/or bugs that you'd get with a "true" install.

Thanks Alan. I pretty much already knew that was the answer, but told the powers that be I would ask the question anyway. Appreciate your response.

You can do loose installs to run different versions of TM1, but I'm not sure why you'd want to - especially in a production environment. I think it'd cause headaches. I am also not sure if "different versions" extends to 32 bit vs 64 bit, that's a bigger difference than 9.1 vs 8.6.

You can, however, run the same version with multiple sets of data, creating lots of servers. So you can have an accounting server, a marketing server, a sales server.... As many as you like on one box. On a site I've just been working on we had half a dozen live and test servers running on the one 64 bit machine. Ran like a dream.

Martin

Michelle,

I actually think you should definitely be able to do this (license-permitting, of course). There are separate installations for the 32-bit and 64-bit versions of TM1, both of which are capable of installing a TM1 Server. Both installations will install to different folders, the 32-bit ending up in the WOW (Windows-on-Windows) location of C:\Program Files (x64)\Cognos\TM1\ folder and the 64-bit ending up in the normal C:\Program Files\Cognos\TM1\ folder, by default.

From there, you'll just want to make sure that both TM1 Server Services are configured to use different ports - you may only be able to run against one Admin Server service (since that is more of a fixed port which cannot be changed), and that could be the only wrinkle in this whole possibility if the 32-bit cannot communicate with the 64-bit Admin Server service, or vice-versa.

As far as running multiple TM1 Versions on the same machine, that is more difficult - the installation packages definitely don't allow you to do this as far as I know, though we have quite often been able to do this by keeping copies of the \bin folder where all TM1 program files exist and creating/running instances of TM1 clients or services off of that on the same machine (I do this on my laptop regularly). I know this type of thing wouldn't be supported, but it has worked pretty well when we need to switch around between different versions and don't want to deal with the always unbelievably large installation package/process.

Hope that helps.

Regards,

Just to add my $0.02, that isn't not really relevant. I just feel the need to comment because it is one of the few things I learned during the conference. The next release of TM1, 9.4, is suppose to be a monster and NOT backwards compatible. This makes me want to since we all know 9.1 was when it came off the bloaks. Once you take the plunge that is it! So keep that in mind for when you are ready to check it out. The reason why it is not backwards compatible escapes me at this point in time. I will post when I remember.

My recollection of what support told me was that you can still do the replace the \bin directory thing (up to 9.1 at least).

AFAIK The only part that requires the complete uninstall is (shudder) TM1 web.

J

Eric,

The next release of TM1, 9.4, is suppose to be a monster and NOT backwards compatible

Could you please be more explicit about this ?
It's those kind of sentences thas give me nightmares

Regards,

Renaud.

9.4 Monster release = Monster issues. I'll get my tin hat then!

Seriously has anybody got any pre-release detail?

Renaud MARTIAL wrote:

Eric wrote:The next release of TM1, 9.4, is suppose to be a monster and NOT backwards compatible

Could you please be more explicit about this ?
It's those kind of sentences thas give me nightmares

IBM/Cognos has said that there is a data file format change between 9.1 and 9.4. When you start up a database with 9.4 for the first time, it will convert the files to the new format. This conversion cannot be reversed and old versions cannot read it, so they're recommending that you save a copy of the model before you upgrade (as if anyone wouldn't do this for any upgrade). So, everyone needs to do an extra-thorough validation on 9.4 before you deploy it to production to ensure you don't encounter any show-stopper issues with your environment. If you uncover something after you've been running it for a week (and users/processes have manipulated the data, invalidating your backup), you're screwed.

Matt

Hope they've invested in a ruddy good quality and testing team, 'cos i sure as hell ain't beta testing it for them!

i sure as hell ain't beta testing it for them

You'll not be upgrading then

When you start up a database with 9.4 for the first time, it will convert the files to the new format. This conversion cannot be reversed and old versions cannot read it

Anyone remember Version 7 <FX runs for the door screaming >

I remember a certain unamed fromer Applix pre-sales consultant doing a demo of 7 when I was at aerospace. It proceeded to collapse like a house of cards around him! (Now we are going back a bit.)

Matt summed it up rather well.

FYI: The change of file format in 9.4 is changing the character set to UNICODE. They apparently need to support some additional character sets, like traditional Chinese, and UNICODE is the best way to support that. There is one advantage for those of us (most) that do not need the additional languages: strings can be longer than 255 characters. In summary:

Pros
- Support for additional languages
- Strings can be longer than 255 characters

Cons
- DB files not backwards compatible. Automatically upgraded on first run (not sure if there is any warning)
- File and memory footprint larger due to the multi-byte character sets which will probably lead to it being a little slower

Be warned and test thoroughly.

Haven't they been supporting Unicode for a while, but as a separate installer? Why force everyone to move to Unicode when most of us currently are quite happy without it? Certainly no way we'll be moving from 9.0 for some time to come

My thoughts exactly Steve

There has been a Kanji version since at least 2003 to my certain knowledge.

But hey - come on we'll get strings > 255 in exchange for the slower performance and non-backward compatibility so we mustn't grumble.

More memory usage??? NNNNOOOOOOOO

I am having to rip bits out of our models since upgrading to 9.1 as it is. I don't need something will take more memory. That will be a disaster. But hey, at least I'll be able type what I think of the change in more than 255 characters! That may not be of that much benefit as most swear words are actully quite short.

Does anyone know whether strings >255 also implies that variables in TI will be able to hold strings >255 chars? This would be good in relation to the ODBCOutput function, where the current limit is really a pain.

Michel