64-bit or 32-bit / physical or virtual

[jimicron]

Hi all,

Our company has experienced various issues with TM1 10.1.1 and I was wondering a couple general things to get an idea of what others in the community are using/doing.

1) Do you use 32-bit TM1 on 64-bit server O/S -or- do you use 64-bit TM1 on 64-bit server O/S
2) Have you had any issues with using 64-bit TM1 on 64-bit server O/S (if you use this)
3) Is your TM1 server on a Physical server or on a Virtual Machine?
4) Any pros or cons to either?

Thanks a lot!

[stephen waters]

Jimicron,
Pretty much all of our TM1 clients (ranging from version 9.0 to 10.1.1) are running 64 Bit TM1 on 64 Bit server hardware and O/S. The exceptions are those on legacy Applix licensing or those with old server boxes.

Quite a few of our clients use virtual machines ( as do we for demos) but a lot still use physical machines. Virtual machines give more flexibility in terms of configuring available resources but are said to have a slight performance overhead.

Can you give more information about
- your problems\issues
- whether this is an upgrade and if so from which version
- hardware and software infrastructure.
If so, you may be able to get some more specific responses.

[declanr]

I am in a similar position to Stephen on this with most clients being 64-bit servers and ranging the same 9 -10 TM1 versions.

Most recently a few clients have been put on 10.1.1 64-bit virtual servers (both new developments and transfer of existing environments) and this has only happened as a result of myself having been pleasantly surprised that I have actually found 10.1.1 to be one of the most reliable versions yet... I was also happy with the stability of 9.5.2 so here's hoping that IBM continue the trend with 10.2...

[foogy]

We're running 10.1.1 as 64bit version on 64bit virtual machines. No problems so far. Physical hardware is probably still faster, but if you operate a TM1 infrastructure for a longer time (and not only provide development and then leave to next customer), you'll also evaluate things like reliability, scalability, snapshots, fail-over and other aspects that a server farm can handle very well. And: memory/HDD can be assigned/unassigned as needed without unmounting the server.

We declined physical hardware from the inital TM1 server setup mainly for reasons of operations effort, reliability and flexibility. This still makes sense if the TM1 virtual machine is the only one running on the VMWare physical machine.

[Gabor]

I did tests against several VMware farms around the world in the last few years.
An optimized physical server (e.g. 4 years old DL 380 G6 /2x X5570 2.93GHz (quadcore) 8 MB SmartCache / 48 GB Memory DDR3-1333) was always at least twice as fast as the virtualized device, regardless of what you do in TM1.
(latest processor features are usually turned off under VM, the processor is more or less mid class, VM itself consumes a little calc power etc.)

As long as you have no performance problems this will not hurt you (TM1 itself is fast), but complex rules, feeding or TI's processing large amount of data can really get benefit of fast physical hardware.

[macsir]

1) Do you use 32-bit TM1 on 64-bit server O/S -or- do you use 64-bit TM1 on 64-bit server O/S
64bit TM1 on 64-bit server O/S
2) Have you had any issues with using 64-bit TM1 on 64-bit server O/S (if you use this)
No
3) Is your TM1 server on a Physical server or on a Virtual Machine?
Phayscial with around 100G Ram
4) Any pros or cons to either?
Better Hardware, Better performance

[jimicron]

Thanks a lot for your input so far everyone. I am sorry that I didn't reply sooner. I forgot to check the "Notify me when a reply is posted" box so just figured no one had replied to this. It's interesting that you all seem happy with 10.1.1 and feels it's stable. It's not been our experience.

So, that makes me think our environment isn't set up correctly and/or some of our TM1 server (where the models are built) are doing things that Tm1 is struggling with. The person that developed on those did so starting with 9.5.1 and then we went to 9.5.2 - both of which we felt were fine and didn't have issues. Where we really started to have issues is when we went to 10.1.1. The two servers that were there in 9.5.2, we can't even open Performance Modeler with those two. However, with other newer servers, we can open PM just fine. The person that built in 9.5.1 & 2, rebuilt everything on 10.1.1, but did a lot of copying/pasting from her objects/rules/ect on 9.5.2 to 10.1.1 because we ran in parallel for a little bit.

We have a couple cases open with IBM (have had for several months) about processes taking 5 hours to run when in 9.5.2 it only took about 1-3 minutes to run the same process. I have built on the same server that you can't open up PM with in a Test environment... and when I delete all the old stuff that was built from 9.5.2 and leave just the stuff I have built, I can then open up PM.

Some specs on our server:

Model: VMware Virtual Platform
OS Version: Windows 2008 R2
Service Pack: Service Pack 1
Processor: Intel Xeon CPU x5675 @ 3.07GHz 3.07 GHz (4 processors)
Installed Memory (RAM): 96 GB
System Type: 64-bit

Let me know if there is anything else that would be helpful to know Also, I have now checked the box for notifications!! haha.

[jimicron]

Hi all,

I had some more questions of curiosity... it just seems like very corner we take, we keep on running into issues or roadblocks, but it doesn't seem like others do (have). I was wondering:

1) What are your datasources when you load data into TM1? Do you use .csv, .xlsx, Oracle and/or SQL servers?
2) Does anyone use ODBC drivers to pull data from a .csv? (advantage to this is you can use SQL to filter using a WHERE clause)
3) How complex are your models? I am assuming with the brilliance of some of you, that your models are quite complex and do some great things!!
4) An error we get quite frequently is "client name does not exist on server" and with an open case with IBM and our own internal IS support, this one has been a struggle
5) Has anyone else experienced much longer processing time when running a process after going to 10.1 from 9.5.2?

Feel free to answer just those that you want to. I realize that not all will apply to all folks. We even have some IBM folks coming this Thursday on site to address some issues as we have had issue after issue after issue since going to 10.1.1.

Thanks again!

[declanr]

1/ Depends entirely on the customer's set-up but have always used ODBC connections (normally standard SQL) but also CSVs. Never had any noticeable problems attributable to the server being virtual. I have some quite serious import TIs working against large ODBC queries and have always found that well written TIs perform extremely well/quickly.

2/ Never felt the need to use an ODBC against a CSV file personally... it would have to be a pretty big CSV file to require using a where clause... I would just use a CSV and stick in a simple

Code: Select all

If ( v1 @<> 'Thingymajig');
ItemSkip;
EndIf;

3/Complexity is in the eye of the beholder.

4/Sounds like an issue with active directory or integrated logon setup - this definitely works perfectly well on physical or virtual boxes so it would be worth you providing more info on your specific setup around that.

5/When IBM brought out 10.1 they set it so that parallel interaction is automatically set to True... try changing it (or adding and setting it) to false in the config file as this can cause a significant change.

[jimicron]

Hi Declan!

In regards to number 2 below... light bulb just went off! We are a very heavy user of Cognos Analyst, which is 32-bit and it is very engrained in me to just ping .csv's (which we pull from SAP) and then just use SQL to filter out what we don't need (or what we need) so end users don't have to manipulate the files that are pulled from SAP. It's such a commonplace as we have around 100 ODBC DSN's that are either using Microsoft Excel Driver or Microsoft Text Driver. Those 100 DSN's ping around 500 load files. But, Analyst is 32-bit. We have played around with uploading .csv to a SQL server but that adds complexity and (IMO) an unncessary layer. So, it's been a struggle, but I hadn't thought of just using "Text" data source and then creating a variable and then using code there! It's not as easy as just writing one SQL syntax statement, however, it's better than addding the complexity of publishing to a SQL server.

I had not thought of that so I really appreciate that!

So, with that in mind, one of the columns I pull in is "Profit Center" and so on the variables tab I called it "ProfitCenter" and then created a variable and called it "vProfitCenter" in which I tried this code:

Code: Select all

vProfitCenter=
     If(SUBST (ProfitCenter,6,1) <> 'A';
ItemSkip;
EndIf;

But it's telling me "Line 2: Syntax error on or before: SUBST (ProfitCenter, logicial expression

In looking at the developer's guide, it appears that I have it correct and then using what you gave me

Any ideas on that? I know this strays away from my original post but your input on this specific certainly provides hope and a ray of sunshine on accomplishign what I want without adding the complexity. Thanks again!

[macsir]

vProfitCenter=
If(SUBST (ProfitCenter,6,1) @<> 'A';
ItemSkip:
EndIf;

[jimicron]

Ha! Macsir.... I was just re-reading and saw that... And was just trying it when I got notified you posted.. haha.

I changed it to:

Code: Select all

vProfitCenter=
If(SUBST (ProfitCenter,6,1) @<> 'A';
ItemSkip;
EndIf;

But, still an error "Line 2: Syntax error on or before: ; missing comma

Thanks!

[macsir]

vProfitCenter=
If(SUBST (ProfitCenter,6,1) @<> 'A');
ItemSkip;
EndIf;


My bad, I didn't notice it either.

[jimicron]

Thanks again!! Whew... I didn't catch that last closing parenthesis either

But, unfortunately, still not working (totally siderailing the original topic, but again... this is going in a new direction that I hadn't thought of and will at least alleviate a little pain, although it does have a learning curve for me).

The code is:

Code: Select all

vProfitCenter=
If(SUBST(ProfitCenter,6,1) @<> 'A');
ItemSkip;
EndIf;

Same error as before. I also have another one that I am doing in this process as I work on this one and that one also has the same error. That dog gone "missing comma"

Thanks again!

[macsir]

A question,
What is vProfitCenter=? Is it connecting to the IF statement?

[macsir]

I am guessing, this is the code you are trying to write?

Code: Select all

vProfitCenter=SUBST(ProfitCenter,6,1);
If( vProfitCenter @<> 'A');
ItemSkip;
EndIf;

[declanr]

Jimicron/Macsir,

I think the issue here is in the "where" rather than the "what".

It appears that Jimicron is using the formula by typing it into the "formula" column of the "variables" tab.
I however had meant (and should have clarified in my original reply) that the "ItemSkip" needs to be used in the "metadata" and/or "data" tabs as the first statement. That way the data tab for that record will start to process but if the formula evaluates to "ItemSkip" it will then jump straight ahead to the next data record.

So assuming that "ProfitCentre" is already a variable you could just use (in the metadata and/or data tabs):

Code: Select all

If(SUBST(ProfitCenter,6,1) @<> 'A');
   ItemSkip;
EndIf;

or you could use the formula column in the variables tab to make:

Code: Select all

vProfitCentre = Subst ( ProfitCentre, 6, 1 );

and use the following in the metadata and/or data tabs:

Code: Select all

If ( vProfitCentre @<> 'A' );
     ItemSkip;
EndIf;

Or you could use Macsir's previous code in the metadata and/or data tab:

Code: Select all

vProfitCenter=
If(SUBST(ProfitCenter,6,1) @<> 'A');
ItemSkip;
EndIf;

Just remember that it needs to be at the top of the tab/s.

HTH

[jimicron]

Thanks all!

Correct. I was trying to do this on the Variables tab. When I pull in my data from the .csv file, it has a column called "Profit Ctr" and then on the Variables tab, I renamed it to "ProfitCenter" - so the one that is "ProfitCenter" is my original data. The one that is "vProfitCenter" is the one with the forumula that points to the original "ProfitCenter."

I'd actually really like to do it on the Variables tab (for different reasons). I write SQL against .csv's a TON in our Analyst models so I would like to stay similar to that. But also, to clarify, in my .csv, it has all profit centers... but when I pull in my data into TM1, I want all profit centers BUT the ones where it has an "A" as the sixth character. Otherwise, I want the full 10 digit profit center - just not the ones where "A" is the sixth digit.

So, in the screenshot below, the code that is in the "Forumula" (one outlined in red) is:

Code: Select all

vProfitCenter=
If(SUBST(ProfitCenter,6,1) @<> 'A');
ItemSkip;
EndIf;

Again, greatly appreciate your help!

[jimicron]

I think I just got it

I used this code:

Code: Select all

vProfitCenter=
If(SUBST(ProfitCenter,6,1) @<> 'A', ProfitCenter,'');

Is this okay and acceptable?

[jimicron]

Hmmm. Maybe I see why you want to do it in your Advanced tabs? I ran the process but I got an error box versus 'completed successfully' and it's showing the ones I am trying to exclude... I don't like end users to get this error box on a regular basis because then they get immune to it and when there truly is an error, then they ignore it. Do you guys agree? (Sorry if this seems to basic to you - definitely still learning and I appreciate your patience!)