TM1 Forum

Hi All,

When I am trying to open a websheet in TM1 web,for the first time it takes 2 mins to open.After that with the web sheet open, when I am trying to open the respective cube from TM1 architect it takes around 3 mins. Can anyone please let me know what are the possible reasons for the cube taking so much time to load. Once it opens from next time its loading instantly.

Please advice.

Regards,
Rama Mishra

The discrepancy between the first "slow" time and subsequent "faster" times is caching; 2 types of caching the main one being all of TM1's rule calculations and consolidations and the other less significant but still important is the websheet formatting etc being rendered.

You will need to look at standard efficiency stuff of which there are many topics on here you can study or you could provide more detail about your cube structure and rulea for help.
Most likely though you will find it is something to do with the rules being calculated; possibly over-feeding etc.

The websheets will be easy to make more efficient if there are issues; rules and feeders can be harder as they may warrant a whole redesign. Especially as it can be due to a chain of rules that go through many cubes.
Or you could possibly just be trying to view a very big accumulation of data.

Thanks a lot Declanr for the helpful reply.

I will go through the documents related to TM1 system performance monitoring. Please provide if you have any docs.

But I am surprised with one thing that - 'with the websheet open(this means the cube is loaded in memory) , the cube in TM1 Architect still takes 3 mins to load'. Could you please give some idea on this? Please let me know if you need any details.

Thanks and Regards,
Rama Mishra

RMishra wrote:But I am surprised with one thing that - 'with the websheet open(this means the cube is loaded in memory) , the cube in TM1 Architect still takes 3 mins to load'.

This points to performance issues potentially resulting from a bad cube design.

Does the cube have rules? If so, is there a SKIPCHECK statement?

How big is the cube in terms of its number of dimensions, the *.cub file size on disk and how much memory it consumes (you can check that last value by looking at the properties pane of Cubes in Server Explorer)?

Hi,

Yes the cube contails rules.
Rules has skipcheck statement.
It has 9 dimensions.
.cub file size in disk is 106mb.
It consumes 1.40mb memory.

Thanks,
Rama

RMishra wrote:.cub file size in disk is 106mb.
It consumes 1.40mb memory.

I find it hard to believe unless your cube is set to load on demand and is currently unloaded (or perhaps not all of it gets loaded into memory because one or more dimensions lack required elements). The size of the cube in RAM is typically at least a few times bigger than on disk. Could you check in the }CubeProperties if your cube is set to load on demand? If yes, that would explain why you need to wait a few minutes for initial results. However, it would not explain why you need to wait more than once in any given server session.

RMishra wrote:But I am surprised with one thing that - 'with the websheet open(this means the cube is loaded in memory) , the cube in TM1 Architect still takes 3 mins to load'. Could you please give some idea on this? Please let me know if you need any details.

This is likely because TM1 only caches what has been asked for by a client. It doesn't cache the whole friggin cube at once. If your web sheet does not pull back the same slice of data from the cube that your default view in Architect does then the default view data has not been cached yet and will have to be calculated instead of just retrieved from memory.

Thanks for the replies.

Below information I got from }StatsbuCube control cube
Total memory used by the cube is: 7423915008
Memory used for calculation: 5964292864
Memory used for feeder: 6442450944
Memory used for input data: 981464064
Number of populated numeric cells: 2753076
Number of fed cells: 207558116

Loadondemand is selected 'No' in ?cubeproperties cube.

Regards,
Rama

RMishra wrote:Below information I got from }StatsbuCube control cube
Total memory used by the cube is: 7423915008
Memory used for calculation: 5964292864
Memory used for feeder: 6442450944
Memory used for input data: 981464064
Number of populated numeric cells: 2753076
Number of fed cells: 207558116

From these stats I can tell you that in all likelihood:

a) Based on your feeder-to-input-data ratio (75 fed cells for each populated cell) you probably have overfeeding, which would negatively affect cube performance and cube size.

b) Based on your average populated cell size (356 bytes) you have a sub-optimal dimension order, which would negatively affect performance and cube size.