TM1 Forum

Steve Rowe wrote:I'm actually more interested in the way the cache of calculated results is preserved as I think in the heavy business modelling apps that I prefer to work in this can be more important than locking (lower user base, more calc cycles). My understanding is that pre 9.1 the cache was discarded if something changed at the server level and post 9.0 it was at the "related cubes" level.

Just like the locking...

Not sure if thsi is actually true of course.

Yes, this is my understanding too. I think you and I both commented to each other at the last Cognos day that we had always been under the impression that the caching had been intelligent about discarding results, but then we heard about an improvement that was going to do just that - implying that previously any change dropped all the cached results.

Wish I'd know that for a certain magazine publisher's model.

Martin

Manny Perez' note dated Dec 1st 2005 from the old Center of Excellence has this passage:

Any time an input cell is changed, all stored calculations of the cube where the change is made and all dependent cubes are invalidated.

A cube X is dependent on cube Y, if any of its values are calculated (via rules) using values in Y.

I think he probably knew

NB - this was pre 9.1

That maybe so David but I'm pretty sure that I've had my positon confirmed to people who ought to know too...
Maybe 8 was "related cube structures" and 9 was whole server??
Who knows for sure!

Manny vs who?
Having been present at the most memorable Q&A session at Scottsdale 2006 where Manny said (about what I don't remember - might have been locking!) we'll never do that and Chandra ??? stood up and flatly contradicted him (cue for sharp intake of breath round the room). Chandra was leading the 9.1 team and look how well they did. He didn't last that long. So yes, 9.1+ is very likely to be different.

Steve

Our experience with 9.1 is that with sizable queries and multiple users, performance does suffer, eg 3 users kick off the same query at the same time. First user gets a result in 10 seconds, second in 20, and third in 40 seconds. If caching were operating I would expect all to be about the same. There are reported problems with multi-user response in 9.1.

These are apparently fixed in 9.4, however, we are only just beginning to test 9.4, and as we don't have access to load testing software, we will only know for sure when I can get a number of users to connect on, which won't be for a while yet.

The bit about needing a SaveDataAll after a View creation in TI is interesting. We have had problems with an intermittent error along the lines of 'View Array out of date with dimension'. Clicking the OK button to acknowledge the message crashes the Client, taking Excel with it. This does not go down well with our users. This seems to relate to dimensions being updated, but it doesn't always happen. I have got it to happen by updating the dim in the dim editor, not just TI.

I have actually recently removed SaveDataAll from our GL loads as the SaveDataAll inevitably failed every time, and generated a ProcessError file. Cognos said that this is a spurious error, and that data was being saved. However, the time required to save was quite long and we can do without the overhead of saving. In the worst case, we load the current year from the GL every 10 minutes, so if the server does crash, it is not long before it is up to date again.

It is not a perfect situation. I can only hope that 9.4 is more stable.

Regards


Paul Simon