TM1 Forum

Now that 9.4 has been out for a while and there appears to be a significant number of users. I've set up a poll to see what peoples views are on the object-locking model in 9.4. Please feel free to post you comments as well if you want to add some more detail.

Is this working??? 35+ hits and only 1 person has voted, maybe I should have added a don't care option!

Steve Rowe wrote:Is this working??? 35+ hits and only 1 person has voted, maybe I should have added a don't care option!

I think perhaps the problem is the reverse of the last option. The number of people on 9.4 would probably still be a minority. While I've tested 9.4, I have to admit that I haven't looked very closely at the locking model and since I've decided that my next move is likely to be only to 9.0 (given that it's the memory bump in 9.1 that scares the bejebbers out of me) it became academic. Had I been voting, I'd therefore have been looking for an option of "I don't have enough experience of 9.4 to base my answer on". (OK, not STRICTLY true since it's actually "It's not something that I looked at when I was testing 9.4", but near enough.)

Alan Kirk wrote:

Steve Rowe wrote:Is this working??? 35+ hits and only 1 person has voted, maybe I should have added a don't care option!

Had I been voting, I'd therefore have been looking for an option of "I don't have enough experience of 9.4 to base my answer on". (OK, not STRICTLY true since it's actually "It's not something that I looked at when I was testing 9.4", but near enough.)

M'Kay, well... hey, at least it boosted the voter turnout!

Can I ask those viewing but not answering the poll, to perhaps post and say why they are not voting.

I'm just curious if it's because you are not aware of the differences in the object locking model or are not even aware of what a object locking model is.

Obviously of the 100 hits some will be the same people coming back, but if you feel you can't vote in the above please fell free to say why, that way I can make the next poll I do more inclusive.

Thanks for taking the time to respond.
Cheers

I had no voted because my TM1 experience started at 9.1 and the options were geared towards 9.0 and below.

Having gone through 3 of the 9.1 versions and now on 9.4 MR1 FP1 I can tell you that they have made very good strides in the locking models. 9.1 was horrible with concurrent access, but they at least fixed the crashing bugs by about 9.1 SP2. So far 9.4 has been a big update. I have already recieved multiple comments from the user base about increased performance. We have a model that relies on 2 mega cubes that have 12 disparate locations and somewhere between 50 to 75 users on at any given time. During peak processsing I used to receive calls all the time about the system having erratic performance, this was all based on others recalcing at the same time. These have been significantly reduced with 9.4.

Thanks George,
Understood, I kind of blank 9.1 from my mind...

From my point of view 9.1 and 9.4 have the same locking model, it's just that in 9.4 it works much better. It's reassuring that you say 9.4 is significantly faster than 9.1 since my understanding was that 9.1 was slower (a lot?) than 9.0 without a redesign to take advantage of the changes in the locking model. Also since you have the mega cube approach you have not tried to use the changes in the locking model between 9.0 and 9.1 to your advantage.

If a 9.0 style model performs at the same level in 9.0 and 9.4 on the same hardware, then a lot of my fear of 9.4 disappears.
Thanks, anyone else care to comment?

For the purposes of the poll please consider 9.4 and 9.1 the same thing with similar levels of stability.

Cheers,

[quote="Steve Rowe"]Thanks George,
Understood, I kind of blank 9.1 from my mind...
[quote]

I wish I could, we developed everything on the inital release of 9.1 and the first training\demo experience with about 10 users went horribly badly. One of them would see something and the rest would flock like lemmings to it. The moment about 3 of them clicked on it at the same time the server would crash and there I am with my 3 months of TM1 experience and the deer and headlights look.

Good Times.

Hi

We are using 9.1 in production though we are upgrading to 9.4 soon. I have only worked with 9.4 in development so I cannot vouch for its stability, but it hasn't caused any problems so far.

By comparison, we have had a lot of problems with 9.1 If 3 users query the same view I can guarantee that the first user to hit enter will get the fastest response and the last user the slowest, with a differential of a factor of 3. This should not be the case as the view should be cached.

I have had to implement my own locking model to prevent crashes due to concurrent processes in 9.1.

I am pleased to hear that performance in 9.4, is better, particularly multi-user performance.

Having said that, when budgeting alongside day time loads, the performance of 9.1 is still better than 9.0, even if 9.0 was faster for straightforward queries.

The trend with TM1 applications is towards

larger numbers of users
larger amounts of data
mix of reporting and budgeting
day time loads from source systems, and also dimension updates

I believe that all this requires a locking mechanism that goes beyond the server level locking of 9.0. However, you do need to modify your cube design and rules design to get the benefits of the 9.1/4 locking mechanism. That frequently means relying on processes instead of rules. However, the cut off for a process does tend to naturally fit in with the submission point of a budget.

Regards

Paul Simon

I have heard horror stories when it comes to 9.1.

So, 9.4 is definitely an improvement.

Is there anything we can pen down in terms of 'rule of thumb' when designing a model in 9.4?

jonathan.d wrote:I have heard horror stories when it comes to 9.1.

So, 9.4 is definitely an improvement.

Is there anything we can pen down in terms of 'rule of thumb' when designing a model in 9.4?

Have seen a number of posts with similar thoughts - and I completely disagree with going down a completely different model of implementation based on the perceptions of users on this (and other) forums. Until there's some official release or feedback from IBM there's no way of knowing whether the locking model is still being developed or whether there are still issues to be ironed out (I've assumed this is the precise reason why no official information has been released). As another post alluded to, if a future release with further development changes to the locking model is released then you're potentially left with a model optimised ONLY for the current 9.4 release.

belair22 wrote:Have seen a number of posts with similar thoughts - and I completely disagree with going down a completely different model of implementation based on the perceptions of users on this (and other) forums. Until there's some official release or feedback from IBM there's no way of knowing whether the locking model is still being developed or whether there are still issues to be ironed out (I've assumed this is the precise reason why no official information has been released). As another post alluded to, if a future release with further development changes to the locking model is released then you're potentially left with a model optimised ONLY for the current 9.4 release.

Enjoy the wait If Cognos (nee Applix) has taught us anything, the only time you know for sure what's in the next release's feature list is when you see it in the product. Locking and performance improvements have been quoted in every release I've used, and they're in the list discussed for 9.5 too....

Unless there is a significant reason for him to upgrade, jonathan.d might stay on 9.4 for years-- if that's the case, why not optimize for 9.4? At some point, you have to move forward with an implementation. If you wait for the perfect release, you'll be waiting forever.

Unfortunately, I don't have any advice to offer on 9.4. 9.1 has been pretty stable for us, and the features offered by 9.4 aren't enough to make me want to upgrade. I'll re-evaluate when I see 9.5. In our case, an upgrade involves eight servers and a few hundred users, so the improvements would have to be substantial to be worth the effort and risk.

Matt