TM1 Forum

Hello,


we are at the moment in an update process from tm1 9.5.2 to 10.2.2.
Since we are using tm1web in WAN we are checking performance and behavior changes from tm1web 10.2.2 vs. 9.5.2 (using IIS with http compression).

With default parameters it looks like that we mostly have less traffic at initial page load in 10.2.2 as in 9.5.2.

With increased paging parameters WebsheetRowThreshold, WebsheetColumnThreshold, WebsheetRowBufferSize, WebsheetColumnBufferSize it looks like we have more initial traffic with 10.2.2.


Therefore it would be a lot of work to test the beavior of tm1web, with different settings, different websheets, different bandwith and latency and different usages, i would like to hear about other experiences. I think this may be interesting for many TM1 users.

What are your experiences with tm1web 10.2 and greater especially in WAN?
What are your experiences with tm1web paging parameters? How does tm1web paging work out on limited WANs? Any recommendations for settings of this parameters?


Kindly regards
Peter

CEL3K5 wrote:tm1web 10.2.2 vs. 9.5.2 (using IIS with http compression).

You're not using IIS with 10.2.2.

Sorry. We are using only TM1 9.5.2 with IIS, 10.2.2 with Tomcat. I thought it would be obvisous that only 9.5.2 is with IIS.

CEL3K5 wrote:we are at the moment in an update process from tm1 9.5.2 to 10.2.2.
Since we are using tm1web in WAN we are checking performance and behavior changes from tm1web 10.2.2 vs. 9.5.2 (using IIS with http compression).

IMO you are comparing apples and oranges due to the change in stack (IIS vs Tomcat). Also, the cacheing parameters available in 10.2.2 weren't available in some earlier versions so this also prevents it being a straight contest.

CEL3K5 wrote:With increased paging parameters WebsheetRowThreshold, WebsheetColumnThreshold, WebsheetRowBufferSize, WebsheetColumnBufferSize it looks like we have more initial traffic with 10.2.2.

When I looked at this earlier in the year, it appears that these parameters should generally speed things up. However, once you've set them up in the config it is rather limiting because some 'legacy' websheets may be larger than the thresholds set in the parameters. Our sheets that were larger than the thresholds suffered poor loading and paging performancce. If you set the parameters to something approaching the maximum size of your current set of websheets then my casual observation was that the general performance degraded noticeably. In fact, 10.2.2 TM1 Web seemed a bit on the slow side based on the batch of initial tests with older websheets. Perhaps if you are starting from scratch and follow a revised set of best practices then it is going to be quicker than the IIS and .NET-based implementation; but if you have a lot of existing sheets then you probably should just keep testing with different configs, incrementing the thresholds as you go until you reach a happy medium (or equality of misery, as some might say). Time pressures prevented me from doing this, but it seems a sensible way forward. Before you go for the tests on latency etc you might want to do simple testing on localhost on 9.5.2 vs 10.2.2 just to get a baseline.

CEL3K5 wrote:Sorry. We are using only TM1 9.5.2 with IIS, 10.2.2 with Tomcat. I thought it would be obvisous that only 9.5.2 is with IIS.

Sorry as well but nothing is obvious on this forum. As to performance testing, I just don't see the purpose. It's not like you can do anything about it since this isn't a beta testing version. This is what you've got to work with. As long as you follow best practices with websheets, like not publishing a bunch of superfluous rows and not leaving DBRWs that will result in KEY*ERR when opening then you should be OK. I have not done testing but my experience has been that the performance of the two versions of Web you are talking about are not that much different.