TM1 Forum

I have a TM1 Server with the latest patches on Windows 2016 Server, TM1 is installed as a HyperV VM. With 128GB of Ram the database is around 80GB or60GB in size. It takes 7 hours to reboot the TM1 Service. How can this be resolved ?

nightwing2099 wrote: ↑Thu Dec 02, 2021 9:51 am I have a TM1 Server with the latest patches on Windows 2016 Server, TM1 is installed as a HyperV VM. With 128GB of Ram the database is around 80GB or60GB in size. It takes 7 hours to reboot the TM1 Service. How can this be resolved ?

This raises more questions than suggestions. Is this only a recent phenomenon? Did the model start in a brisk enough manner before? Have you changed rules / feeders recently? Are you using persistent feeders? Have the feeder files blown out massively in size recently? Can you use multi-threaded startup in your server config file? If so, are you, and what are you using exactly? MTCubeLoad? MaximumCubeLoadThreads? Have you turned on performance monitor? Are any of the memory usage stats ridiculously high?

Assuming that you have a dev box if I were in your shoes I'd probably copy it to dev, dump any persistent feeders, try to use a multithreaded load if possible and see what the result is.

As Alan says, we need a lot more detail to help you. I'd start by dissecting the server message log and seeing which cube rules take the longest to generate feeders for and go from there. I would say 7 hours does seem excessive and likely can be resolved but there is likely a big redesign required. I recently helped a client go from a startup time of 5 hours without persistentfeeders down to 2 minutes, but it took 8 weeks of hard graft and a lot of testing.

Short term solution to reducing the startup time is to enable persisentfeeders but it will only mask the problems with the model.

Start with the cube that takes the longest to calculate feeders for, and have a look at the dimensionality and have a think, is there anything that could be done from consolidations as opposed to feeders. Can intermediate calculation steps be removed to reduce the number of feeders. Also check is the cube is in an optimal dimension order.

If you can share your cube structure and respective dimension sizes perhaps we could offer some tips.