Enables multi-threaded loading of individual cubes.
Parameter type: optional, dynamic
Default value: F (disabled)
MTCubeLoad uses the MTQ multi-threaded framework to achieve improved speeds when compared to the previous approach using MaximumCubeLoadThreads. Configuration settings MTCubeLoad.MinFileSize, MTCubeLoad.Weight, MTCubeLoad.UseBookmarkFiles, and IndexStoreDirectory are used to optimize performance.
MTCubeLoad also eliminates the risk of changing feeder generation sequences. When MTCubeLoad is enabled, the MaximumCubeLoadThreads configuration option is ignored. Since MTCubeLoad leverages the MTQ framework, it still relies on the MTQ configuration to set the number of concurrent threads.
TM1 server administrators can use the PreallocatedMemory.Size, PreallocatedMemory.ThreadNumber, and PreallocatedMemory.BeforeLoad settings to configure preallocation memory and optimize scale-up and performance results. Allocating memory for a TM1 server can help avoid contention effects related to varying operating system memory allocation. For best results, consider the amount of RAM consumed by TM1 server to configure the preallocation memory settings.
The following Planning Analytics database configuration illustrates the configuration of multi-threaded cube loading:
MTCubeLoad=T MTQ=All MTCubeLoad.UseBookmarkFiles=T IndexStoreDirectory=c:\Cubes\CubeFolder\indexStore # 30GB of RAM to preallocate PreallocatedMemory.Size=30000 # Run preallocation in parallel to cube cell/feeder loading PreallocatedMemory.BeforeLoad=F # Window 2012 patches as of Dec 2016 worked most efficiently with a single thread PreallocatedMemory.ThreadNumber=1 # Disable TM1 performance counters to speed up MTQ. PerfMonIsActive=F
To enable logging for multi-threaded loading of individual cubes, enter the following lines in the tm1s-log.properties file:
- Captures individual timing of TM1® Server loading stages: DeSerializeDimensions, DeSerializeAttributeCubes, DeSerializeRegularCubes, GenerateServerSecurity.
- Captures cube loading time, separately for .cub and .feeder files.