Memory Tuning Parameters to improve Indexing speed

[jcoyne]

Hi Chaps,

 

I love the functionality of Intella but have sleepless nights about the indexing process.

 

I'm generally a bit disappointed that Intella does not use more of the available power of the machine to index quicker. I'm currently conducting the indexing on a dual processor (total 8 cores) with 28 GB RAM and a 512GB SSD for the intella database and 4TB spindle drive for the source files (mix of pst's and general office files).

 

Intella never gets more than 30-50% of the system utilisation and never maxes out the RAM / IO / CPU and I seem to think there must be a way to improve the speed and get it to use more power.

 

I've experimented with the settings in the ini files, but I don't really notice much difference.

 

Has anyone got any recommendations of where to move the memory tuning setting to? or anything else within the system that make the indexing process quicker?

 

Best regards,

 

 

Jason Coyne.

[Andrej]

Hello Jason,

 

Generally speaking, there are three parameters that influence memory allocation and processes usage.

Parameters -Xmx and -Dintella.serviceMaxHeap define the maximum memory allowed for main Intella process and Intella service processes.

Parameter -Dintella.maxExtractorServices defines the maximum amount of Intella service processes that can be created.

 

For optimal indexing speed, we recommend to allocate the available memory in a way that would enable more service processes.

The rule of thumb is that the total memory allocated for Intella (Xmx + serviceMaxHeap * N) should be less than physical RAM size minus 1GB. 

For 16GB RAM, -Xmx7500M, –Dintella.serviceMaxHeap=3750M would work for 2 processes, 6000M/3000M for 3 processes and so on...

 

Please note that having 100% system utilization might not be the best solution as this might be a sign of performance and speed degradation.

 

Kind regards,

Andrej Jesenko