Need information regarding memory and disk

[Sindhukarthik]

Hi Qlik,

Need more information regarding memory and disk tuning in replicate.

[SushilKumar]

Hello Team,

As of Now there is no Such recommendation in term on memory and Disk Tunning. As it requires analysis of Customer Environmnet. Mainly the kind of data and Data Volume and participating endpoint (Databases)

However, The Default Setting which are put in place are able to cater moderate Replication Load.

You may find some infrastructure related recommendations on below link.

Recommended hardware configuration #Recommended hardware configuration | Qlik Replicate Help

Regards,

Sushil Kumar

[Alan_Wang]

Hi @Sindhukarthik 

We recommend keeping CPU usage under 50% and memory usage under 80%. If any incoming changes are seen moving to the 'On Disk' section of the task, then it would help to allocate more memory to help the task process the incoming changes.

This option in the Change Processing Tuning section can be increased to a high value to maximize the use of memory instead of having transactions stored on the disk. 60000 seconds is a good number to ensure that the allocated memory is used before any changes are moved to the disk.

Transaction Offload Tuning -> Transactions duration exceeds (seconds): 

The value set in the option below should be above the memory value that you have set in the 'Batch Tuning' section

Total transactions memory size exceeds (MB):

Batch size tuning will be dependent on endpoints and your environment. It will require trial and error to find the most efficient settings. If you would like Qlik to assist your team with in-depth performance tuning, our Professional Services team can be engaged as a consulting service for further expertise. Please refer to this article if you would like to get in touch with our Professional Services team.

https://community.qlik.com/t5/Official-Support-Articles/How-and-when-to-contact-Qlik-s-Professional-...

[Barb_Fill21]

Hi @Sindhukarthik 

@Alan_Wang  and @SushilKumar  are both correct.  My perspective is that in order to avoid latencies, its best to keep the transactions in memory until they can be written to the target. So things like changes being made to the source should be committed and committed often.  Then when Replicate receives the changes, and if it is high volume, with lots of tasks etc., then make sure your Server has plenty of RAM 64gig+ to accommodate many tasks with wide tables and high volumes. If you do the tuning to keep transactions in memory, then you will need to use more RAM. Each task can behave differently which may need special tuning, and may require more RAM just for one task for example. 

Sincerely

Barb