Question on TM1 replication

[sureerat]

Hi all,

My model that I'm implementing is complicate and have big size. I'm finding way to improve performance by reduce unnecessary data stored in source server. I think to create another instance to keep inactive data from source server. I used replication feature in TM1 but I have some question.
If I have replicated data from source to target already and I decide to delete replicated data from source. I found data in target server will be deleted too.

How do I order replicate to copy only new added data and not modify data that was replicated before.

Anyone have other idea for this?

Thank you for your all opinions.

[rkaif]

You can adopt a few approaches here. If you dont want your data to be deleted from the Target, then you can Disable the Chore after initial Replication has been completed. This will stop the data to be deleted from the Target.

Otherwise, you can create a replica of the cube which you are trying to replicate on the existing server. This replica will act as an Archive cube and will have all the data of the prior years. Set this replica cube to 'Load on Demand' so that it does not take the memory all the time. While in the original cube you can keep the data of existing year and couple of prior years (depending upon the business requirement). In this way you dont have to maintain two separate servers.

[sureerat]

Hi rkaif,

Thank you for your reply. I have questions to ask suggestion from you.

1. From your suggestion, If I disable chore as you said. It's just prevent TM1 replicate update data to target by automatic right?
But when I need to purge inactive data from original to replica cube again. Does it should be make data loss like automatic replicate?

2. "Load on demand" that you said it's mean use TI order TM1 unload replica cube from memory right?
If yes, this process will require user to manual execute TI to unload cube after server has started right?
My concerned is TM1 server loading time. If I have replica cube within existing server, these cubes will be loaded on server start and unloaded by TI after server already run. I fear server will consume much loading time. Does it has way to config TM1 does hasn't to load replica cubes?

Please forgive me if I miss understand your suggestion.

Thank you.

[Kyro]

Replication simply opens the transaction log from both source and target servers and applies the makes the changes that are logged. So why not simply turn logging off before removing the redundant data from your production server, turn it on again afterwards and the changes wont be replicated.

My actual suggestion would be to export the whole cube to csv then delete the redundant years data. its easy enough to reload it if anyone wants it again later.

Have you tried dimension re-ordering?

[mattgoff]

Is your inactive data within the same cube (e.g. old versions) as active data or are there special cubes that are not often accessed? Have you tested to see if eliminating data makes the server faster? I'm skeptical that you're going to see much performance improvement that couldn't be found elsewhere (specifically by tuning rules and feeders). If you're set on removing data, I would probably recommend two isolated servers and transfer data via TI/ODBO over replication. Of course, I try to dissuade people from replication unless absolutely necessary, so grain of salt, etc.

Matt