Forum Discussion
Connecting to Analysis Services - Maximum Allowable Memory Limit
Direct Query/Live connection is a good method to keep memory usage down. You need to go after the Import mode datasets. Remember that these datasets use up TWICE the amount of RAM , so a P2 can accommodate a maximum of ONE 25 GB dataset plus a couple stragglers before it falls over.
Use the Performance Metrics template to identify the culprits and in a friendly-aggresive way help the developers of these datasets to get their memory cost down. This is not a one time exercise either - it needs to be ongoing.
It is not always the case that datasets use TWICE the amount of RAM. That would only happen if you were not using Incremental refreshing.
A P2 has a total of 50GB of RAM available, even with a 25GB model there is more than enough capacity for dataset refreshes and other datasets. It all depends on how they are consumed and used.
- lbendlin6 years ago
Super User
Yes, should have mentioned the Incremental refresh part - although it would be interesting to hear Microsoft's perspective on that. In my view incremental refresh only manages the dataset's storage partitions. You still need to load the _entire_ dataset into memory (at least until they implement the promised paging) for rendering. It will certainly help for refreshing (only current partition) but when it comes to the blitting that's when it gets interesting. I would think they need to load the entire new copy into the SKU and then remove the old copy.
- GilbertQ6 years ago
Super User
Hi there
I can confirm that when you use Incremental refreshing it ONLY uses the memory required for those partitions.
It will certainly NOT need to load the entire dataset into memory twice to refresh the dataset with Incremental refreshing.
Whilst a lot of people might think that it needs to copy entire datasets, this is not true and there are a lot of smart things that happen behind the scenes which I have seen happen.