Forum Discussion
Memory issue after Power BI April update ?
Can you provide any additional details on your PBIX file, data source, I'm assuming SSAS? Direct Query, import, etc.
My model is built on Power BI Desktop, importing data from sql server DB. the model size is not so large :45 Mb.
the visual that trigers the vast memory usage is a pareto% :
TotalNetSalesAllClients := CALCULATE([Total Net Sales],All(DimClients))
AccountRank := RANKX(ALL(DimClients),[Total Net Sales],,DESC)
ParetoValueAccount := SUMX(TOPN([AccountRank], ALL(DimAccounts),[Total Net Sales]),[Total Net Sales])
Pareto%Accounts := DIVIDE([ParetoValueAccount],[TotalNetSalesAllAccounts],BLANK())when writing it I see that there is an inconsistancy with the use of DimClients vs DimAccounts (they are the same one ..).
basicly the original name was DimClients wich I later change to DimAccounts, assuming that the model can handle it-and it did.
I now assume that the upgrade of the model file didn't handle that change correctly and was kinda "searching" a non existing DimClients until memory error.
I am checking this now and will write back if this is correct.
My Bad. the source code was taking from my documentaion and not from the model.
the measures from the model all point to the relevant DimAccounts.
My assumption was wrong..
BIXL How many Total Clients / Accounts do you have?
You may have to do this calculation with a summary table and Calculated Columns instead of Measures?
The ParetoValue and Pareto% Measures iterate over the enitre dataset for EACH Client / Account
I recently had to do this...
Thanks for your detailed alternative solution, but I'm not sure that it's right thing to do bypassing this issue with alternative solution.
first, as for my best knowledge, a measure consumes CPU, not memory (as appose to calculated columns). so it just doesn't make sense to replace a measure with calculated column when having memory problem, right ?
second. this measure worked just fine until April update. so it's beyond me. I sended a frown to Microsoft, hopefully they can figure it out.
thanks again.
