Hello,
I am using the native Power BI connector to connect to CosmosDB to a collection that is partitioned.
I am able to connect to the collection, navigate the complex objects within it (indexes etc) and create my tabular table from the JSON file strucutre.
There is an ISO DateTime that I want to filter on, and I have created a new DateTime column in my query without (much of an) issue - it is working.
My issue is: the model load is extremely slow. When I load the model in either Power BI Desktop or VS (i have made the same model twice in both tools so i can troubleshoot on both sides), the query runs for 30-60 minutes, or more, and is only returning between 10-100 records (depending on how tightly i restrict the datetime filter).
Both CosmosDB and Azure SSAS Tabular (1400 level) are supposed to be extremely fast for data reads, so I am not sure why the data loads/ model processing is taking so long.
My queries are not overly complex, and I am filtering the data on a date (my original data set is 4.5 million rows) and I am pulling back less than 1,000.
I tried adding and index column and filtering on that (pull back only top 2000 rows) to see if it is the datatime filter that is the issue, but using the index column as a filter is just as slow.
Is this a connector issue - i know this connector is in beta, but it seems very slow.
My internet connection is sufficient to handle the queries.
I did notice that my VS version is only 32-bit, but my PBID is 64-bit.
Any thoughts on why the model loads are so slow?
I am NOT using the collection partition value (client) as a filter right now, but there are not many partitions, and I thought that CosmosDB was supposed to hande the partition manamgent/optimization natively - if there are best pracitces, is there a guide somehwere?
Thank you!
