Forum Discussion
Power Query engine performance issue
I had to issue 3 REST API calls (over a corporate fiber link on a Dell workstation with 64GB of RAM, and one of the latest Intel CPUs, with 3 SSDs, etc, so the horsepower should be there):
1_get the auth token
2_get 1,000 ID records
3_get the data linked to these IDs, which is a 1,000 x 96 dataset (rows x columns)
It imported very slowly. Looking at the diagnostis, the PQ engine seems to complete each API calls in ms, so it's got to be the engine processing all the ETL transformations that must be slow. To give you an idea, it takes almost 2 hours to ingest this dataset, and that's before clicking Close&Apply! What is puzzling, the total number of transformations is really small too, with a final pivot as the last step.
I am surprised that a 1,000 x 96 would incur such a performance penalty. It's scary to think what would happen if the dataset to ingest had 10s of 1,000, or 100s of 1,000 rows, for ex. Power Query would be running for more than a week!
It is a lot faster to ingest data from a DB than a REST API. So Power Automate to dump the raw data into a DB, then importing from that DB might be the speedier and preferred solution.
Nonetheless, does anyone have any idea why a 1,000 x 96 dataset would import this slowly from REST API calls that complete in ms?
Without seeing your query code, it's a bit difficult to diagnose for sure, but given the size of the data, I'd recommend buffering the table after loading and before any transformations like pivoting. Scroll down to the Buffering section in this article for some more detail about the Table.Buffer function.
Other posts related to buffering:
https://community.powerbi.com/t5/Desktop/Using-Table-Buffer/td-p/1535407
ImkeF has a nice list of various recommendations for improving query performance.
I'd consider anything that can comfortably fit into your RAM to not be a large table.
As far as buffering, I suggest buffering right after the API stuff so that any further transformations don't attempt to trigger API calls again. You could do this at this step:
rt_table = Table.Buffer(Table.AddColumn(type_change0, "RT_DATA", each getRTData([handleID], token), type record)),If this table fits in memory nicely, then any subsequent basic transformations should be pretty fast.
Hi AlexisOlson - I am wary of this suggestion because Table.Buffer may not like nested Tables or Binary objects. I have seen this happen with Dataflows.
Element115 - it will help to buffer before running the very expensive Table.Pivot function.I glad it starting to help. One thing that can slow the performance is the API Throttling Limits. You should check how many send and receives you can make per second or minute.
There are two other things would try, but this will depend on whether original data and API data must fully updated each time.
- try an incremental load using the LastReportTime to avoid reload all records each week.
- if ID_External and Handle_ID are not unique then try running them in a distinct batch to avoid using the API to call the same data more than once.
Buffering loads the table to memory as it exists at that particular step. Deciding when and where to do this is more of an experimental art rather than an exact set of rules to follow, especially without a deep understanding of exactly how the query optimization engine works.
Just because you have some version of the table loaded into memory doesn't mean that there's never a need buffer again after that point. If you do expensive calculations or extensive transformations on a table, sometimes it's worth buffering those intermediate results before doing any further steps so that you have those calculations/transformations stored in a format that can be referenced efficiently.
A rather extreme example is this function I wrote here. As ImkeF points out, after I've done some initial transformations to set up some chunks to loop through, buffering them to memory helps a lot since it's doing nested iterations on those chunks. Only buffering the initial input wouldn't be nearly as fast.
It's possible that buffering both before and after the pivot is the fastest but that's something that needs to be tested in your specific situation. Don't go too crazy with buffers though. Take them out anywhere they don't help.
Element115 -
For 1 - it depends, as AlexisOlson says it is experimental. However there is one firm rule that you should follow. If you are connecting to foldable datasource like a database, don't buffer until after Query Folding breaks. Buffering at the very start would break folding and effectively load the entire table to temporary memory.
For 2 - Firstly, I believe Pivot is the more expensive transformation. Second, I would want Power Query to full complete all the steps before starting the Pivot transformation. Hence, my strategy would be to place it before Pivot. However, testing might show that there is very little impact from using either approach.
29 Replies
Without seeing your query code, it's a bit difficult to diagnose for sure, but given the size of the data, I'd recommend buffering the table after loading and before any transformations like pivoting. Scroll down to the Buffering section in this article for some more detail about the Table.Buffer function.
Other posts related to buffering:
https://community.powerbi.com/t5/Desktop/Using-Table-Buffer/td-p/1535407
ImkeF has a nice list of various recommendations for improving query performance.
Buffering with Table.Buffer or Table.StopFolding?
I already tried both, and the latter gives a slight perf improvement, but nothing earth shattering. The pivot step and the type casting is holding up everything. A PQ Engine issue?
And is a table with, say, 100K data points considered a large table? In my book, that's not large.
Here is the code in full--comment at the very end explain the last step, a type cast, that caused the error, which went away once the type cast step was removed. Now: I am casting from text to logical using text values of 'False' and 'True'. Used to work in the past if I remember correctly.
let source1 = Sql.Databases("DB_NAME"), source2 = source1{ [Name="###"] }[Data], dbo_vOLC = source2{ [Schema="dbo", Item="vOLC"] }[Data], #"Changed Type" = Table.TransformColumnTypes(dbo_vOLC,{ {"ID_External", Int64.Type} }), #"Renamed Columns" = Table.RenameColumns(#"Changed Type", { {"FK_ID_###", "ID_###"} } ), remove_cols = Table.StopFolding( //Table.Buffer( // Table.RemoveColumns( #"Renamed Columns", { "ID" , "ID_OLC" , "Longitude" , "Latitude" , "Is_remotely_managed" , "Model" , "FirmwareVersion" , "HardwareAddress" , "ControlSystem" , "InstallationDate" , "CommunicationStatus" , "NumberOfMeteringChannels" , "CLO_Enabled" , "LastReportTime" , "Is_metered" , "UserSwitchType" , "FactorySwitchType" } ) ), project_id = fnProjectID(), getToken = () => let URL = "https://api.###.com", body = "app_key=###&app_secret=###&service=###&scope=###", response = Web.Contents( URL, [ RelativePath = "oauth/accesstoken" , Headers = [ #"Authorization"="Basic ###" , #"Content-Type"="application/x-www-form-urlencoded" ] , IsRetry = true , Content = Text.ToBinary(body, BinaryEncoding.Base64) ] ), jsonResponse = try Json.Document(response), token = if jsonResponse[HasError] then error jsonResponse[Error][Message] else jsonResponse[Value][token] in token, getHandleID = (external_ID as number, token as text) => let URL = "https://api.###.com", relative_path = "###" & project_id & "###" & Number.ToText(external_ID) & "###", headers = [ #"Authorization"="Bearer " & token , #"Content-Type" = "application/json" ], external_ID_txt = try Number.ToText(external_ID), http_resp = if external_ID_txt[HasError] then error "cannot convert external_ID from number to text" else try Web.Contents( URL , [ RelativePath = relative_path , Headers = headers , IsRetry = true ] ), response = if http_resp[HasError] then error http_resp[Error][Message] else try Json.Document(http_resp[Value]) in if response[HasError] then response[Error][Message] else try response[Value]{0}[handleId] otherwise response[Value][Message], token = getToken(), add_handleID = Table.AddColumn(remove_cols, "handleID", each getHandleID([ID_External], token), type text), type_change0 = Table.TransformColumnTypes( add_handleID, { {"handleID", type text} } ), getRTData = (handle_ID as nullable text, token as text) as record => let URL = "https://api.###.com", relative_path = "###" & project_id & "###", handleID_num = try Number.From(handle_ID), rt_data = if handleID_num[HasError] then [ERROR="Number.From(handle_ID) cannot convert; hanleID=" & handle_ID] else let headers = [ #"Authorization" = "Bearer " & token , #"Connection" = "keep-alive" ], query = [handleId = handle_ID], http_resp = try Web.Contents( URL , [ RelativePath = relative_path , Headers = headers , IsRetry = true , Query = query ] ) , data = if not http_resp[HasError] then try Json.Document(http_resp[Value]) else error http_resp[Error][Message], result = if data[HasError] then [ERROR=data[Error][Message]] else if Value.Is(data[Value], type list) and List.IsEmpty(data[Value]) then [ERROR="no data: wait for the next server refresh"] // the API returns a list, ie [] JSON array, when data is available else if Value.Is(data[Value], type list) and not List.IsEmpty(data[Value]) then data[Value]{0} // the API returns a record, ie {} JSON object, only if there was an error else if Value.Is(data[Value], type record) then [ERROR=data[Value][Message]] else null in result in // returns a record of real-time values or an error msg record rt_data, rt_table = Table.AddColumn(type_change0, "RT_DATA", each getRTData([handleID], token), type record), expand_data = Table.ExpandRecordColumn(rt_table, "RT_DATA", {"status", "DateTime", "properties"}, {"status", "DateTime", "properties"}), expand_prop_list = Table.ExpandListColumn(expand_data, "properties"), #"Expanded properties" = Table.ExpandRecordColumn(expand_prop_list, "properties", {"Key", "Value", "Unit"}, {"Key", "Value", "Unit"}), #"Removed Columns1" = Table.RemoveColumns(#"Expanded properties",{"Unit"}), #"Filtered Rows" = Table.SelectRows(#"Removed Columns1", each ([Key] <> null)), type_change1 = Table.TransformColumnTypes(#"Filtered Rows",{{"Key", type text}, {"Value", type text}}), #"Pivoted Column" = Table.Pivot(type_change1, List.Distinct(type_change1[Key]), "Key", "Value"), #"Replaced Value" = Table.ReplaceValue(#"Pivoted Column","No","False",Replacer.ReplaceText,{"CLO Enabled"}), #"Replaced Value1" = Table.ReplaceValue(#"Replaced Value","Yes","True",Replacer.ReplaceText,{"CLO Enabled"}) // and here was the step to cast this [CLO Enabled] column to type logical, after the addition of which I got the OLE DB type mismatch error in #"Replaced Value1"Element115 - this is how I would approach this to see if it improves performance.
remove_cols - insert a Table.Buffer after this step. Please check this column is folding as the previous change of data type might break.getToken - this can be moved to separate query/function. It is not necessary to include in this Query. There is potential issue with this step because we want to get one token for all subseqent requests. There is a risk that you are generating a token for each query. The latency for each token request will add up!. Text.Buffer would be helpful, but sadly doesn't exist. You could try to return the Json.Document with the getToken function (i.e return the jsonResponse step instead of text). In the main query, just add the following: BufferToken = Binary.Buffer( getToken() ), Token = Json.Document( BufferToken )[Value][token], This should force Power Query to get token once.add_handleID - this steps refers to [ID_External] in the remove_cols, is this correct? This only the ID column. Or does the project_id step do something?type_change0 - this step is unnecessary, but this is a good time for the second Table.Buffer so you know have the SQL table and the HandleID.getHandleID - this could be moved to separate query/funciton. I am not sure about the complexity of the error handling. This approach may be chatty. I.e. to couple the steps it needs to execute the web call more than once.#"Expanded properties" - add the 3rd buffer step after this step, so all the previous steps are completed before the Pivot. Note the #"Remove Column1" can be avoided by excluding the Unit column.
