The feature is similar to that of redshift's Distribution style all. This will reduce data transfer between distributions/nodes and consequently improve performance.

Database project from Visual Studio is useful to manage DDLs, schema compare between the project and database, etc. Please create a project template dedicated for DW or support DW development from Database Project.

Polybase should add support for Azure Data Lake. I'm hopeful that will give us an option to do Polybase on top of files which are encrypted at rest.

Support the MERGE sql statement to support management of SCD1 / SCD2

The fact IDENTITIES and SEQUENCE Types are missing makes it very difficult to generate surrogate keys. It seems the only pattern available is to combine ROW_NUMBER() with a select max from table, which is horribly clunky and won't work if parallel treads are loading into the table.

Setting reject rows in Polybase to ignore the header row is a hack. And if you have all string columns your header won't be rejected. Please add an IGNORE_HEADER_ROWS = <numRows> option to Polybase

It would be great to implement Row-Level Security on SQL Data Warehouse.
That would be a great solution for scenarios where (for example) we have clients like PowerBI using data from SQL warehouse, and persons who have limited data access.

Currently Azure DWH is only accessible using the host name of logical SQL Server and all traffic goes through public Internet.

Would be great to have an endpoint with a reserved IP within the Virtual Network to ensure given logical SQL Server is seen as an internal resources within the corporate network.

Let's say you define an external table with:

REJECT_VALUE = 1000,
REJECT_TYPE = value

That means that it can reject and discard 1000 rows without failing the entire command.

Please enhance Polybase to give us visibility into which rows were rejected and why. I would propose you can define an error output file where rejected rows get written. Either that or let us query some DMV to get a copy of the complete row as a string and the error reason.

If a line end is present within a text field (qualified) importing data using Polybase will fail.

Loading this line will work fine:
1;2.1;"zzzz"

Loading this will fail:
1;2.1;"zz
zz"

Code:

CREATE EXTERNAL FILE FORMAT text_file_format_raw
WITH
(
FORMAT_TYPE = DELIMITEDTEXT,
FORMAT_OPTIONS
(
FIELD_TERMINATOR =';',
STRING_DELIMITER = '0x22',
USE_TYPE_DEFAULT = TRUE
)
);

CREATE EXTERNAL TABLE dbo.test_external
(
col1 INT,
col2 DECIMAL(2,1),
col3 NVARCHAR(5)
)
WITH
(
LOCATION = 'test_failing.txt',
DATA_SOURCE = azure_storage,
FILE_FORMAT = text_file_format_raw
);

CREATE TABLE dbo.test
WITH (CLUSTERED COLUMNSTORE INDEX, DISTRIBUTION = ROUND_ROBIN)
AS
SELECT * FROM dbo.test_external;

Support for Entity Framework in Azure SQL Data Warehouse would allow for greater usability in general, but it would also allow for better support of other frameworks/protocols. Specifically, we have a need to serve our data through an OData feed, and Entity Framework's deferred execution would allow us to handle OData's Query Options in a much simpler way.

Enable elastic query for SQL DW - this would allow cross database query in SQL DataWarehouse.

Spatial data shlould be supported by SQL Data Warehouse the same way it is supported by SQL Server.

SQL Data Warehouse scale stepping should be much smaller (10 DWU or even 1 DWU), than 100 DWU!
Please fix that ASAP!
Thank you!

Because we cannot do SELECT @MyVar = Col1, we are forced to utilize the same code to execute the exact same sql, to assign multiple variables.

Able to read the JSON and XML file as part of external system OR provide functionality like Extractor in Data Lake analytics so we can write our own Extractor to read respective data format.

It would be great if you can add support for recursive CTEs for computing hierarchical data