MDX效率，来自同一层次结构的度量

Question

I am fairly advanced at VBA / SQL .

I am working on an extract from an SSAS cube in Excel to replace a pivot table. I don't have access to the SQL server or SSMS . For the first time I am extracting this information directly from the cube using MDX .

Much of the reason for this is to remove manual cube pivot table updates where issues have occurred with individuals overtyping the pivot table and / or not extending additional formulas down to match it.

I have produced a query which works for me, it is provided below (as loaded into the strMdx string in VBA ):

    strMDX = "WITH MEMBER [Measures].[Sector] as ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Sector]).NAME "
    strMDX = strMDX & "MEMBER [Measures].[Industry] as ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Industry]).NAME "
    strMDX = strMDX & "MEMBER [Measures].[Sub Industry] as ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Sub Industry]).NAME "
    strMDX = strMDX & "MEMBER [Measures].[Account] as ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Account]).NAME "
    strMDX = strMDX & "MEMBER [Measures].[Sub Account] as ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Sub Account]).NAME "
    strMDX = strMDX & "MEMBER [Measures].[Level] as ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Level]).NAME "
    strMDX = strMDX & "SELECT NON EMPTY ( {[Measures].[Sector], [Measures].[Industry], [Measures].[Sub Industry], [Measures].[Account], [Measures].[Sub Account], [Measures].[Level], "
    strMDX = strMDX & "[Measures].[iCounts] }) ON 0, "
    strMDX = strMDX & "NON EMPTY( [Date].[DayYear].[Year],[Date].[MonthYear].[MonthName]) ON 1, "
    strMDX = strMDX & "NON EMPTY( DESCENDANTS([Sector].[IndH],,LEAVES)) HAVING [Measures].[iCounts] >0 ON 2 FROM ( SELECT ( "
    strMDX = strMDX & "[Date].[MonthYear].[Date].&[" & Format(strStartDte, "yyyymmdd") & "]&[" & Format(strStartDte, "dd mmm yyyy") & "] : "
    strMDX = strMDX & "[Date].[MonthYear].[Date].&[" & Format(strEndDte, "yyyymmdd") & "]&[" & Format(strEndDte, "dd mmm yyyy") & "] ) "
    strMDX = strMDX & "ON 0 FROM [IndustryCube])"

This is the MDX :

WITH 
MEMBER [MeASures].[Sector] AS 
    ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Sector]).NAME 
MEMBER [MeASures].[Industry] AS 
    ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Industry]).NAME 
MEMBER [MeASures].[Sub Industry] AS 
    ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Sub Industry]).NAME 
MEMBER [MeASures].[Account] AS 
    ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Account]).NAME 
MEMBER [MeASures].[Sub Account] AS 
    ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Sub Account]).NAME 
MEMBER [MeASures].[Level] AS 
    ANCESTOR([Sector].[IndH].CURRENTMEMBER,[Sector].[IndH].[Level]).NAME 
SELECT 
NON EMPTY 
( 
    { [MeASures].[Sector]
    , [MeASures].[Industry]
    , [MeASures].[Sub Industry]
    , [MeASures].[Account]
    , [MeASures].[Sub Account]
    , [MeASures].[Level], 
      [MeASures].[iCounts] }
) ON 0, 
NON EMPTY 
    (
     [Date].[DayYear].[Year]
    ,[Date].[MonthYear].[MonthName]) ON 1, 
NON EMPTY
    (DESCENDANTS([Sector].[IndH],,LEAVES)) 
HAVING [MeASures].[iCounts] > 0 ON 2 
FROM ( 
    SELECT ( 
        [Date].[MonthYear].[Date].&[ & Format(strStartDte, yyyymmdd) ]&[ & Format(strStartDte, dd mmm yyyy) ] : 
        [Date].[MonthYear].[Date].&[ & Format(strEndDte, yyyymmdd) ]&[ & Format(strEndDte, dd mmm yyyy) ] ) 
        ON 0 
    FROM [IndustryCube]
)

There are two key aspects of this query to note:

1. It is a 3 axis query - which works perfectly for me in writing it using loops in VBA after.
2. I need to split out and use multiple levels from the same Hierarchy ( [Sector].[IndH] ), therefore I need to flatten the hierarchy. I have done this using Descendants then by defining each level as it's own measure. This does work to flatten out the hierarchy but I believe is a bit of a workaround.

Other than the two points above I think it's a pretty standard MDX statement.

The key issue I have is that if I run this for say, a month or two, the query runs perfectly well, although it takes just over a minute. If I run it for over 5 years (which I need to) then it takes a significant amount of time and I have never got it to run successfully.

Is there anything I can do to this query to improve it from a performance perspective?

I have a suspicion the issue is to do with flattening the hierarchy. Unfortunately there is no way around this as the cube has not been set up with separate dimensions outside the hierarchy, or another hierarchy that contains the fields I need that I can pull from.

Answer 1

I'd be tempted to move it into a couple of mdx scripts within your vba , to try to make things a little more readable more than anything else.

I'm guessing that in vba you're looping through the members that are in the third axis. If so take that out of your big script and create a small mdx script that returns these members. Maybe like:

SELECT 
NON EMPTY 
    [MeASures].[iCounts] ON 0,  
NON EMPTY
    DESCENDANTS([Sector].[IndH],,LEAVES)  ON 1 
FROM ( 
    SELECT ( 
        [Date].[MonthYear].[Date].&[ & Format(strStartDte, yyyymmdd) ]&[ & Format(strStartDte, dd mmm yyyy) ] : 
        [Date].[MonthYear].[Date].&[ & Format(strEndDte, yyyymmdd) ]&[ & Format(strEndDte, dd mmm yyyy) ] ) 
        ON 0 
    FROM [IndustryCube]
)

The outer loop could iterate over the members returned from the above and feeds each into the following:

SELECT 
NON EMPTY 
    [MeASures].[iCounts]  ON 0, 
NON EMPTY 
    [Date].[DayYear].[Year]
    *[Date].[MonthYear].[MonthName]
    *[Sector].[IndH].[Sector]
    *[Sector].[IndH].[Industry]
    *[Sector].[IndH].[Sub Industry]
    *[Sector].[IndH].[Account]
    *[Sector].[IndH].[Sub Account]
    *[Sector].[IndH].[Level] ON 1 
FROM 
( 
    SELECT  
        [Date].[MonthYear].[Date].&[ & Format(strStartDte, yyyymmdd) ]&[ & Format(strStartDte, dd mmm yyyy) ] : 
        [Date].[MonthYear].[Date].&[ & Format(strEndDte, yyyymmdd) ]&[ & Format(strEndDte, dd mmm yyyy) ]  
        ON 0, 
        [Sector].[IndH].&[ ....strFromOuterLoopHere...  ON 1
    FROM [IndustryCube]
);

---

One thing to be wary about when using mdx are braces (..) - this indicates to the processor that you want to create a tuple; the script in your question has quite a few extra braces.