tim laqua dot com

Excel 2007 Hangs When Trying to Edit an OLAP Cube Filter

Tim — Wed, 07 Apr 2010 03:38:16 +0000

Ok, it doesn't hang every time you try to edit an OLAP cube filter, but sometimes - it appears to. In reality, I've never seen it permanently hang - just kind of go away for a while. Here's the basic symptom that the business will report to you:

"Excel freezes (or hangs) when I try to change this filter"

Most of us have seen this at one point or another and shrugged it off as a busy time or processing is going on or there are cats clogging up the tubes, etc. Tonight, I finally decided to figure out what's causing it.

When you have an OLAP cube filter in your pivot table and expand the items list to select your filter items, Excel fires a bunch of Discover commands at the cube (these are pretty light in most cases). If you don't have anything selected yet (so you just have the default member selected) Excel isn't doing much work here and the list pops right up. Here's two basic flow examples for what discover commands are fired off when you DO have an item selected:

If you have one level 1 member selected

MDSCHEMA_CUBES (whatever cube you're connected to)
MDSCHEMA_CUBES (again, same thing)
MDSCHEMA_MEMBERS (LEVEL_UNIQUE_NAME of the filter), no TREE_OP
MDSCHEMA_MEMBERS (MEMBER_UNIQUE_NAME of the selected member) TREE_OP 4, only immediate parent
MDSCHEMA_MEMBERS (MEMBER_UNIQUE_NAME of the selected member's parent) TREE_OP 8, returns itself
MDSCHEMA_MEMBERS (MEMBER_UNIQUE_NAME of the selected member's parent) TREE_OP 1, returns all children

If you have one level 2 member selected in a hierarchy

MDSCHEMA_CUBES (whatever cube you're connected to)
MDSCHEMA_CUBES (again, same thing)
MDSCHEMA_MEMBERS (LEVEL_UNIQUE_NAME of the filter), no TREE_OP
MDSCHEMA_MEMBERS (MEMBER_UNIQUE_NAME of the selected member) TREE_OP 4, only immediate parent
MDSCHEMA_MEMBERS (MEMBER_UNIQUE_NAME of the selected member's parent's parent) TREE_OP 4, immediate parent
MDSCHEMA_MEMBERS (MEMBER_UNIQUE_NAME of the selected member's parent's parent) TREE_OP 8, returns itself
MDSCHEMA_MEMBERS (MEMBER_UNIQUE_NAME of the selected member's parent's parent) TREE_OP 1, all children
MDSCHEMA_MEMBERS (MEMBER_UNIQUE_NAME of the selected member's parent) TREE_OP 8, returns itself
MDSCHEMA_MEMBERS (MEMBER_UNIQUE_NAME of the selected member's parent) TREE_OP 1, returns all children

So, that's six requests for a level 1 member and 9 requests for a level 2 member - both of these take milliseconds to complete the whole flow, so no big deal. When does this become a big deal? When you do something like click the "Select multiple items" checkbox, expand a hierarchy and UNCHECK something... like if you opened the Adventure Works 2008 cube, added Delivery Date.Date as a filter, expanded the "All Periods" level and deselected the first day:

Now click OK and expand the items box - we get a familiar popup:

It comes back after a few seconds with the expected item list, but why did it take so long? Well, based on our flows above, we have a whole bunch of level 1 members selected - 1187 of them to be exact. That means there will be 1187 requests for their parent member (TREE_OP 4) which will all return the [All Periods] member. To illustrate this, I traced that exact operation:

The profiler captured 1192 discover requests - 1187 of them were getting the parents of selected members, two were MDSCHEMA_CUBES requests, one was getting [All Periods], one was getting all the children of [All Periods], and one was getting the Unique Name of the root level. Not cool. Imagine what happens when your talking about tens of thousands of members at a particular level! Granted, that seems unlikely but understand that the logic behind this is "I don't want to see this", "hide just this member", and so forth - in some cases people have no idea what actually happened was they unchecked one of 6,000 members.

So what to do about this? Educate. To fix the immediate problem of a seemingly broken Excel pivot table - just wait for the filter to load, it'll come back eventually. In the future when working with Excel filters, always favor selecting vs. deselecting and filtering at higher levels in the hierarchy whenever possible.

Locating Rampant Database Growth

Tim — Fri, 19 Mar 2010 17:40:02 +0000

Every now and then you get a helpful alert from one of your database servers about disk space being low (you're monitoring that, right?), or a bunch of autogrowth alerts (you're monitoring that too, right?) - but what happens when you get these for a database that you don't expect growth in? Further, what happens when that database is growing rampantly (say like 1GB/hr in my case) and it's a canned database from a 3rd party product? This time it was the database that SolarWinds uses for collection - and apparently it was collecting a lot of something that it wasn't collecting before.

you send out an email asking the end users of the system (IT in this case) if anything changed
adjust so said system stops trying to fill up your drives

Unfortunately, there's often quite a few possibilities for "why" a database is growing - and when it's a canned product, you don't always have the best understanding of why it does what it does when it does it. As a Database Admin, you can help diagnose the problem by letting everyone know what exactly is growing:

First, figure out what table is growing:
spSpaceUsed.sql

CREATE TABLE #Temp_SpSpaceUsed
	([name] VARCHAR(255)
	,[rows] INT
	,[reserved] VARCHAR(255)
	,[DATA]	VARCHAR(255)
	,[index_size] VARCHAR(255)
	,[unused] VARCHAR(255)
	,[CreatedTime] DATETIME DEFAULT GETDATE())
 
EXEC sp_msforeachtable 'INSERT INTO #Temp_SpSpaceUsed ([name],[rows],[reserved],[data],[index_size],[unused]) EXEC sp_spaceused [?]'
 
WAITFOR DELAY '00:01:00'
 
EXEC sp_msforeachtable 'INSERT INTO #Temp_SpSpaceUsed ([name],[rows],[reserved],[data],[index_size],[unused]) EXEC sp_spaceused [?]'
 
;WITH SpSpaceUsed AS
(
	SELECT 
		[name]
		,CAST(LEFT([DATA], LEN([DATA]) - 2) AS INT) AS [DATA]
		,[CreatedTime]
		,ROW_NUMBER() OVER (PARTITION BY [name] ORDER BY [CreatedTime] ASC) AS [RowNumber]
	FROM #Temp_SpSpaceUsed	
)
SELECT 
	 a.[name]
	,b.DATA - a.DATA AS [GrowthKB]
	,(CAST(b.DATA - a.DATA AS FLOAT) / DateDiff(MILLISECOND, a.CreatedTime, b.CreatedTime)) * 1000.0 * 60.0 * 60.0 AS [GrowthKBPerHour]
FROM SpSpaceUsed a
	INNER JOIN SpSpaceUsed b
		ON a.[name] = b.[name]
			AND a.RowNumber = 1
			AND b.RowNumber = 2
ORDER BY
	GrowthKBPerHour DESC
 
DROP TABLE #Temp_SpSpaceUsed

Next, look at the culprit (assuming there's one table doing all the damage - I've found that's usually the case) and try to decipher the column names. Finally, run a few queries against the table - WITH(NOLOCK) of course - hey, this table is pretty busy. Now try and see where all the action is coming from. In this case, one of the columns was [IP] so I just did a COUNT(1) by IP for the last hour's records. Sure enough, 99% of the records were coming from a single IP, sent that IP off to the guys who know what IPs mean, and they knew exactly which change caused the growth.

Bottom line - nobody knows everything about everything, but we can learn pretty quick with the right information.

Script sp_configure Output To Migrate Settings Between SQL Servers

Tim — Sat, 27 Feb 2010 23:22:30 +0000

Sometimes when setting up a new server, all you want is for it to work and be configured exactly like the old server. In some cases, the new server is almost identical in every way already. If you've worked much with SQL Server, you know very well that there are a billion switches and knobs and every time you do this, you forget at least one. I had a similar situation recently and figured I'd be lazy about it and just turned the output of sp_configure in to a script I could execute on the new server:

scriptSpConfigure.sql - run this on the old/source server

EXEC SP_CONFIGURE 'show advanced options' , 1;
GO
RECONFIGURE;
GO
DECLARE @spConfigureOutput TABLE
	([name] VARCHAR(255)
	,[minimum] INT
	,[maximum] INT
	,[config_value] INT
	,[run_value] INT)
 
INSERT INTO @spConfigureOutput
EXEC SP_CONFIGURE
 
SELECT	'EXEC sp_configure ''' + name + ''', ' + CAST(config_value AS VARCHAR)
FROM	@spConfigureOutput
GO
EXEC SP_CONFIGURE 'show advanced options' , 0;
GO
RECONFIGURE;
GO

Copy the selected rows and paste in to another SSMS window, add the show advanced options switch above and below - should end up with something like this after you remove all the settings you don't care about or don't want on the new server:

EXEC SP_CONFIGURE 'show advanced options' , 1;
GO
RECONFIGURE;
GO
 
EXEC SP_CONFIGURE 'Ad Hoc Distributed Queries', 1
EXEC SP_CONFIGURE 'Agent XPs', 1
EXEC SP_CONFIGURE 'clr enabled', 1
EXEC SP_CONFIGURE 'cost threshold for parallelism', 5
EXEC SP_CONFIGURE 'cursor threshold', -1
EXEC SP_CONFIGURE 'Database Mail XPs', 1
EXEC SP_CONFIGURE 'default full-text language', 1033
EXEC SP_CONFIGURE 'default trace enabled', 1
EXEC SP_CONFIGURE 'fill factor (%)', 80
EXEC SP_CONFIGURE 'ft crawl bandwidth (max)', 100
EXEC SP_CONFIGURE 'ft notify bandwidth (max)', 100
EXEC SP_CONFIGURE 'max full-text crawl range', 4
EXEC SP_CONFIGURE 'max server memory (MB)', 10000
EXEC SP_CONFIGURE 'max text repl size (B)', 65536
EXEC SP_CONFIGURE 'min memory per query (KB)', 1024
EXEC SP_CONFIGURE 'min server memory (MB)', 10000
EXEC SP_CONFIGURE 'nested triggers', 1
EXEC SP_CONFIGURE 'network packet size (B)', 4096
EXEC SP_CONFIGURE 'PH timeout (s)', 60
EXEC SP_CONFIGURE 'query wait (s)', -1
EXEC SP_CONFIGURE 'remote access', 1
EXEC SP_CONFIGURE 'remote login timeout (s)', 20
EXEC SP_CONFIGURE 'remote query timeout (s)', 900
EXEC SP_CONFIGURE 'scan for startup procs', 1
EXEC SP_CONFIGURE 'server trigger recursion', 1
EXEC SP_CONFIGURE 'SMO and DMO XPs', 1
 
EXEC SP_CONFIGURE 'show advanced options' , 0;
RECONFIGURE;
GO;

Script RESTORE DATABASE … WITH MOVE Stub

Tim — Sat, 27 Feb 2010 18:29:26 +0000

I think most people setup the drives on their Development servers to match their Production servers - this is so that restores go smoothly and files go where they're supposed to, things match up nicely, etc. Unfortunately, when you create a full backup all the backup file contains is the logical name of all the files - no physical paths. This means that if the database doesn't exist on the destination server yet, the engine has absolutely no idea where to put the files or what to name them.

The following script can be executed in the context of the source (original) database that was backed up to script out the MOVE statements if your plan is to put the files in the same place on the destination server.

scriptRestoreWithMove.sql - run this in the context of the source database

SET NOCOUNT ON
 
DECLARE @MoveOption AS TABLE
	([Id] INT IDENTITY(1,1)
	,[MoveOption] VARCHAR(MAX))
 
PRINT 'RESTORE DATABASE ' + DB_NAME(DB_ID())
-- Edit this to match where you're restoring from
PRINT 'FROM DISK = ''' + DB_NAME(DB_ID()) + '.bak'''  
PRINT 'WITH'
 
INSERT INTO @MoveOption ([MoveOption])
SELECT
    'MOVE ''' + a.name + ''' TO ''' + a.FILENAME + '''' AS [MOVE OPTION]
FROM
    dbo.sysfiles a
 
DECLARE 
	 @LastId INT = 0
	,@MoveOptionText VARCHAR(MAX)
 
WHILE EXISTS (SELECT TOP 1 1 FROM @MoveOption WHERE [Id] > @LastId)
BEGIN
	SELECT 		TOP 1
			 @MoveOptionText = [MoveOption] 
			,@LastId = [Id]
	FROM		@MoveOption
	WHERE		[Id] > @LastId
	ORDER BY	[Id] ASC
 
	PRINT CASE WHEN @LastId = 1 THEN '' ELSE ',' END + @MoveOptionText
END
 
PRINT 'GO'
 
PRINT 'ALTER DATABASE ' + DB_NAME(DB_ID()) + ' SET MULTI_USER'
PRINT 'GO'

And the output looks like this:

RESTORE DATABASE AdventureWorksDW
FROM DISK = 'AdventureWorksDW.bak'
WITH
MOVE 'AdventureWorksDW_Data' TO 'D:\AdventureWorksDW\AdventureWorksDW_Data.mdf'
,MOVE 'AdventureWorksDW_Log' TO 'T:\AdventureWorksDW\AdventureWorksDW_Log.LDF'
GO
ALTER DATABASE AdventureWorksDW SET MULTI_USER
GO

If you end up wanting to do something similar with existing databases, make sure to add a command before the RESTORE to set SINGLE_USER mode.

VersionOne Nag: Reminding Us To Burn Our Points

Tim — Fri, 12 Feb 2010 04:19:41 +0000

You'll forget - we all do. After a hard day of work, the last thing that pops in to your head is "hey, I should go update this story in VersionOne before I go home." Oh no, the only thing you're thinking about is how bad traffic will be, what you're having for dinner, your trip to Seattle, which bar has the best drink specials tonight, etc. So accepting that we forget to burn points now and then, why don't we just make a script to remind us? While you're at it, might as well add a reminder for when you accidentally close a story without setting the ToDo points to zero.

Luckily, since you obviously use VersionOne (because it's the gold standard), there's an API for that:

Note: The following script will request your VersionOne username and password via Read-Host - if you want this to run every day at 4pm, you need to embed your username and password somewhere (securestring support in powershell is helpful, but still - nothing's perfect in this scenario unless you can use LDAP auth)

PowerShell script: v1Nag.ps1

$emailFrom = "whoever.replies@shouldgoto.com"
$emailSMTPServer = "your.smtpserver.com"
$teamName = "Your Team Name in VersionOne"
 
$url = "https://wwwXXX.v1host.com/yourhostingid/"
$username = Read-Host "User"
$password = Read-Host -assecurestring "Password"
 
$password = [System.Runtime.InteropServices.Marshal]::PtrToStringAuto([System.Runtime.InteropServices.Marshal]::SecureStringToBSTR($password))
 
$apiClient = resolve-path "versionone.sdk.apiclient.dll"
$objectModel = resolve-path "versionone.sdk.objectmodel.dll"
[Reflection.Assembly]::LoadFrom($apiClient) > $NULL
[Reflection.Assembly]::LoadFrom($objectModel) > $NULL
$v1 = new-object VersionOne.SDK.ObjectModel.V1Instance( $url, $username, $password)
$v1.Validate()
 
$sf = new-object VersionOne.SDK.ObjectModel.Filters.StoryFilter
$v1.Get.Iterations($null) | ? { $_.IsActive } | % { $sf.Iteration.Add($_) }
$v1.Get.Teams($null) | ? { $_.Name -match $teamName } | % { $sf.Team.Add($_) } 
$stories = $v1.Get.Stories($sf)
 
$memberDetail = @{}
 
foreach ($story in $stories) {
	foreach ($owner in $story.Owners) {
		if (!$memberDetail.Contains($owner.ID)) { 		
			$md = New-Object PSObject | select Email, DetailEstimate, ToDo, Effort, EffortToday, DoneWithToDo
			$md.Email = $owner.Email;
			$md.DetailEstimate = 0;
			$md.ToDo = 0;
			$md.Effort = 0;
			$md.EffortToday = 0;
			$md.DoneWithToDo = @()
			$memberDetail.Add($owner.ID, $md)
		}
 
		$memberDetail[$owner.ID].ToDo += $story.ToDo;
		$memberDetail[$owner.ID].DetailEstimate += $story.DetailEstimate;
 
		if ($story.ToDo -ne 0 -and $story.Status.CurrentValue -eq "Done") {
			$memberDetail[$owner.ID].DoneWithToDo += $story;
		}
 
		foreach ($effortRecord in $story.GetEffortRecords($null)) {
			$memberDetail[$owner.ID].Effort += $effortRecord.Value;
			if ($effortRecord.CreateDate -ge [DateTime]::Today) {
				$memberDetail[$owner.ID].EffortToday += $effortRecord.Value;
			}
		}
	}
}
 
$smtp = new-object Net.Mail.SmtpClient($emailSMTPServer)
 
$memberDetail.Values | ?{ $_.EffortToday -eq 0 -and $_.ToDo -gt 0} | % { $smtp.Send($emailFrom, $_.Email, "V1 Nag: No Points Burned Down Today", "Please make sure you have recorded any effort completed on your stories for the day.") }
 
$memberDetail.Values | ?{ $_.DoneWithToDo.Count -gt 0 } | % { $smtp.Send($emailFrom, $_.Email, "V1 Nag: Story Marked As Done With ToDo Balance", "Please make sure you have zero'd out ToDo on your Completed/Done stories.") }

You'll need to download the .NET SDK and build the requisite DLLs from:
http://community.versionone.com/Downloads/default.aspx

Clone Analysis Services Partitions with PowerShell

Tim — Fri, 15 Jan 2010 20:45:20 +0000

Most of us with large Analysis Services cubes partition our cubes by month or year or some other time-based slice and we have all, at one point or another, developed some way to create partitions for new months on-demand. Often, the solution to this seems to be a C# console application or SSIS package using AMO to create a new partition based off an existing partition. The problem I see with this is that maintaining it requires opening up the project or package, making changes, re-compiling, deploying, testing, deploying to production, verifying, etc. It also requires that whoever is going to maintain it is comfortable with C#.

To simplify the maintenance and get rid of the "black box" factor that utility apps like this tend to have, I put together a PowerShell script to do the same thing and a stored procedure to call the script. Really, it doesn't matter what you use as you're most likely using an almost identical chunk of code to get your new partition created - my argument is that using PowerShell instead of C# or SSIS reduces the cost of maintenance, improves readability, and facilitates better understanding throughout your team.

The PowerShell Script: cloneSSASPartition.ps1

param(
   [string] $ServerName, 		<# The name of the Analysis Services Instance #>
   [string] $DatabaseName, 		<# The name of the Database our Cube is in #>
   [string] $CubeName,     		<# The name of the Cube that our Measure Group is in #>
   [string] $MeasureGroupName, 	<# The name of the Measure Group that our Partition is in #>
   [string] $SourcePartitionName, 	<# The name of the Partition we want to merge in to another partition #>
   [string] $NewPartitionName, 	<# The desired name of the new partition #>
   [string] $NewPartitionQuery 	<# The Query source for the new partition (will use the same datasource as the SourcePartition #>
)
 
[System.Reflection.Assembly]::LoadWithPartialName("Microsoft.AnalysisServices") >$NULL
 
$server = New-Object Microsoft.AnalysisServices.Server
$server.Connect($ServerName)
 
$database = $server.Databases.GetByName($DatabaseName)
$cube = $database.Cubes.GetByName($CubeName)
$measureGroup = $cube.MeasureGroups.FindByName($MeasureGroupName)
$sourcePartition = $measureGroup.Partitions.GetByName($SourcePartitionName)
$sourceQueryBinding = $sourcePartition.Source
 
$newPartition = $sourcePartition.Clone()
$newPartition.ID = $NewPartitionName
$newPartition.Name = $NewPartitionName
$measureGroup.Partitions.Add($newPartition) > $NULL
$newPartition.Source = New-Object Microsoft.AnalysisServices.QueryBinding($sourceQueryBinding.DataSourceID, $NewPartitionQuery)
$newPartition.Update()
 
$server.Disconnect()

The stored procedure to call the script: [Utility_CloneSSASPartition_S01] - assumes you placed cloneSSASPartition.ps1 at the root of the C: drive. If you put it elsewhere, update the proc to reflect the actual location.

SET ANSI_NULLS ON
GO
 
SET QUOTED_IDENTIFIER ON
GO
 
/*************************************************************************************
***
*** Procedure:  		[Utility_CloneSSASPartition]
*** Purpose:		Executes PowerShell script to clone a SSAS partition
***			
***			
*** Author:		tl
*** Date Created:		2010-01-15
*** 
*** Revision History
*** Date		Author			Description
*** 2010-01-15	tl			Created
*************************************************************************************/
 
CREATE PROCEDURE [dbo].[Utility_CloneSSASPartition_S01]
(
	@ServerName AS VARCHAR(255),
	@DatabaseName AS VARCHAR(255),
	@CubeName AS VARCHAR(255),
	@MeasureGroupName AS VARCHAR(255),
	@SourcePartitionName AS VARCHAR(255),
	@NewPartitionName AS VARCHAR(255),
	@NewPartitionQuery AS VARCHAR(255)
)
AS 
 
SET NOCOUNT ON
DECLARE @cmd VARCHAR(8000)
SET @cmd = 
	'powershell.exe -Command "& {'
	+ ' C:\cloneSSASPartition.ps1'
	+ ' -ServerName:\"' + @ServerName + '\"'
	+ ' -DatabaseName:\"' + @DatabaseName + '\"'
	+ ' -CubeName:\"' + @CubeName + '\"'
	+ ' -MeasureGroupName:\"' + @MeasureGroupName + '\"'
	+ ' -SourcePartitionName:\"' + @SourcePartitionName + '\"'
	+ ' -NewPartitionName:\"' + @NewPartitionName + '\"'
	+ ' -NewPartitionQuery:\"' + @NewPartitionQuery + '\"}"'
 
PRINT @cmd
EXEC xp_cmdshell @cmd

And, finally, we execute it

EXEC [dbo].[Utility_CloneSSASPartition_S01]
	 @ServerName = 'TESTSERVER'
	,@DatabaseName = 'TestDatabase'
	,@CubeName = 'TestCube'
	,@MeasureGroupName = 'Test Measure Group'
	,@SourcePartitionName = 'Test Partition - 200912'
	,@NewPartitionName = 'Test Partition - 201001'
	,@NewPartitionQuery = 'SELECT * FROM dbo.FactTableMonth(''1/1/2010'', ''2/1/2010'')'

Again, I'm not saying there's anything "wrong" with using C# or SSIS Packages to do this sort of thing, I'm just pointing out that C# isn't always the answer - especially when someone other than you needs to maintain this thing in the future. Keep it simple.

Trending SQL Server Agent Job Duration by Hour

Tim — Wed, 23 Dec 2009 03:07:38 +0000

Earlier today I noticed a SQL Server Agent job taking a little longer than usual (or what I thought was longer than usual). Let's face it, we're not staring at the Job Activity monitor all day, so unless you've written a report to monitor job run times - on occasion you ask yourself "is that a normal run time for this thing?" The job I was curious about happened to be a job that runs throughout the day and should only have real work to do once or twice an hour - and it should run roughly the same amount of time on any given business day for a given hour (i.e. at 1:00 PM on any given business day, this thing should do the same amount of work).

So I came up with the following query to PIVOT the run duration on the hour the job executed:

DECLARE @JobName VARCHAR(255) = 'Agent Job Name'
 
;WITH JobHistory AS
(
  SELECT
     a.run_date
    ,a.run_time / 10000 AS [HOUR]
    ,(a.run_duration / 10000 * 60 * 60	-- Hours
     + a.run_duration % 10000 / 100 * 60	-- Minutes
     + a.run_duration % 100			-- Seconds
     ) / 60.0 AS [DurationMinutes]
  FROM
    msdb.dbo.sysjobhistory a WITH(NOLOCK)
    INNER JOIN msdb.dbo.sysjobs b WITH(NOLOCK)
    ON 
      a.[job_id] = b.[job_id]
      AND b.[name] = @JobName
      AND step_id = 0
      AND run_status = 1
)
SELECT *
FROM
  JobHistory	
  PIVOT
  ( SUM(DurationMinutes)
    FOR [HOUR] 
    IN  ([00],[01],[02],[03],[04],[05]
        ,[06],[07],[08],[09],[10],[11]
        ,[12],[13],[14],[15],[16],[17]
        ,[18],[19],[20],[21],[22],[23])
  ) AS p

UPDATE: Added AND run_status = 1

Then made a table (not sure why, but I can't resist doing the heatmap thing):

click the image to enlarge

And, of course, the chart so you can point at a picture should you have to explain something to somebody in the future:

click the image to enlarge

Wouldn’t it be fun if Cubes could talk?

Tim — Thu, 19 Nov 2009 19:47:21 +0000

I didn't say "wouldn't it be useful" because after putting a test together, asking a cube questions with no context tends to return answers that it probably shouldn't have returned. In BI, it is incredibly important to understand what exactly it is you're asking for - if we just say we want "sales" and return an answer, nobody really knows what we meant by "sales." Sure, in various circles, "sales" means the same thing - but once you start talking to different areas, departments, etc - the meaning of the word starts to shift.

But I digress - asking cubes questions is still pretty fun and some of the random things it returns when you point it at your own cubes can be flat out hilarious.

Here's a few questions thrown at the Adventure Works cube in the Adventure Works DW 2008 Analysis Services database

what was sales for nichole nara
SELECT {[Measures].[Internet Sales Amount]} ON 0  FROM (SELECT {[Customer].[Customer].[Nichole Nara]} ON 0 FROM [Adventure Works])
$13,295.38
what were sales for nichole nara in cy 2004
SELECT {[Measures].[Internet Sales Amount]} ON 0  FROM (SELECT {[Date].[Calendar Year].[CY 2004]} ON 0 FROM (SELECT {[Customer].[Customer].[Nichole Nara]} ON 0 FROM [Adventure Works]))
$2,419.06
what were sales for nichole nara in cy 2004 in north america
SELECT {[Measures].[Internet Sales Amount]} ON 0  FROM (SELECT {[Sales Territory].[Sales Territory Group].[North America]} ON 0 FROM (SELECT {[Date].[Calendar Year].[CY 2004]} ON 0 FROM (SELECT {[Customer].[Customer].[Nichole Nara]} ON 0 FROM [Adventure Works])))
No Results Found
what were sales for nichole nara in cy 2004 for bikes
SELECT {[Measures].[Internet Sales Amount]} ON 0  FROM (SELECT {[Product].[Category].[Bikes]} ON 0 FROM (SELECT {[Date].[Calendar Year].[CY 2004]} ON 0 FROM (SELECT {[Customer].[Customer].[Nichole Nara]} ON 0 FROM [Adventure Works])))
$2,384.07
what were sales in north america in fy 2003 for bikes
SELECT {[Measures].[Internet Sales Amount]} ON 0  FROM (SELECT {[Product].[Category].[Bikes]} ON 0 FROM (SELECT {[Date].[Fiscal Year].[FY 2003]} ON 0 FROM (SELECT {[Sales Territory].[Sales Territory Group].[North America]} ON 0 FROM [Adventure Works])))
$1,739,306.95
sales for cy 2004 in cy q1, cy q2, and cy q4 in north america and canada
SELECT {[Measures].[Internet Sales Amount]} ON 0  FROM (SELECT {[Sales Territory].[Sales Territory Group].[North America],[Customer].[Country].[Canada]} ON 0 FROM (SELECT {[Date].[Calendar Quarter of Year].[CY Q1],[Date].[Calendar Quarter of Year].[CY Q2],[Date].[Calendar Quarter of Year].[CY Q4]} ON 0 FROM (SELECT {[Date].[Calendar Year].[CY 2004]} ON 0 FROM [Adventure Works])))
I don't know - Query Failure: Members belong to different hierarchies in the  function.
sales for cy 2004 in cy q1, cy q2, and cy q4 in north america
SELECT {[Measures].[Internet Sales Amount]} ON 0  FROM (SELECT {[Sales Territory].[Sales Territory Group].[North America]} ON 0 FROM (SELECT {[Date].[Calendar Quarter of Year].[CY Q1],[Date].[Calendar Quarter of Year].[CY Q2],[Date].[Calendar Quarter of Year].[CY Q4]} ON 0 FROM (SELECT {[Date].[Calendar Year].[CY 2004]} ON 0 FROM [Adventure Works])))
$3,967,371.16
sales between cy 2003 and cy 2004 for large resellers
SELECT {[Measures].[Internet Sales Amount]} ON 0  FROM (SELECT [Large Resellers] ON 0 FROM (SELECT {[Date].[Calendar Year].[CY 2003]:[Date].[Calendar Year].[CY2004]} ON 0 FROM [Adventure Works]))
$19,561,960.04
sales for top 50 customers in fy 2004 in fy q1 and fy q3
SELECT {[Measures].[Internet Sales Amount]} ON 0  FROM (SELECT {[Date].[Fiscal Quarter of Year].[FY Q1],[Date].[Fiscal Quarter of Year].[FY Q3]} ON 0 FROM (SELECT {[Date].[Fiscal Year].[FY 2004]} ON 0 FROM (SELECT [Top 50 Customers] ON 0 FROM [Adventure Works])))
$177,263.41

The above questions returned pretty much exactly what I was looking for except for the one that threw us an error about members belonging to different hierarchies. That one was because "Canada" was in [Customer].[Country] and "north america" was in [Sales Territory].[Sales Territory Group]. Yeah, it was a dumb question because canada is in north america, but it illustrates that when you don't qualify things, you can get some pretty confusing results.

How's it work?

Enumerate all the Attribute Values for each Attribute Hierarchy and store them as "keywords"
Get some input
Look for the aforementioned "keywords" in the input string
Construct an MDX query based on what "keywords" were found in the string
Make sure there aren't any extra (non-noise) words in the string
Run Query
Return Results

How the queries are constructed

Attribute Values just create a subcube - it creates a new nested subcube for each list, range, or named set it finds
It only supports one measure as written and it just selects that from whatever quagmire of subcubes it's created

There are tons of issues with this methodology when we have role playing dimensions because the attribute values will be the same for these dimensions - so only the first one enumerated will get in (there's a way to ignore certain dimensions in the app.config - ignore all but one role playing dimension). It throws away context and calls it "noise" - now that's dangerous; if you throw away the context and just look for keywords, it's pretty easy to answer a question that wasn't really asked (oops).

As far as configuring this thing goes - take a look at the app.config and throw in appropriate values for your setup. You'll have to tweak the *Pattern values to fit things you want to match or exclude in your particular cube. It's configured for the Adventure Works cube right now, but I didn't work with it much - I just flat out excluded a few dims and hierarchies that would obviously confuse it.

Also note that, as written, it needs an Analysis Services 2008 cube.

Let me know if you want to give it a try and I'll be more than happy to work with you on configuring/modifying it for your setup. Also post comments on interesting observations you have when asking cubes unqualified questions

Download the VS 2008 Project: SmartCube.zip

Also note, the project uses an interface called IODoodad for getting questions and responding - this has two public methods, GetInput() and Respond(). The original theory here is that we could make an old-school IRC bot type thing. We did manage to hook it to Yammer for its IO, respond to private messages and public messages prefixed with @cube, etc. I didn't really explain the config for YammerIO (there's a custom config section for that that needs a bunch of OAuth stuff). Again, if you're interested in getting that working, let me know and I'll post instructions. The project above is set to use ConsoleIO.

I don't plan on doing any more development here, it was just a quick PoC to see what it looked like and get some initial feedback. If anyone's interested in moving forward with it and making it smarter (the project arguably should have been titled DumbCube at this point), let me know so I can post a link to your work and fork away

Charting Analysis Services 2008 Trace Events

Tim — Sat, 24 Oct 2009 15:05:06 +0000

The other day I was running some Analysis Services traces for some reason or another and ran across Chris Webb's post on (Visualising Analysis Services Trace Information in Reporting Services). After looking over that post, I thought it'd be interesting to visualize various types of processing, queries, etc - basically take a trace and then graph out what happened and when. Here's a few samples of what the project yielded:

Red: Cube Processing on a 2 Minute schedule
Green: Cube Processing on a 1 hour schedule
Blue: Cube Processing on a 1 hour schedule
Black: Query Activity

Most of the activity here is from SSRS subscriptions firing around 8AM

This is the report filtered to only show events that had a duration greater than 10,000 MS (10 seconds) - Red cube is a trouble maker

And this is general SSAS activity over a few hours with the nearly instant events filtered out

The following query assumes that the trace data ended up in a table:

@StartMillisecond: INT; Minimum relative start time (millisecond value, kinda have to guess and check)
@EndMillisecond: INT; Maximum relative start time (millisecond value, kinda have to guess and check)
@MinDuration: INT; Minimum duration in milliseconds

;WITH [ProcessEvent] AS
(
SELECT
	'Process' AS [Event]
	,cast([TextData] AS XML).value
		('declare namespace B="http://schemas.microsoft.com/analysisservices/2003/engine";           
		(//B:DatabaseID)[1]'
		,'VARCHAR(1024)') AS [DatabaseID]
	,cast([TextData] AS XML).value
		('declare namespace B="http://schemas.microsoft.com/analysisservices/2003/engine";           
		(//B:DimensionID)[1]'
		,'VARCHAR(1024)') AS [DimensionID]
	,cast([TextData] AS XML).value
		('declare namespace B="http://schemas.microsoft.com/analysisservices/2003/engine";           
		(//B:MeasureGroupID)[1]'
		,'VARCHAR(1024)') AS [MeasureGroupId]
	,[StartTime]
	,[Duration]
FROM dbo.zTrace_MSAS_20091023_2 T
WHERE 
	T.[EventClass] = 16 
	AND T.[EventSubclass] IN (3, 12)
	AND [TextData] LIKE '%Process%'
),
[QueryEvent] AS
(
SELECT
	'Query' AS [Event]
	,[DatabaseName] AS [DatabaseID]
	,NULL AS [DimensionID]
	,NULL AS [MeasureGroupId]
	,[StartTime]
	,[Duration]
FROM dbo.zTrace_MSAS_20091023_2 T
WHERE T.[EventClass] = 10
),
[Event] AS
(
	SELECT * FROM [ProcessEvent]
	UNION
	SELECT * FROM [QueryEvent]
)
SELECT
	 ROW_NUMBER() OVER (ORDER BY [StartTime] ASC) AS [EventId]
	,DATEADD(HOUR, -5, [StartTime]) AS [StartTime]
	,DATEDIFF(MILLISECOND, (SELECT MIN([StartTime]) FROM Event), [StartTime]) [RelativeStartTime]
	,[Duration]
	,[Event]
	,[DatabaseID]
	,CASE
		WHEN [DimensionID] IS NOT NULL THEN 'Dimension'
		WHEN [MeasureGroupID] IS NOT NULL THEN 'Measure Group'
		ELSE 'None'
	 END AS [ObjectType]	,CASE
		WHEN [DimensionID] IS NOT NULL THEN [DimensionID]
		WHEN [MeasureGroupID] IS NOT NULL THEN [MeasureGroupID]
		ELSE 'None'
	 END AS [ObjectId]	
FROM [Event]
WHERE
	DATEDIFF(MILLISECOND, (SELECT MIN([StartTime]) FROM Event), [StartTime]) >= @StartMillisecond
	AND DATEDIFF(MILLISECOND, (SELECT MIN([StartTime]) FROM Event), [StartTime]) < @EndMillisecond
	AND Duration >= @MinDuration

A few notes on the report:

[EventId] is the Series Group
This is a Range chart type, the Bottom value was [RelativeStartTime] and the Top value was [RelativeStartTime] + [Duration]

Coloring the series and markers uses the following expression

=Iif(Fields!Event.Value = "Process"
	,Iif(Fields!DatabaseID.Value = "SSASDB"
		,"Red"
		,Iif(Fields!DatabaseID.Value = "OtherSSASDB"
			,"Green"
			,Iif(Fields!DatabaseID.Value = "YetAnotherSSASDB"
				,"Blue"
				,"Pink"
			)
		)
	)
	,"Black"
)

And that's about it - if you have any more questions about it just let me know. Just seemed like an interesting way to look at trace data It also wouldn't be that difficult to generate events for other things, like your ETL processing and such so you could see that on top of everything else.

Estimating the Size of a Table in SQL Server 2008

Tim — Fri, 16 Oct 2009 02:35:00 +0000

I have read this (http://msdn.microsoft.com/en-us/library/ms178085.aspx) article at least 6, maybe 7 times in the past - and every time I say to myself "this is ridiculous - someone has to have written a script to do this by now" and every time, I google for hours and fail to find anything. So I finally gave up and wrote something to do it. Note, I've only verified it on 100% fixed width tables. I compared its output to a 600+ million row table and it came out somewhere around 3% higher - fine with me as I'd rather over-estimate space requirements than under-estimate.

First, we have to create a table type to hold our table column definitions. the datatype is required, length is required for anything that allows length (MAX is not handled for any types), and the worst case scenario is always used so if you know you won't be filling those VARCHAR(255)s, feel free to use a smaller, more average, number. IsKey is a BIT value that tells the proc to use those columns as part of the index key. The rest of it is all calculated.

CREATE TYPE TableColumnDefinition AS TABLE
	([DataType] VARCHAR(16) NOT NULL
	,[Length] INT
	,[IsKey] BIT
	,[DataLength] AS
		CASE 
			WHEN [DataType] = 'BIT' THEN 0.125
			WHEN [DataType] = 'TINYINT' THEN 1
			WHEN [DataType] = 'SMALLINT' THEN 2
			WHEN [DataType] IN ('INT', 'SMALLMONEY', 'REAL', 'SMALLDATETIME') THEN 4
			WHEN [DataType] IN ('BIGINT', 'DATETIME', 'MONEY') THEN 8
			WHEN [DataType] = 'UNIQUEIDENTIFIER' THEN 16
			WHEN [DataType] = 'NVARCHAR' THEN [Length] * 2 + 2
			WHEN [DataType] = 'VARCHAR' THEN [Length] + 2	-- DOESN'T SUPPORT MAX
			WHEN [Datatype] = 'DATE' THEN 3 
			WHEN [Datatype] = 'TIME' THEN 5 -- Worst case
			WHEN [Datatype] = 'DATETIME2' THEN 8 -- Worst case
			WHEN [Datatype] = 'DATETIMEOFFSET' THEN 10 -- Worst case
			WHEN [DataType] = 'CHAR' THEN [Length]
			WHEN [DataType] = 'NCHAR' THEN [Length] * 2 + 2
			WHEN [DataType] = 'BINARY' THEN [Length]
			WHEN [DataType] = 'VARBINARY' THEN [Length] + 2
			WHEN [Datatype] = 'TIMESTAMP' THEN 8
			WHEN [DataType] = 'XML' THEN [Length]	-- Estimate an average size in bytes, 2GB max
			WHEN [DataType] IN ('DECIMAL', 'NUMERIC') THEN 
				CASE 
					WHEN [Length] BETWEEN 1 AND 9	THEN 5
					WHEN [Length] BETWEEN 10 AND 19 THEN 9
					WHEN [Length] BETWEEN 20 AND 28 THEN 13
					WHEN [Length] BETWEEN 29 AND 38 THEN 17
				END
			WHEN [DataType] = 'FLOAT' THEN
				CASE
					WHEN [Length] BETWEEN 1 AND 24 THEN 4
					WHEN [Length] BETWEEN 25 AND 53 THEN 8
					WHEN [Length] IS NULL THEN 8	-- Default n is 53
				END
			ELSE 0
		END
	,[IsVariableLength] AS
		CASE 
			WHEN [DataType] IN ('VARCHAR', 'NVARCHAR', 'VARBINARY') THEN 1
			ELSE 0
		END)

Now for the proc that attempts to replicate the methodology outlined on the aforementioned MSDN article. The results appeared to be accurate or over the actual value, which was exactly what I needed, so I went with it and moved on. I'm completely open to corrections, suggestions, improvements, etc. Just make a comment and I'll integrate them (assuming they work .

/*************************************************************************************
***
*** Procedure:  		[Utility_EstimateClusteredIndexSize]
*** Purpose:		    Estimates the amount of space needed for a given clustered index
*** Methodology:		http://msdn.microsoft.com/en-us/library/ms178085.aspx
***			
***			
*** Author:		      Tim Laqua
*** Date Created:	  2009-10-15 
*** 
*** Revision History
*** Date		Author			Description
*** 2009-10-15	tl				Created
*************************************************************************************/
ALTER PROCEDURE [dbo].[Utility_EstimateClusteredIndexSize]
(
	 @Num_Rows FLOAT
	,@ClusteredIndexIsUnique BIT
	,@Fill_Factor FLOAT
	,@TableColumn TableColumnDefinition READONLY
)
AS 
 
SET NOCOUNT ON
 
DECLARE
	 @Num_Cols FLOAT = 0					-- total number of columns (fixed-length and variable-length)
	,@Fixed_Data_Size FLOAT = 0			-- total byte size of all fixed-length columns
	,@Num_Variable_Cols FLOAT	= 0			-- number of variable-length columns
	,@Max_Var_Size FLOAT	= 0				-- maximum byte size of all variable-length columns
	,@Variable_Data_Size FLOAT = 0
 
DECLARE @VAR TABLE
	([Name] VARCHAR(64)
	,[VALUE] FLOAT
	,[DisplayValue] AS
		CASE
			WHEN LOWER([Name]) LIKE '%size%' OR LOWER([Name]) LIKE '%space%' THEN
				CASE 
					WHEN [VALUE] < 1024 THEN CAST(CAST(ROUND([VALUE], 3) AS NUMERIC(16,0)) AS VARCHAR(16)) + ' B'
					WHEN [VALUE] < 1024*1024 THEN CAST(CAST(ROUND([VALUE]/1024, 3) AS NUMERIC(16,3)) AS VARCHAR(16)) + ' KB'
					WHEN [VALUE] < 1024*1024*1024 THEN CAST(CAST(ROUND([VALUE]/1024/1024, 3) AS NUMERIC(16,3)) AS VARCHAR(16)) + ' MB'
					ELSE CAST(CAST(ROUND([VALUE]/1024/1024/1024, 3) AS NUMERIC(16,3)) AS VARCHAR(16)) + ' GB'
				END
			ELSE CAST(FLOOR([VALUE]) AS VARCHAR(16))
		END)
 
 
-- Leaf level data
SELECT 
	 @Num_Variable_Cols = COUNT(1)
	,@Max_Var_Size = ISNULL(SUM([DataLength]), 0)
FROM @TableColumn WHERE [IsVariableLength] = 1
 
SELECT 
	 @Num_Cols = COUNT(1) + @Num_Variable_Cols
	,@Fixed_Data_Size = CEILING(ISNULL(SUM([DataLength]), 0))
FROM @TableColumn WHERE [IsVariableLength] = 0
 
 
INSERT INTO @VAR VALUES ('@Fixed_Data_Size', @Fixed_Data_Size)
 
IF @ClusteredIndexIsUnique <> 0
BEGIN
	SET @Num_Cols += 1
	IF @Num_Variable_Cols > 0
		SET @Num_Variable_Cols += 1
	IF @Max_Var_Size > 0
		SET @Max_Var_Size += 4
END
 
DECLARE @Null_Bitmap INT = CEILING(2 + ((@Num_Cols + 7) / 8))
INSERT INTO @VAR VALUES ('@Null_Bitmap', @Null_Bitmap)
 
-- if there are variable columns
IF @Num_Variable_Cols > 0
	SET @Variable_Data_Size = 2 + (@Num_Variable_Cols * 2) + @Max_Var_Size 
INSERT INTO @VAR VALUES ('@Num_Cols', @Num_Cols)
INSERT INTO @VAR VALUES ('@Num_Variable_Cols', @Num_Variable_Cols)
INSERT INTO @VAR VALUES ('@Max_Var_Size', @Max_Var_Size)
 
DECLARE @Row_Size FLOAT = CEILING(@Fixed_Data_Size + @Variable_Data_Size + @Null_Bitmap + 4 )
INSERT INTO @VAR VALUES ('@Row_Size', @Row_Size)
 
DECLARE @Rows_Per_Page FLOAT = FLOOR(8096 / (@Row_Size + 2))
INSERT INTO @VAR VALUES ('@Rows_Per_Page', @Rows_Per_Page)
 
DECLARE @Free_Rows_Per_Page FLOAT = FLOOR(8096 * ((100 - @Fill_Factor) / 100) / (@Row_Size + 2) )
INSERT INTO @VAR VALUES ('@Free_Rows_Per_Page', @Free_Rows_Per_Page)
 
DECLARE @Num_Leaf_Pages FLOAT = CEILING(@Num_Rows / (@Rows_Per_Page - @Free_Rows_Per_Page) )
INSERT INTO @VAR VALUES ('@Num_Leaf_Pages', @Num_Leaf_Pages)
 
DECLARE @Leaf_space_used FLOAT = 8192 * @Num_Leaf_Pages
INSERT INTO @VAR VALUES ('@Leaf_space_used', @Leaf_space_used)
 
-- Index Information
DECLARE
	 @Num_Key_Cols FLOAT = 0			-- total number of key columns (fixed-length and variable-length)
	,@Fixed_Key_Size FLOAT = 0			-- total byte size of all fixed-length key columns
	,@Num_Variable_Key_Cols FLOAT = 0	-- number of variable-length key columns
	,@Max_Var_Key_Size FLOAT = 0		-- maximum byte size of all variable-length key columns
 
 
SELECT 
	 @Num_Variable_Key_Cols = COUNT(1)
	,@Max_Var_Key_Size = ISNULL(SUM([DataLength]), 0)
FROM @TableColumn 
WHERE 
	[IsVariableLength] = 1 
	AND [IsKey] = 1
 
SELECT 
	 @Num_Key_Cols = COUNT(1) + @Num_Variable_Key_Cols
	,@Fixed_Key_Size = ISNULL(SUM([DataLength]), 0)
FROM @TableColumn 
WHERE 
	[IsVariableLength] = 0
	AND [IsKey] = 1
 
IF @ClusteredIndexIsUnique <> 0
BEGIN
	SET @Num_Key_Cols += 1
	IF @Num_Variable_Key_Cols > 0
		SET @Num_Variable_Key_Cols += 1
	IF @Max_Var_Key_Size > 0
		SET @Max_Var_Key_Size += 4
END
INSERT INTO @VAR VALUES ('@Num_Variable_Key_Cols', @Num_Variable_Key_Cols)
INSERT INTO @VAR VALUES ('@Max_Var_Key_Size', @Max_Var_Key_Size)
INSERT INTO @VAR VALUES ('@Num_Key_Cols', @Num_Key_Cols)
 
-- Who would seriously have a nullable key column?  ridiculous.
DECLARE @Index_Null_Bitmap FLOAT = 2 + ((@Num_Key_Cols + 7) / 8) 
INSERT INTO @VAR VALUES ('@Index_Null_Bitmap', @Index_Null_Bitmap)
 
DECLARE @Variable_Key_Size FLOAT = 0
 
IF @Max_Var_Key_Size > 0
	SET @Variable_Key_Size = 2 + (@Num_Variable_Key_Cols * 2) + @Max_Var_Key_Size 
INSERT INTO @VAR VALUES ('@Variable_Key_Size', @Variable_Key_Size)
 
DECLARE @Index_Row_Size FLOAT = @Fixed_Key_Size + @Variable_Key_Size + @Index_Null_Bitmap + 1 + 6
INSERT INTO @VAR VALUES ('@Index_Row_Size',@Index_Row_Size )
 
DECLARE @Index_Rows_Per_Page  FLOAT = FLOOR(8096 / (@Index_Row_Size + 2))
INSERT INTO @VAR VALUES ('@Index_Rows_Per_Page', @Index_Rows_Per_Page)
 
DECLARE @Non_Leaf_Levels FLOAT = 1
 
-- Log base @Non_Leaf_Levels hack
WHILE (@Num_Leaf_Pages / @Index_Rows_Per_Page) > POWER(@Index_Rows_Per_Page,@Non_Leaf_Levels)
	SET @Non_Leaf_Levels += 1
 
SET @Non_Leaf_Levels += 1
INSERT INTO @VAR VALUES ('@Non_Leaf_Levels', @Non_Leaf_Levels)
 
-- Summation hack
DECLARE 
	 @Num_Index_Pages FLOAT = 0
	,@Summand INT = @Non_Leaf_Levels
 
WHILE @Summand > 0
BEGIN
	SET @Num_Index_Pages += CEILING(@Num_Leaf_Pages/POWER(@Index_Rows_Per_Page, @Summand))
	SET @Summand -= 1
END
INSERT INTO @VAR VALUES ('@Num_Index_Pages', @Num_Index_Pages)
 
DECLARE @Index_Space_Used FLOAT = 8192 * @Num_Index_Pages 
INSERT INTO @VAR VALUES ('@Index_Space_Used', @Index_Space_Used)
 
DECLARE @Clustered_Index_Size FLOAT = @Leaf_Space_Used + @Index_Space_used
INSERT INTO @VAR VALUES ('@Clustered_Index_Size', @Clustered_Index_Size)
 
SELECT * FROM @VAR

Now for an example of how to run this doodad. This particular estimate was for a staging table - I needed the UNIQUEIDENTIFER as the key and the CreatedTime field had to be part of the clustered index for the partition scheme.

DECLARE @TableColumn TableColumnDefinition
 
INSERT INTO @TableColumn
VALUES 
	 ('UNIQUEIDENTIFIER', NULL, 1)
	,('DATETIME', NULL, 1)
	,('INT', NULL, 0)
	,('INT', NULL, 0)
	,('INT', NULL, 0)
	,('INT', NULL, 0)
	,('INT', NULL, 0)
	,('INT', NULL, 0)
	,('INT', NULL, 0)
	,('INT', NULL, 0)
	,('INT', NULL, 0)
 
EXEC [dbo].[Utility_EstimateClusteredIndexSize]	
 @Num_Rows = 750000		-- number of rows in the table
,@ClusteredIndexIsUnique = 1 	-- is it?
,@Fill_Factor = 90			-- page fill factor
,@TableColumn = @TableColumn	-- definition of table columns

As for what the output looks like, I wanted it to output all the variables used so I could validate along the way with th emanual method. The results look something like this

Name                                                             Value                  DisplayValue
---------------------------------------------------------------- ---------------------- -------------------
@Fixed_Data_Size                                                 60                     60 B
@Null_Bitmap                                                     5                      5
@Num_Cols                                                        12                     12
@Num_Variable_Cols                                               0                      0
@Max_Var_Size                                                    0                      0 B
@Row_Size                                                        69                     69 B
@Rows_Per_Page                                                   114                    114
@Free_Rows_Per_Page                                              11                     11
@Num_Leaf_Pages                                                  7282                   7282
@Leaf_space_used                                                 59654144               56.891 MB
@Num_Variable_Key_Cols                                           0                      0
@Max_Var_Key_Size                                                0                      0 B
@Num_Key_Cols                                                    3                      3
@Index_Null_Bitmap                                               3.25                   3
@Variable_Key_Size                                               0                      0 B
@Index_Row_Size                                                  34.25                  34 B
@Index_Rows_Per_Page                                             223                    223
@Non_Leaf_Levels                                                 2                      2
@Num_Index_Pages                                                 34                     34
@Index_Space_Used                                                278528                 272.000 KB
@Clustered_Index_Size                                            59932672               57.156 MB

So now you may ask "but what about estimating Non-Clustered Indexes and Heaps?" Well, good question as this is the methodology for estimating the size of a Clustered Index. I ran this for a fact table with two non-clustered indexes in addition to the clustered index by basically defining the index (so pretending the index was a table for the purposes of this proc - included columns would just be non-key columns). All in all, it came out +3.31% on 600+ million rows, so I called that methodology good enough