Prologika (Teo Lachev's Weblog)

Where is RDLOM in R2?

2010-03-09T02:17:00Z

After covering the cool new stuff in R2, it's time to talk about the uncool. One R2 change that really vexes me is RDLOM. RDLOM, short for Report Definition Language Object Model, lets developers program RDL in object-oriented way as opposed to using XML. RDLOM could save you a lot of effort when you have to meet more advanced requirements that involve pre-processing RDL. For example, in one of my projects we had to implement a wizard that would walk business users through steps to generate an ad-hoc report. Behind the scenes, the wizard would generate the report definition by using a custom RDLOM which we developed since back then Microsoft didn't have their own. I included a scaled-down version of our implementation in my TechEd 2007 code sample. Although it took us a while to get our RDLOM implemented, it turned out to be invaluable down the road. Since we didn't have to use XML API to manipulate RDL, it simplified programming RDL. For another project, we used RDLOM to let the customer specify which sections they want to see on the report. Once the selection was made, we would remove the unwanted sections from a report template. The list of real-life scenarios that could take advantage of RDLOM goes on …

Apparently, I wasn't the only one bugging Microsoft about RDLOM that ships with SSRS and they promised that this will happen at TechED 2007 USA. Sure enough, SSRS 2008 brought us an unsupported version which I blogged about and covered more extensively in my Applied SSRS 2008 book. Although unsupported, the SSRS 2008 RDLOM was very useful. It was implemented in a single assembly (Microsoft.ReportingServices.RdlObjectModel.dll), exposed all RDL elements as objects, and supported serializing and deserializing RDL.

Alas, this will come to pass with R2. Upgrading to R2, you'll find that the classes in Microsoft.ReportingServices.RdlObjectModel are marked as internal so you cannot use this assembly. Instead, RDLOM got moved into the Microsoft.ReportingServices.Designer.Controls.dll assembly. The good news is that there are rumors that RDLOM may get documented. Now the bad news:

Clearly, R2 RDLOM was not meant to be distributed. While you need to reference explicitly only Microsoft.ReportingServices.Designer.Controls and Microsoft.ReportingServices.RichText assemblies you'll find that they are dependent on other assemblies and you need to distribute a total of 25 assemblies resulting in a whopping 25 MB of referenced code. One way to avoid shipping so many assemblies to deploy your application in the BIDS or Report Builder 3.0 folders.

R2 RDLOM will be 32-bit only because BIDS and Report Builder are 32-bit only.
R2 RDLOM doesn't support serializing the object model. I found a hack that uses .NET reflection to invoke the internal Serialize method of the old RDLOM 2008.

The RDLOM R2 C# console application bundled in this code sample demonstrates how you can use the "new" RDLOM. Besides referencing the new assemblies, you need to use the Load method to deserialize RDL to the object model:

report = Report.Load(fs);

And the above-mentioned hack to serialize it back to disk or stream.

using (FileStream os = new FileStream(ReportPath, FileMode.Create)) {

Assembly assembly = Assembly.GetAssembly(typeof(Report));

object mc = assembly.CreateInstance("Microsoft.ReportingServices.RdlObjectModel.Serialization.RdlSerializer");

Type t = mc.GetType();

MethodInfo mi = t.GetMethod("Serialize", new Type[] { typeof(Stream), typeof(Report) });

report = (Report)mi.Invoke(mc, new object[] { os, report });

}

Other than that, RDLOM remains virtually unchanged from 2008 excluding that it now supports the R2 enhancements to RDL. What if you don't need this overhead but you need to make small changes to RDL that don't necessarily require an object-oriented layer. I've found that the easiest way to do this and avoid XML programming is to use LINQ to XML. The LINQ to XML project included in the same sample demonstrates this approach. For example, the following code changes the data source reference of the first data source.

static void Main(string[] args) {

XElement report = XElement.Load("report.rdl");

string dns = "{" + report.GetDefaultNamespace() + "}";

XElement ds = report.Element(dns + "DataSources").Element(dns + "DataSource").Element(dns + "DataSourceReference");

ds.Value = "/Data Sources/AdventureWorksAS2008";

report.Save("report1.rdl"); }

Not happy about the RDLOM story? Join the movement and vote for my suggestion on connect.microsoft.com to have a supported RDLOM.

How Windows UAC Got Me Again

2010-03-02T21:17:00Z

User Access Control (UAC) is a feature of Windows Vista, Windows 7, and Windows Server 2008 that prevents you from running applications with administrator rights even if you are a local admin on the box. Although I have to admit I hate UAC, I am convinced that it's better to have it around especially after my home desktop got infected a few months ago with a rootkit virus so badly that I had to restore the disk from a backup. So, UAC is a necessary evil. But since it's relatively new, not many application developers test for UAC and this can get you as it happened in the following story.

A coworker complained that Reporting Services SharePoint integration doesn't work on a distributed environment where SharePoint is installed on the front end Windows 2008 server and the SharePoint configuration database is installed on a separate database server. As usual, my first stop for troubleshooting SharePoint integration with Reporting Services was to examine the log file which gets generated after installing the Reporting Services Add-in for SharePoint, as I explained in this blog. On Windows Server 2008, the log file is located in the Temp folder of the user running the add-in setup, such as C:\Users\<user>\AppData\Local\Temp. To make the quest even more interesting, the AppData folder is actually hidden by default on Windows Server 2008 and Windows Vista/7 (another security gotcha), so you have to use Folder and Search Options to enable hidden files so you can see it. Are you having fun yet?

In this case, the log file showed an error:

User: SYSTEM

Installing Report Server feature.

Calling copyappbincontents command.

copyappbincontents command completed successfully.

Adding Report Server feature to farm.

Install error: The EXECUTE permission was denied on the object 'proc_putObject', database 'IWSharePointCluster_Config', schema 'dbo'.

So, SharePoint can't connect to the database server. What's interesting is that the log file reported that the user is SYSTEM and not the logon of the interactive user. Luckily, Prash Shirolkar from the SSRS team blogged about this and provided a workaround for this issue which will get fixed in R2 RTM. Gotta love UAC…

New Reporting Services Data Providers

2010-02-28T15:39:00Z

Two new data providers will be included in the SQL Server 2008 R2 release of Reporting Services, Microsoft SQL Azure and Microsoft SQL Server Parallel Data Warehouse, that will let you report from SQL Azure and SQL Server 2008 R2 Parallel Data Warehouse (code named Madison) databases respectively. I decided to take the Microsoft SQL Azure data provider for a spin. As you probably know by now, SQL Azure is a part of the Microsoft Windows Azure SaaS cloud offering. SQL Azure lets you host your SQL Server relational databases in the Microsoft data centers that provide scalability and fault tolerance services. This is great for companies of all sizes which don't have resources or expertise to purchase and configure expensive hardware and software in order bo build scalable and highly available solutions. IMO, SQL Azure will be increasingly important, especially after Reporting Services and Analysis Services become available on SQL Azure although Microsoft doesn't say yet when this will happen. At this point, you can only use Reporting Services as a client to author reports from SQL Azure-hosted databases.

Strictly, speaking if your database is hosted in SQL Azure, you can use the SQL Server or OLE DB providers (already included in Reporting Services) to send queries to SQL Azure. Just like on-premise solutions, SQL Server Azure clients retrieve data using the native SQL Server Tabular Data Stream (TDS). However, if you decide to use the native SQL Server providers, you will run into some issues with retrieving SQL Azure metadata. Hence, Microsoft added the SQL Azure data provider. To test it out, I followed these steps:

I signed up for SQL Azure. If you MSDN Universal subscriber, you can sign up to test SQL Azure for eight months for free.
I deployed the AdventureWorksDW2008R2 and AdventureWorksLT2008R2 databases to my SQL Azure server. The Adventure Works database deployment scripts for SQL Azure are available on codeplex. They will create AdventureWorksDWAZ2008R2 and AdventureWorksLTAZ2008R2 database on your SQL Azure server.
Installed SQL Server 2008 R2 Release Candidate which is not publicly available.
Created a new report in BIDS. As usual, I added a data source and selected the Microsoft SQL Azure provider.
In the Connection Properties dialog box, I entered the name of my SQL Azure server and my standard security credentials (Windows security is not supported on SQL Azure). For some reason, the database dropdown is not automatically populated so I had to enter the name of the database AdventureWorksDWAZ2008R2 manually.
Once connectivity has been taken care of, life is as usual. The Graphical Query Designer (toggle the Edit as Text button if the generic query designer starts first) shows the database schema.

One thing that surprised me was the performance of data retrieval. Even when I opted to get all data from FactInternetSales, which has thousands of rows, the query executed pretty fast. One caveat though with this approach is that you have open port 1433 on your corporate network because this is the port SQL Server uses for client connections. This is not required if your application is installed on Windows Azure and connects to the SQL Azure database on the server.

Beware Hidden Calculated Members

2010-01-18T23:06:07Z

Here is another one of those "by design" issues (see the Not All Calculated Members Are Born Equal blog for another recent gotcha) that are not documented but can sure waste you precious time figuring out. The issue was first submitted on connect.microsoft.com by Greg Galloway almost a year ago where it was left in an investigation phase (what can we do to get the SSAS team actually review the connect cases and post findings?). Then, I submitted a similar issue and official support case since I was 100% sure it was a bug and I needed a closure. Alas, Microsoft came back to declare it "by design":

"…per feedback from several AS 2005 Beta customers, for compatibility with pre-existing AS 2000 behavior, AS 2005 RTM (and later releases) was changed to not include hidden calculated members in cell calculations and scopes (except when a hidden calculated member is explicitly referenced in the calculation scope)."

The issue affects hidden calculated members and scope assignments where a scope assignment simply ignores hidden calculated members if they are not explicitly referenced by the script. You can easily repro the issue by making the following changes to the Adventure Works script:

/*-- Allocate equally to quarters in H2 FY 2005 ------------------------*/

CREATE MEMBER CURRENTCUBE.[Measures].[Visible Quota] AS null,

FORMAT_STRING = "Currency",

VISIBLE = 1 , ASSOCIATED_MEASURE_GROUP = 'Sales Targets' ;

CREATE MEMBER CURRENTCUBE.[Measures].[Invisible Quota] AS null,

FORMAT_STRING = "Currency",

VISIBLE = 0 , ASSOCIATED_MEASURE_GROUP = 'Sales Targets' ;

Scope

(

[Date].[Fiscal Year].&[2005],

[Date].[Fiscal].[Fiscal Quarter].Members

);

this = (ParallelPeriod

(

[Date].[Fiscal].[Fiscal Year], 1,

[Date].[Fiscal].CurrentMember

), [Measures].[Sales Amount Quota]) * 1.35;

Here, I created two identical calculated members with the only difference that the first one is visible and second one hidden. Then, execute the following query:

SELECT {[Measures].[Sales Amount Quota], [Measures].[Visible Quota], [Measures].[Invisible Quota]} on 0

FROM [Adventure Works]

WHERE [Date].[Fiscal].[Fiscal Quarter].&[2005]&[1]

And here are the results (surprise!):

The Visible Quota member is populated as it should be but Invisible Quota refuses to budge as the script simply ignores it. If the script scopes on the Invisible Quota member explicitly or you make it visible, then we get the expected results. Since the issue is by design and a fix won't be available anytime soon, you should be very careful when you use hidden calculated members and expect them to be affected by scope assignments. I use often helper calculated members which I hide since they shouldn't be visible to the end users. This issue hit me really hard after QA reported wrong results and your humble correspondent had to debug a cube with some 1000+ lines of code.

Report Builder 3.0 Edit Sessions

2010-01-18T14:25:00Z

When Report Builder 2.0 shipped , it introduced local and server preview modes. If the report uses local references only, such an embedded data source(s), Report Builder 2.0 executes the query and renders the report locally just like the BIDS Report Designer. However, when the report uses a server reference, such as shared data source(s) or subreports that were saved on the server, Report Builder 2.0 would transparently upload and process the report on the server just like the Report Builder 1.0.

Report Builder 3.0 improves the performance of server preview mode by introducing edit sessions. An edit session is implicitly created when the user runs the report for the first time. Behind the scenes, Report Builder 3.0 calls the new CreateReportEditSession API to instruct the server to create an edit session. The server caches the report dataset(s) in the ReportServerTempDB database (not in memory) and uses the cached data for subsequent previews. In this respect, you can think of edit sessions as the *.data files that the BIDS Report Designer generates to cache the report data in order to speed up report preview. As long as you don't make changes to the report that affect the data, such as changes to the dataset itself, data source, or parameters, the cached copy of the data is used by the report. This means that the user may not see changes that might have taken place in the database for the duration of the edit session. To make sure that the latest data is used, click the Refresh button on the Report Builder 3.0 toolbar. Refreshing the report discards the edit session and executes the database queries.

The administrator can configure two new server properties to manage edit sessions:

EditSessonCacheLimit - By default, the data cache can hold up to five datasets per report and user. Note that if the dataset is parameterized, each parameter combination counts as one dataset. If you use many different combinations of parameter values, the report might need more cached dataset copies. The administrator can allow this by increasing the EditSessonCacheLimit setting.
EditSessionTimeout - By default, an edit session lasts 7200 seconds (2 hours). The servers resets the session every two hours when the user previews the report.

Report Builder 3.0 edit sessions are a performance-related enhancement that makes server-side report preview more efficient.

Report Builder 3.0 Enhancements in R2 and SharePoint 2010 Mode

2010-01-13T03:06:00Z

A couple of Report Builder enhancements in SharePoint 2010 integration mode caught my attention. First, Report Builder 3.0 now launches by default when the Open in Report Builder action is initiated inside SharePoint 2010. Previously, Report Builder 1.0 would be launched.

Second, if the report includes a drillthrough action to another report, the action now works when the action is initiated from Report Builder 3.0 preview mode. As it turns out, a drillthrough action is forwarded directly to SharePoint. Previously, a report drillthrough action will result in an error because the Report Builder would attempt to resolve the action locally.

ReportService2010 Endpoint

2010-01-13T02:29:57Z

Here is a little gem for developers programming Reporting Services. SQL Server 2008 R2 brings a new web service endpoint, ReportService2010, which merges the functionalities of the ReportService2005 and ReportService2006 endpoints. The new endpoint can manage objects on a report server that that are configured for either native or SharePoint integrated mode. Previously, the ReportService2005 endpoint was used for managing objects on a report server that is configured for native mode and the ReportService2006 endpoint was used for managing objects on a report server that is configured for SharePoint integrated mode. This required code changes and retesting if the report server was re-configured in a different mode. These endpoints are now deprecated but still supported. The ReportService2010 endpoint includes the functionalities of both endpoints. When unsupported APIs in ReportService2010 are invoked, such as invoking the CreateLinkedReport API in SharePoint mode which doesn't support linked reports, the endpoint will return an error that the operation is not supported.

Virtual Adventures

2010-01-07T23:48:58Z

Scenario: Mount base and differencing VHD disks using Windows Virtual PC on Windows 7 x64.

Dude, did you run out of BI stuff to talk about? I didn't but I think other folks may benefit from my experience especially MCTs who have discovered that Windows 7 hasn't happened yet to the Microsoft Official Curriculum and resorted to all sorts of hacks to get Virtual Server 2005 or Lab Launcher working with Windows 7. In the process, I've learned a lot about the virtualization technology that goes beyond just mounting disks. OK, I have a hidden agenda to write this blog as well – I want to document these steps.

So, I am preparing to teach 6326A - Implementing and Maintaining Microsoft SQL Server 2008 Reporting Services in an instructor-led environment where I have to install the required software on the student's machine. After downloading the trainer resources, I have discovered that Microsoft provides 6236A-NY-SQL-01.vhd disk which requires a base disk Base08A.vhd. Reading the setup guide included in the course revealed that Microsoft wants me to install Virtual Server 2005 on both the instructor and students machines.

Gotcha 1: Virtual Server 2005 doesn't run on Windows 7. Browsing the MCT private groups I came across posts discussing a hack around this limitation but I didn't want go that way. Not to mention that Microsoft Lab Launcher, which apparently is intended to simplify the VHD deployment to student machines, doesn't run on Win 7 as well. A double-gotcha which picked up my interest and stirred my hacking instincts to get the whole thing working with Virtual PC. After all, I can't control what OS the students run or the training center has installed.

OK, but this is just VHD disk, isn't it? So, it should run on Windows Virtual PC, right? So, I created a new virtual machine and attempted to mount 6236A-NY-SQL-01.vhd but I was greeted with the following informative error:

Cannot attach the virtual hard disk to the virtual machine. Check the values provided and try again.

Then I mounted Base08A.vhd and I was able to boot the VM only to find out that it includes the Windows Server 2008 OS only. Where is the cool SQL Server BI stuff? This is the part where I realized that my virtual knowledge needs brushing up. Enter differencing disks. This of course is not explained in the setup document, so don't try to find it there. I figured that Base08A.vhd is the base disk that has the OS only while the 6236A-NY-SQL-01.vhd is the differencing disk that has the rest (SQL Server 2008 + Reporting Services samples and Adventure Works database) installed. To prove my hypothesis, I used the Sun's VirtualBox. I use VirtualBox for my personal VM work because it supports x64 guess OS, which Virtual PC doesn't support (shame). I used the Disk Manager to add both the base and difference images and … lo and behold, the VM booted up and I was able to verify that SQL Server 2008 and rest of the goodies are installed.

But I didn't want to force the students to install VirtualBox so I continued hacking my way to Windows Virtual PC. In a Eureka moment, I examined the Windows Event log and saw the following message:

The parent virtual hard disk 'C:\Program Files\Microsoft Learning\Base\Base08A.vhd' for the differencing virtual hard disk 'C:\VPC\3263\6236A-NY-SQL-01.vhd' does not exist. Please reconnect the differencing virtual hard disk to the correct parent virtual hard disk.

So, the base disk path is hardcoded in the differencing image. The following commands confirmed this.

Went to command prompt and entered (the actual commands are in bold):
C:\>Diskpart
DISKPART> select vdisk file="c:\vpc\6236\6236A-NY-SQL-01.vhd"
DiskPart successfully selected the virtual disk file.
DISKPART> detail vdisk
Device type ID: 2 (VHD)
Vendor ID: {EC984AEC-A0F9-47E9-901F-71415A66345B} (Microsoft Corporation)
State: Added
Virtual size: 64 GB
Physical size: 9 GB
Filename: c:\vpc\6236\6236A-NY-SQL-01.vhd
Is Child: Yes
Parent Filename: C:\Program Files\Microsoft Learning\Base\Base08A.vhd
Associated disk#: Not found.
I moved the based image (Base08A.vhd) to C:\Program Files\Microsoft Learning\Base. With a baited breath, I repeated the process of creating a new virtual machine in Windows Virtual PC pointing to the existing 6236A-NY-SQL-01.vhd but I got the first error:
Cannot attach the virtual hard disk to the virtual machine. Check the values provided and try again.
No more details in the Windows event log. At this point, I posted a question to the Windows 7 Virtualization discussion forum and I was clued in the disk virtual size (not physical size) could be the issue since Virtual PC doesn't support disks larger than 127 GB. Although after attaching the 6236A-NY-SQL-01.vhd using the Windows 7 Disk Manager I could see that the virtual size is 64 GB, I compacted the base disk using these steps. That didn't help and the hour was getting late. So I threw the white towel on the differencing disk approach (for now) and decided to merge the differencing disk into the base disk.
Assuming the commands in step 1), I executed the following commands:
DISKPART>detach vdisk
DiskPart successfully detached the virtual disk file.
DISKPART>merge vdisk depth=1
100 percent complete
(after a long wait …)
DiskPart successfully merged the differencing chain.

Now that I had the disks merged I was able to create a new virtual machine successfully using the base disk image (Base08A.vhd) which combined both the original and differencing disks. Since the differencing disk was merged, I could delete 6236A-NY-SQL-01.vhd.

Gotcha 2: After booting the virtual machine, I was able to log in using the pre-defined Student account. However, attempting to run the SQL Server Management Studio, I got an error that the SQL Server 2008 evaluation period has expired. I attempted to reset the Windows Server 2008 (guest OS) system date, but after a few seconds the date would revert to the present date. More binging the Internet solved the issue.

Shut down the VM.
Open the vmc file in Notepad.
Chang the last six digits to a year ago (070109)
<time_bytes type="bytes">57002400060004070109</time_bytes>
Further down the file, disable guest-to-host OS time synchronization:
<host_time_sync>
<enabled type="boolean">false</enabled>
<frequency type="integer">15</frequency>
<threshold type="integer">10</threshold>
</host_time_sync>
Boot VM. Now change the system date to 07-01-09.

Success!

Most Requested Features

2009-12-21T21:06:51Z

You can use the Microsoft Connect website to find most requested features. Unfortunately, the search doesn't let you specify a product so the search results may be related to other products. For example, searching on reporting services may bring in results from Analysis Services and reporting. Nevertheless, it was quite interesting to find the top voted suggestions. For example, the following query shows the top suggestions for Reporting Services (flip to the Suggestions tab):

https://connect.microsoft.com/SQLServer/SearchResults.aspx?KeywordSearchIn=2&SearchQuery=%22reporting%22+AND+%22services%22&FeedbackType=1&Scope=0&SortOrder=10&TabView=0&wa=wsignin1.0

Reporting Services-Recognize multiple result sets returned from a stored procedure – (50 votes)
Merging / Linking datasets on report level (50 votes) – No 6 on my SSRS Top 10 wish list.
SQL Reports should support stylesheets (43 votes) – No 9 on my list.
Support for XML Paper Specification (XPS) Output Format (29 votes) – I am personally surprised about this one.
Reporting Services Security Using Membership and Roles (29 votes), and so on

Running ReportViewer Under Trusted Account

2009-12-16T00:33:00Z

Scenario: You have an Internet-facing web application that uses the Visual Studio ReportViewer control to render reports. The application takes care of authenticating the users. You want all report requests to the report server to go under a single (trusted) Windows account. There are at least two approaches to implement this scenario:

Basic Authentication

You can configure ReportViewer for basic authentication by implementing the Microsoft.Reporting.WebForms.IReportServerCredentials interface, as follows:

[Serializable]

public class MyReportServerCredentials : IReportServerCredentials

{

public MyReportServerCredentials()

{

}

public WindowsIdentity ImpersonationUser

{

get

{

return null;

}

public ICredentials NetworkCredentials

{

get

{

return new NetworkCredential(

ConfigurationManager.AppSettings["reportUser"],

ConfigurationManager.AppSettings["reportPassword"]

, "<your domain here>" – comment out this line if a domain account is used as a trusted account

);

}

public bool GetFormsCredentials(out Cookie authCookie, out string user, out string password, out string authority)

{

authCookie = null;

user = null;

password = null;

authority = null;

return false;

}

Specifically, you need to implement the NetworkCredentials property of IReportServerCredentials. In my case, I read the trusted account credentials from web.config. Once IReportServerCredentials is implemented, you need to configure ReportViewer to use it. This takes one line of code:

reportViewer.ServerReport.ReportServerCredentials = new MyReportServerCredentials();

The advantage of using IReportServerCredentals if flexibility. You can retrieve the account credentials from any place, such as security service, configuration file, database, etc. In addition, the impersonation is scoped for ReportViewer only. Other network calls made by the applications are unaffected.

Process Identity

The issue with IReportServerCredentals is the application must deal with the credentials of the trusted account, including the password. For added level of security, I try my best to avoid handling passwords and use Windows authentication whenever I can. Therefore, my preferred approach is to change the process identify of the web application. One way to do this is to change the identity of the IIS application pool the web application is running under, as follows:

If you don't have a domain account to use as a trusted account, create a new domain account.
On the web server, add the account to the IIS_WPG local Windows group to give it the necessary rights to be used as an application pool account.
In the IIS Manager, expand the server name, right-click the Application Pools folder, and click New Application Pool.
Give the new pool a name. In the pool Properties page, click the Identity tab, select the Configurable option, and enter the credentials of the trusted account.
Open the web application properties and change its application pool to the pool you just created.
If present, remove <identity impersonate="true"/> element from the application web.config file so the network calls are made under the process identify instead of the anonymous user account, such as IUSR_<MACHINE_NAME>.
At this point, if you run the web application, you may get a permission error because the domain account doesn't have the necessary rights to specific folders, such as the web application folder and the temp folder. Using Windows Explorer, grant the domain account (or IIS_WPG group) rights to these folders. The error message should include the path of the folder you need to grant access to.
Finally, use the Report Manager to grant the necessary rights to the trusted account, such as Browser, to the required folders.

Infragistics Silverlight for Analysis Services

2009-12-09T22:11:39Z

Don Demsak pointed out another Silverlight-based control from Infragistics, which is currently in a CTP phase. There are actually two controls: XamWebDataChart and XamWebPivotGrid. I played a bit with it and it looks great. I liked the Excel-like paradigm with the metadata pane on the right and the PivotChart and PivotTable synchronized. As I've been saying – give me a Silverlight-based Excel Pivot and I am happy.

See my previous blog about the Intelligencia for Silverlight which is another Siliverlight-based control for Analysis Services.

Analysis Services Top 10 Wish List

2009-12-07T18:27:00Z

Analysis Services leads the OLAP server market and it should be your platform of choice for historical and trend reporting. In the spirit of the season, here is my Top 10 Wish List for Analysis Services. All items are related to regular cubes and none to PowerPivot which I am yet to try.

Improved client support – No 1 on my list has nothing to do with the server itself but with how well the Microsoft-provided clients support Analysis Services. To its credit, Microsoft continues improving the Excel OLAP support but still there are major functionality gaps from one tool to another. This is especially evident in the area of Reporting Services (see my No 1 SSRS wish list item). I am not sure if that's possible from a technical standpoint but I wish the SSAS team re-factors the query generation logic into a MDX Query Generator binary of some kind which all tools can share to produce the same results consistently.
Focus on UDM (aka Corporate BI) – Although I see some useful scenarios for self-serviced OLAP with PowerPivot (previously known as Gemini), for various reasons I believe the focus should stay on UDM which now Microsoft refers to as Corporate BI. First, the business problems that I solve with UDM exceed by far Gemini's capabilities. Second, if not a single source for reporting and a single version of the truth, I believe UDM should handle 80-90% of the reporting needs. Finally, based on my experience, data acquisition is the most difficult thing. This is why we have a dimensional model which IT would put together. But once data is cleaned and transformed, building UDM on top of it shouldn't take that long
Silverlight OLAP Viewer control - A Silverlight-based control that ships with Visual Studio to let developers embed an Excel-like Analysis Services browser into their ASP.NET applications.
Improved performance – Version 2008 has brought significant performance enhancement and SSAS performs great when aggregating regular measures. Usually, it is the calculations that will get you in trouble. So, anything that can be done to improve the server performance further will be welcome. For instance, the calculation engine executes the query on a single thread. It will be nice if a more complicated query can be parsed and executed in parallel.
MDX Query Analyzer – Related to performance, it will be great if we finally get a graphical query analyzer (a-la the T-SQL query analyzer) with hints about how to improve the query execution.
Memory-bound partitions – Again related to performance, it will be great if you could configure a partition to be memory-bound for super-fast access, as I mentioned in this blog.
Detail reporting – I wish Analysis Services improves the ability to report on detail level. For example, it will nice to support text measures that the server returns for the lowest grain. I also want detail queries, such as a query that returns accounts for given customer to perform faster and don't give up with "The expression contains a function that cannot operate on a set with more than 4,294,967,296 tuples." error.
Resource governor – Analysis Services should expand on letting the administrator manage resources, such as memory, connections, query timeouts, etc.
Custom security – Support custom authentication that is not Windows-based to facilitate Internet connectivity perhaps similar to custom security in Reporting Services.
Modeling enhancements – OK, this is a catch-all bucket for support issues that I logged and which were declared as "by design" and "future enhancements" , such as aggregatable Many-to-Many relationships, fixing subselect issues, fixing exclusion filter with custom operator issue.

Happy holidays!

Reporting Services Top 10 Wish List

2009-12-07T13:55:00Z

Reporting Services has gone a long way for the past five years and it's getting better. In the spirit of the season, here is my Reporting Services Top 10 Wish List revised from three years ago. This list is based on my work, so your priorities may differ.

Improved Analysis Services integration – Since its debut in SSRS 2005, the MDX query designer hasn't changed. R2 continues the trend, so no hope until at least 2011, that's six years! Yet, the MDX query designer has plenty of deficiencies and bugs to be addressed, such query schema limitations, poor support for parent-child hierarchies, parameter limitations, and so on. I use SSAS heavily in my real-life projects and have to battle these deficiencies on a daily basis. I think it's time for Microsoft to poll customers and MVPs for feedback and improve the integration with SSAS.
Eventing model – Although SSRS 2008 introduced a Report Definition Customization Extension (RDCE) that lets you change the report definition when the report requested, this extensibility mechanism was added as an afterthought and it's somewhat kludgy. What's really needed is server-side events, such as OnReportRender, OnPrint, OnParameterLoad, etc., similar to the ASP.NET programming model. Imagine the flexibility you will have as a report author if the server raises events for various stages of report processing and passes the report definition to let you evaluate conditions and change collections and RDL on the fly.
Supported RDL Object Model – After promising an RDL object model at TechEd 2007, SSRS 2008 brought an unsupported version but R2 is on its way to "undo" it by making it publicly inaccessible. At the same time, there are many requirements that call for pre-processing report definitions. A supported RDL object model will definitely simplify this.
Silverlight-based Report Viewer – Silverlight established itself as a platform of choice for web-based development. A Sliverlight Report Viewer will be a welcome addition.
More interactive reporting – Currently, SSRS requires you to make design changes and preview them. It will be great if a future release blurs design and preview to support more interactive reporting similar to Excel PivotTable.
Dataset enhancements, such as ability to navigate dataset rows and joining datasets at report level.
More user-friendly ad-hoc reporting – Report Builder 2.0/3.0 is one my favorite features that SSRS 2008/R2 brought in. However, some end-user oriented features from Report Builder 1.0 got "lost" along the way. This item is about these features, such as making it easier for the end user to filter report data. When connected to SSAS, for example, SSRS should be as user-friendly as Excel.
ADO.NET dataset binding for server reports – I keep on asking for this feature and I still think it could be very useful to be able to pass ADO.NET datasets to published reports without having to build a custom dataset extension.
Report styling – Ability to style and skin reports.
Decoupling the Report Builder Designer and exposing it as a reusable .NET control which can be embedded in applications.

Happy holidays!

Microsoft Live Labs Pivot

2009-12-02T20:39:45Z

In case you've missed this, the Pivot era has begun. After Excel PivotTable and PivotChart, we'll have PowerPivot in SQL Server 2008 R2. But Pivot evolves… A co-worker showed me today a glimpse of the Pivot future which I guess is the Microsoft Live Labs Pivot. Since grids and charts are not cool anymore we now have pictures and animation. It's hard for me to understand at this point how this would apply to Business Intelligence but the Silverlight app with all these pictures sure looks catchy.

Long live Pivot!

Expert Cube Development with Analysis Services 2008

2009-12-02T14:00:00Z

Now that I've got my copy and read it, I can say a few things about the book Expert Cube Development with Microsoft SQL Server 2008 Analysis Services by Chris Webb, Alberto Ferrari, and Marco Russo. As mentioned in another post, the authors (all MVPs and prominent SSAS experts) don't disappoint and you won't go wrong picking up this book. The book is somewhat small in size (330 pages) but big in wisdom and best practices. It also includes plenty of references to external resources and blogs that the authors and other SSAS experts have written when you need further understanding of the topics discussed. Make no mistake though, this is not a step-by-step book and it doesn't target novice users. As its title suggests, the book is geared toward more advanced users who have a few years of SSAS experience under their belt and are thirsty for more insightful knowledge.

Chapter 1 Designing the Data Warehouse for Analysis Services

The first and most important step for implementing a successful Analysis Services solution has nothing to do with Analysis Services at all. Its success depends on a solid schema design on which you'll build the cube so the book is right on target to hit this topic first. Chapter 1 discusses common schema modeling challenges and provides plenty of tips to tackle them. One grudge I have here is about the following sentence "Beware of the temptation to avoid building a data warehouse and data marts" which may leave the reader with the impression that a data mart is a must. In fact, call me lazy but the first thing that I'd explore when starting a new SSAS project is if it's possible to avoid a separate database. While a classic OLAP solution in many cases will require a new database (whatever you call it a data warehouse or a data mart) and the whole ETL enchilada that this entails, you may need to build a cube on top of simpler schemas, such on top of a DSS (decision support system) schema. In this case, I would explore the option to "dimensionalize" the source schema with a layer of SQL views. As with everything else, the keep-it-simple principle works for me.

Chapter 2 Basic Dimensions and Cubes

This chapter discusses the groundwork required to set up the raw dimensional model. This includes working with data sources, data source views, cubes, and dimensions. My favorite part of this chapter was the attribute design and setting up attribute relations. This often overlooked aspect of SSAS development is crucial for achieving good performance and the book explains why.

Chapter 3 Designing More Complex Dimensions

Next, the book explains how to handle more complex dimension schemas and hierarchies. This includes parent-child, ragged, slowly-changing, referenced, fact, and junk dimensions. The authors' real-life experience shows through. About junk dimensions, my preference is to avoid them wherever possible for performance reasons. Instead, I'd explore the option to keep the low-cardinality attributes in the fact table and use report or URL actions to view them. In my opinion, SSAS is not there yet as far as reporting on details.

Chapter 4 Measures and Measure Groups

The whole purpose of building a cube is to aggregate measures. Analysis Services supports various aggregation functions and the book provides essential coverage. You will learn how to implement semi-additive measures if you cannot afford the Enterprise Edition of SQL Server. Custom aggregation by using unary operators and custom member formulas are presented too. Then the book discusses more complex relationship types, including different grains, linked dimensions and measure groups, role-playing dimensions and fact relationships.

Chapter 5 Adding Transactional Data

Although most users would be interested in aggregated data, sometimes users may need to see the level of details behind a cell. The book explains how you can do this by using drillthrough and actions. Drillthrough options are covered in great details. Excellent coverage is provided for many-to-many relationships as well.

Chapter 6 Adding Calculations to the Cube

Cubes usually include business calculations in the form of MDX expressions. Chapter 6 gives lays the foundation for using MDX. It provides practical examples for calculated members, including YTD, ratios, growths, same period last previous year, moving averages, and ranks. Handling relative dates is a common business requirement. For example, I personally had to implement two custom types of relative dates as explained in this blog. The book provides a great example for implementing a custom calculation dimension for implementing more flexible relative dates that the BI Wizard provides. Finally, the chapter discusses static and dynamic named sets.

Chapter 7 Adding Currency Conversion

I consider myself lucky for never having to support international users because of the complexities surrounding the issue. If you have to, SSAS gives you the necessary features to tackle various localization requirements. This chapter walks you through the details of implementing currency conversion using the BI wizard, measure expressions, and writeback. My advice is to avoid writeback due to its limitations. Instead, consider an URL action that navigates the user to a custom web page for handing data changes and proactive caching for automatically updating the cube.

Chapter 8 Query Performance Tuning

SSAS bends backwards to provide great performance. You can help by following best practices for cube design. This chapter teaches you how to implement partitions and aggregations. I like the sections for monitoring partition and aggregation usage and building aggregations manually. The chapter presents useful tips to diagnose and resolve performance issues, such as by using the MDX Studio, named sets, calculated members, and caching.

Chapter 9 Securing the Cube

Chances are you will need to provide restricted access to the cube for groups of users. This chapter shows you how to implement and test roles. It covers cell and dimension security. It presents solutions for implementing custom dimension data security which are similar to the ones I discussed in this article.

Chapter 9 Productionization

Your SSAS solution is ready and it's time to see daylight. This chapter covers the tasks associated with deploying and managing your SSAS solution in production environment. These include managing and generating partitions and processing. I liked the coverage of different processing options.

Chapter 10 Monitoring Cube Performance and Usage

As an administrator, your job doesn't end once the solution is deployed. You need to monitor it on a daily basis to ensure that it performs optimally. The last chapter explains different server monitoring techniques, such as performance counters, tracing, Resource Monitor, SQL Profiler, and using Dynamic Management Views (DMVs). I personally liked the memory management tips.

In summary, I highly recommend this book to anyone who has several years of experience in SSAS development and need a compass check to make sure that he follows the right track. Kudos to the authors for taking time to write it!