The Microsoft ISV Developer Evangelist Team blog recently announced a free all-day training event covering Windows Azure:

In response to overwhelming feedback, we’ve organized a special, all-day FREE Windows Azure Firestarter event April 6, 2010, 8:30 AM – 5:00 PM (PT) at the Microsoft campus in Redmond, Washington to help developers learn more about how they can take full advantage of cloud services.  This one-of-a-kind event will focus on helping developers get ‘cloud ready’ and will include sessions about Microsoft’s strategic vision for the cloud, the end-to-end view of the Windows Azure platform from a developer’s perspective, as well as how to migrate your data and existing applications (regardless of platform) onto the cloud.

If you’ve attended a MSDN Windows Azure Event or watched a Windows Azure Webcast, this event is the perfect next step for you.  Not only will you get trained on Windows Azure, you’ll have a great opportunity to hear from some of our top Windows Azure experts and ask them questions about how to get started in the cloud with Windows Azure.

Seating is limited, so please register today to attend this exciting live event.  If you cannot make it to Redmond, the entire event will be simulcast online; registration is required to attend the online event, so sign up today to ensure you don’t miss a minute of this exciting event!

It sounds like it could be a good opportunity. You can sign up here.

Traditionally, as stated in the previous post, when Microsoft is referring to the three-screen concept they are talking about PC, mobile, and TV. The actual devices behind each of these can vary. In fact I would go so far to say that the devices themselves are not the important component. How a user interacts with the device and how the device is connected to the cloud ultimately should determine how it is categorized within the concept. So, let me start with the PC and what that screen represent (at least in my opinion).

When you’re targeting the PC as one of the “screens” it is common to think of it as a computer with a monitor, keyboard, and mouse. If, however, you look at it as a primarily stationary device that is always connected and where users actively interact with it in close proximity of the display it starts to open up to more devices beyond the classic desktop PC. So, what kind of devices are we talking about?

How about a multi-touch PC mounted in a kitchen? A user walks up to it and interacts directly with the screen as opposed to a keyboard and mouse. But, the interaction takes place within the same general proximity as a traditional desktop. So, from a display perspective it needs to be optimized for close viewing.

An end-user is not going to sit down and type a novel on this device. Instead, they would use it to display recipes, maintain a family calendar, or watch video with a few taps of the finger. Yet, this is essentially still a PC and I do not think many would argue against it falling into the PC screen category. But, it does start the move away from what is thought of as the typical PC. And, I hope it starts you thinking about other possibilities as well. So, let us move a little bit further from the traditional PC again. How about a Microsoft Surface device?

Most of the Surface devices that have shown up in the wild have been in hotel lobbies, bars, and retail stores. And, for the most part they have been little more than a novelty. But, there is reason to believe that there can be more to this than what we have seen so far.

Several months ago I watched a video (of course I can’t seem to find a link to it now) where Microsoft demoed a Surface desk that was tailored to a factory supervisor. The interface itself was more like a sophisticated dashboard than traditional GUI and was fully multi-touch. On this dashboard the supervisor was able to see at a glance a schematic of the factory floor. Individual equipment on the schematic was color coded to indicate if there was a problem. For example, maybe a line was not meeting the production quota that is has been assigned. Or, perhaps maintenance had been recently skipped on the machinery. With a few taps the supervisor is able to see historical performance information as well as future data such as employee shift schedules.

The demo continued by showing additional cool features of the Surface device. For instance, the supervisor was able to perform and enterprise search for an electronic document by simply placing the hardcopy face down on the desk surface. All of this is very interesting and shows some real value.

Ultimately, the interface was seamless, customizable, and geared directly towards complementing the supervisor’s natural workflow. Here again, we have many of the same concepts of the traditional PC. But, it is really starting to open up some exciting possibilities; especially for BI. Certainly there are some big differences between this and the display that sits on your desk. Yet, I’d contend that this is still the same “screen”.

Microsoft has been touting their technology vision of three-screens and the cloud for about a year now (much longer if you don’t include the formal name). For those not familiar with the concept it is essentially a vision that embraces the union of technologies across the PC, mobile, and TV. The common connector in this vision is usually the cloud.

Over the last several months the bits and pieces that Microsoft has shown of this vision have been fragmented and lacked a sense of completion. However, with the recent announcement of the Windows Phone 7 things are starting to come together; at least on the consumer side of things.

This week the GDC (Game Developers Conference) is taking place and Microsoft is showing off the gaming portion of their vision with announcing XNA Game Studio 4. One of the demos that they recently presented showed a game that worked seamlessly across the PC, Xbox 360, and Windows Phone. Even more impressive than simply having a game that can be ported across these platforms is that the game was able to push it’s save state to the Cloud. This allows the end user to pick up where they left off across all of their devices.

This is certainly cool and is exactly the type of thing that gets my imagination going. I innately have a curiosity for all things dealing with technology. And, when I see something like this my mind immediately starts racing with different scenarios of how this can be used.

With that in mind, I had originally intended to take this post and talk about how this technology vision could be applied to business intelligence; more specifically, how Microsoft intends for it to apply to their product lines. But, after some additional thinking I decided that I wanted to take another direction. Instead, I’d talk at a more abstract level. I’d like to bring up some ideas on how to expand this vision and perhaps redefine it a bit. The idea behind this is to allow your curiosity and imagination to run wild as opposed to having me just dictating to you. To me this is a far more powerful prospect.

With this change in direction the amount of content has grown larger than what I’d consider appropriate for a single blog post. So, I am going to break this up into a small series. It is my hope that you will find this not only interesting but that it will prompt you to imagine and, perhaps, help create the future!

Here is a fairly common challenge: perform a complete refresh of the data in a target table. It seems pretty basic and easy. All of the existing data is replaced with the new load. There is no change detection required; just a simple load.

Yet, more often than not, when I go into a client that is having ETL issues I find that they have tackled this challenge with a poor design. A poor design that ultimately results in errors and ongoing headaches for their support staff. This is so common that I sometimes wonder if I have been following the same shortsighted developer from client to client.

This post is my attempt to jump ahead of this mystery developer and hopefully prevent some future headaches. Essentially, the poorly designed ETL process follows this pattern:

So, what is wrong with this? First, by dropping the existing table it effectively makes the table unavailable until the load completes. This can be particularly troublesome when the load process is lengthy. If the load is scheduled during off hours this might not be a problem. Yet, this downtime can be avoided.

Second, if the load fails you are left with an empty table. In my mind this is the most egregious issue with this pattern. An error leaves you with nothing. This can easy be the difference between an emergency and a simple annoyance. If the process had left the original data in place it would be old and stale but at least it is still accessible.

Here is a modified pattern that helps reduce the issues with the original:

This is a classic pattern for handling a complete refresh. The load itself is completed against a staging table. This leaves the existing target in place and doesn’t interrupt use while the load is running. The target table is only ever dropped after a quality check has been performed. This helps ensure that the load was successful and should prevent the empty table issue that the original process had.

There is nothing complicated about this pattern. It is still simple and straight forward. But, unlike the original, some forethought is put into dealing with failure. I hope this encourages you to take a look at how your ETL deals with problems. Does it handle it gracefully, or does it leave you with an emergency?

Microsoft recently released an updated version of the PowerShell Quick Reference Guide. Essentially, it is a printable booklet that has an overview of commonly used PowerShell commands. So, if you’re learning PowerShell or just need a helpful reminder, check it out.

By the way, I cannot stress enough how useful PowerShell can be when working in a Windows Server environment. It really is worth learning even if it isn’t something you plan on using on a daily basis. If you don’t know it, now is a good time to start!

While catching up on blog posts I came across a post by Patrick Husting from last month. The post titled “Better looking charts in Excel 2007/2010” provides a before and after example of a chart in Excel. The simple example is meant to show the importance of adding a little style to your reports.

I think this is an often overlooked aspect of a complete BI initiative. Too often the focus is only on the technical aspects and does not ultimately address user adoption. While there are several things that can contribute to low user adoption having ugly and unintuitive reports doesn’t help.

With this in mind, I would argue that taking the time and effort to design a comprehensive style guide is well worth the investment. If done properly it should help a project realize ROI. So, when you’re planning your next project consider budgeting time for style planning.

Anyhow, check out Patrick’s blog and let me know what you think.

Just a heads up for those who didn’t notice: yesterday Microsoft released the official release date for the next version of SQL Server. It will be generally available May 2010 and should appear on the May pricelist.

I have been looking forward to some of the improvements in the upcoming release and am particularly excited about Master Data Services. Here is a quick overview of the key features:

• Master Data Services
• PowerPivot
• Application & Multi-Server Management
• Improved Hyper-V Support – Live migration between hosts
• StreamInsight
• New Data Center Edition – Increased capacity (256 logical processors, max OS RAM, etc.)
• New Parellel Data Warehouse Edition

If you’d like to give the last CTP a try you can download it here.

I would have thought that this would be fairly common knowledge but I am often asked for a quick example of how to strip off the time element of a SQL datetime type. So, here is an example followed by an explanation of how and why it works:

SELECT DATEADD(d,DATEDIFF(d,0,GETDATE()),0)

Technically, this doesn’t strip the time but instead sets it to midnight of current date. So, what is this snippet doing? To start we need to know a little bit about how datetime is stored in SQL Server. The datetime type is an 8 byte type. Essentially, it is stored as two distinct 4 byte integers. The first 4 byte integer represents the number of days since the base date for SQL Server. The base date for the datetime type is 01/01/1900. The second 4 byte integer that represents the milliseconds since midnight.

With that in mind, the following snippet should return this base date:

SELECT CAST(0 as datetime)

Now, back to dissecting the original snippet, I am using the built-in DATEDIFF function to return an integer. The integer is the number of days that have elapsed since the base date of 01/01/1900. With the knowledge of how a datetime is stored you might be tempted to just CAST directly from the datetime to an int. However, depending on the time of day, it might cause the int to round up. So, it is better to just rely on the DATEDIFF.

Finally, to convert this integer back to a datetime type I am using the DATEADD function. Once again, we are using the base date and adding the integer to it to give us today’s date with a midnight time component. Pretty simple, right?

A new version of the core SyntaxHighlighter that wp-simplesyntaxhighlighter wraps has been released. As a result, I have released a new version of the WordPress plug-in that reflects this latest revision.

You can find a quick list of the changes here.

This is a continuation on my previous rants on borrowing traditional software development concepts for a BI project. As the title suggests, the topic today is source control.

No matter what you want to call it, a cornerstone of any good software project should be source control. This is pretty fundamental for software projects but it hasn’t been a priority for many in BI. I am not entirely sure why this is. Is it just an oversight? Are BI project managers oblivious to the existence of good source control tools? I’m not sure what the answer is but I am sure that BI teams should be using it.

The benefits of source control should be well known. So, I am not going to rehash them here. However, I would like to go over a few things to keep in mind:

Full Coverage

Source control can be applied to most parts of a BI project. Source control concepts can apply to data models, cube structures, ETL, and even the documentation for the project. Pretty much anything that changes over time can benefit. So, try to think beyond just source code.

Version Labels

Having the ability to identify something makes it infinitely easier to discus it. This same concept applies to components of a project. It doesn’t really matter whether it is some sort of name or just a revision number. The benefit is that it allows you to have a common term to refer to a snapshot of a project.

I strongly suggest that the labeling is universal and span all of the projects components. In other words, version 2.1 could universally refer to a snapshot of the ETL as it does to the data model. And, when you start introducing formal issue tracking and end-user team portals it becomes even more important (more on these topics at a later date).

Ultimately what I am getting here is that checking in revisions isn’t enough. Develop a consistent labeling system for releases. And, use the commenting features of your source control system to identify changes; all changes!

On Cost

There are many source control systems out there. And, many of these are free. So, cost should not be an excuse. Additionally, many of the design tools on the market today have some flavor integrated into the development environment. Don’t make excuses, find something that works for your team and use it!