Cerberos, in Greek and Roman mythology, is a multi headed hound (usually 3 heads) that guards the gates of the Underworld, to prevent those who have crossed the river Styx from ever escaping. Kerberos, required for a Performance Point Extranet topology (https://popbi.wordpress.com/2012/04/01/performance-point-2010-external-access-must-read/) also refers to 3 heads, the client, the server and trusted third party (being the active directory domain controller).
Whats this guy on about ? Ok, Analysis Services is, for now, a 3 headed beast too, the three heads depicted as different types of instances, with an organisation choosing to have three, just two or only one. These 3 heads are Multi-dimensional/Data mining, Tabular and PowerPivot which are three physically separate installations of Analysis Services. You can tell the difference between one or the other by observing the difference in icon style when all 3 are registered in Management Studio, or the Deploymentmode property in the msmdsrv.ini file – as outlined here Differentiating SSAS Instance Types.
Traditional SSAS Multi-Dimensional databases are simply too hard to construct and support for some organisations, so these new approaches offer very good news. PowerPivot /Tabular is a more user friendly option for developing analytical cubes that won’t make you feel like throwing up in the wastebin, unlike MDX for some people.
Each instance type of Analysis Services supports certain data source types, can be programmed with certain design tools, and each has its own model features which are all outlined here in this useful reference Differentiating Features of SSAS Instance Types. PowerPivot and Tabular are fundamentally similar underpinned predominantly by the design tools and subtle model differences. However you can import a Powerpivot database into BISM and publish back out as a Tabular model.
There are some gotchas before you can use the new tabular approach…. As long as your requirements can be modelled without writeback, many to many relationships, very complex calculations or very large datasets then you can use this fantastic new method of developing analytical cubes. You won’t need to know MDX (but instead will need to learn the Excel PowerPivot DAX feature set), and can revel in the new features such as connecting to a range of External Data Sources and blistering performance courtesy of the Vertipaq engine.
There is an excellent whitepaper on DAX for Tabular models here DAX in the Tabular Model Whitepaper and Samples but if you are simply looking for some visual capability demo’s check out the various demo’s like Picnic and Contoso here Power View Demo’s (you will need to have your Window’s Live ID and have Silverlight installed).
Finally, some additional design perspectives are offered by Paul Te Braak here Paul’s Stuff.