(Things I have learned are below)

Private Impressions:

I arrived ​on sunday evening in Seattle and used monday for some tourist impressions:

• Mono-Rail (be sure to have good knowledge about the "complex" network and connections :-) )
  
• Space Needle (where I bought some "space noodles")
  
• You should not wait for the traffic lights for pedestrians to change to greeen (because they change between red and white)
  

• Pike Market (where I bought a raspberry-hat for my daughter)
• at the hotel entry "umbrella bags" are available - people in Seattle are always prepared for rain

Things I have learned:

• SQL Server 2012
  • The next version of SQL Server (codename "Denali") will be named "SQL Server 2012" and will be shipped in the first half of 2012.
  • Project "Crescent" will be shipped as "PowerView".
  • Project "Juneau" will be shipped as "Data Tools"
  • There will is an interesting project Data Explorer" (mayby shipped as "PowerInsert") which manages mash-ups of different datasources through Azure.
• Avoid bad execution plans for queries with where-clauses like "@parm1 is null or @parm1=col1" through use of option recompile.
• Microsoft is aware of that there is the need of bringing the whole BI stack into the cloud.
• Microsoft has innovative ideas of make a better BI experience to the enduser.
• PowerPivot (V2)
  • Despite that PowerPivot can not handle many-to-many-relationships, DAX will help out. PowerPivot V2 will make those formulas easier to read (and maybe faster).
  • PowerPivot contains new import filter capabilities.
  • Tables in PowerPivot Field List will be in alphabetical order by V2 (inspite of chronological order by creation time).
  • Possibility of sorting a column by another column (helpfull for eg. sorting monthname by monthnumer)
  • Formatting options for measures
  • Descriptions are available for tables, columns, kpi, ...
  • Use "Go to" to move from table in Diagram View to same table in Data View.
  • Possibility to maximize table in Diagram View.
  • Perspectives will be available.
• SQL Azure (by end of 2011)
  • max database size will be 150gb
  • management will be done via a metro interface
  • federation will be available
• BISM
  • meant for enterprise BI (e.g. more than 2 GB PowerPivot files, security, partitions, ...)
  • can use PowerShell
  • row security (constants) vs. dynamic security (LOOKUPVALUE())
  • to work in BIDS a temporary "workspace database" on SSAS is automatically created and managed (named with guid postfix)
  • partitions are available
  • change process-option from "default" to "do not process" in SSMS for big data
  • "process recalc" on database level is necessary for the most (does process connected tables)
  • "PushedDataSource" contains copy&pasted data
  • process add shows number of rows in partition (not the amount of added columns)
  • BIDS add-ins "DAX editor for SQL Server" and ​"Create DAX file for model" are available from codeplex
  • "process defrag" should be done if "process incremental" was used heavily
  • works good with servers with high amount of memory and very fast processors (inspite of good disk i/o and high amount of processors)
• SQL Server Parallel Datawarehouse can compete with NoSQL
• DAX (V2)
  • RANKX(<table>, <expression>): e.g. if(ISNOTBLANK([Sum of Sales Amount], RANKX(ALL(Regions), [Sum of Sales Amount]), BLANK())
  • Drill down to Fact-Table is possible through double click in (Power)Pivot-Table
  • SUMX(TOPN(10,Product,[Sum of Sales Amount]), [Sum of Sales Amount)]
  • Tables can be marked as DateTable (and will make the working with TimeIntellignece much mor easier)
  • New nested IF-Statement: SWITCH()
  • Split complex measures up and hide those splitted measures for client tools
  • USERELATIONSHIP() to use inactive relationships of the model
  • bring measures to rows
  • PATH(child, parent)
  • LOOKUPVALUE(column, key, PATHITEM(path, level))
  • M2M / cross filter problem --> two white paper will be published (one by Kaspar DeJonge and another by Russo/Ferarri)
  • EVALUATE returns always a table/rowset
  • DEFINE MEASURE Table[measure]=<expression>
  • ORDER BY
  • CROSSJOIN()
  • ADDCOLUMN()
  • SUMMARIZE()
  • DAX QueryPlan will be available (as Profiler event); logical plan=what ist done by query processor, physical plan=what ist done by vertipaq
  • <task>_vertipaq means that query is directly process by vertipaq (which is a good sign)
  • materializing columns vs. using complex dax expressions: it depends on which is better
• ColumnStore Index
  • sweet spot: good design (e.g. matching data type, no functions), star joins, inner joins, aggregations
  • needs enough memory to build
  • good for scans, not good for seeks
  • consider table partitioning to facilitate updates
  • create from clustered index can be of advantage
  • use MAXDOP>1
  • helps e.g. BISM-DirectQuery
  • only columns are used are read and cached
  • LOB's are cached in consecutive storage
  • contains min&max values
  • not batch mode available for outer join and union
  • covering index vs. columnStore index