Wednesday, June 27, 2012

MSTechEd Europe 2012 Day 1: Optimizing your BI Semantic Model for performance and scale

I joined my first session on tuesday by Dave Wickert. Here's my short version of the session.

VertiPaq is all about Performance, Performance, Performance!
The VertiPaq engine chooses query performance over processor performance.
What the VertiPaq engine does:
  1. Encoding (per column): creates a dictionary with value or hash encoding (self-deciding)
  2. VertiPaq storage
    • dictionary
    • column segment
    • hierarchy
    • relationships
    • partitions
  3. Processing
    • Process Full
    • Process Data
    • Process Recalc (includes calculated columns)
  4. Advanced Processing
    • Process Defrag: updates the dictionary
    • Parallel processing
    • Error handling
  5. Server Memory Map
    • Databases
    • Formula engine (DAX, MDX)
    • VertiPaq caches
    • Storage engine
Defaults that can be adjusted (be careful):
VertipaqMemoryLimit - 60%
TotalMemoryLimit - 80%
VertipaqPagingMode