Thoughts on SQL Server Columnstore

What are your thoughts on using CStore under SSAS MD ROLAP for large fact tables/dimensions?

Ever do any benchmarks on a DW using CStore (relational tables) vs an SSAS MD Molap cube?

What are your thoughts on CStore vs Tabular/PowerBi/AzureAS xVelocity engine?
currently CStore has no number of rows, CPU or memory limitations like xVelocty engine does

Since CStore rowgroups can get fragmented due to deletes/updates
what is your recommended maintenance plan? Rebuild or Reload?

How do CStore indexes handle frequent updates compared to Rowstore and Statistics?
Cstore doesnt use statistics but meta data and delta stores

Do you use Cstore on In-Memory Hekaton databases or whats the point? (Diminishing returns, non-compatible tech)

What are your thoughts on CStore in PDW/APS vs Hadoop?

CStore issues: performance loss due to decompression
AKA, the time to first row slower than rowstore on large queries

Done any Benchmarks on Rowstore Batch Mode vs Columnstore Batch Mode?

Ever Used NCCS on OLTP tables with <1M rows?

Segment elimination: does it work for columns not next to each other?
remember all this data is in one physical MDF file
only when you do NCCs so that each column is addressed separately?

Benchmark:
Columnstore on OLTP/Relational tables
SSAS MD Molap
SSAS Tabular on-prem
Azure Analysis Services
Power BI Premium
Azure Synapse
Azure Data Warehouse
Hadoop/Databricks/Spark

Problem: coming up with a Query and dataset to compare all these

Leave a Reply