QlikView – Hardware Considerations

This blog focuses on some general guidelines and their technical details for QV hardware selection.

  • Intel has faster RAM access than AMD so Intel is recommended.
  • Xeon E5 & Xeon E7 are in the list of recommended Servers (as of 2015).
  • NUMA and Hyper Threading (HT)should be disabled (this was the general recommendation before Xeon processors got introduced I hope) but now enabling HT is recommended for servers like E5 V2 with two sockets.

Understanding little bit more on NUMA and HT will help in seeing their implications.

NUMA – Non-Uniform Memory Access:

Each cores in the server has allocated memory. If any of the core runs out of memory  then it has to request other core for borrowing. This communication might affect the performance of QV Servers.

HT – Hyper Threading:

HT was the technique used before inventing multi-core processors. HT basically disguises as multiple cores but in reality it is only one core. Intel found this technology to effectively use processor when the current task is stalling for something.

Even after inventing Multi-core processors since HT has benefits on many situations, it is managing to survive.

Key critical factors while choosing hardware for QlikView:

  • Clock Frequency and
  • Processing Capacity

QIX (earlier it was called as QlikView engine; after QV12 and Qlik Sense updated engine is called as QIX) engine performs better with Processor having higher clock Frequency. AMD has less clock frequency than Intel.

Processing Capacity would benefit if we have bunch of tasks to process in parallel.

“Xeon E5 V2 Two Socket” has better performance:

Chipset – 2 sockets mounted with 2 processors have direct connection via QPI link so communication happens really quick, this is the reason for better performance.

Note: More sockets (in Chipset) without direct connection result in poor performance.

Bottom line: Hardware selection is subjective to the environment and other technical  requirements so it’s better to take expert advice on top any general guidelines.

 

MDM Fundas

Master Data Management (MDM) is all about maintaining “single source of truth”.

As data gets generated from different systems, data quality gets affected.

Data Growth α Data Quality Issue

Example: Customer data

MDM Categories:

  1. MDM for operational purposes
  2. MDM for analytical puposes

Difference between DW and MDM.

Factors DW MDM
Purpose Aggregating & Reporting Maintain single source of truth for an entity/subject/dimension
Operates on Both Transactional & Non-Transactional data Only Non-Transactional data
Reports Reports for analyzing data Reports on Data Governance, Compliance and Data Quality
Write back on source Data Warehousing has nothing to write on source system Often MDM has something to correct in source system