Case study · P—11
OLAP Scaleout: Load-Balanced Multidimensional Cube Cluster
- Role
- Cluster design · Senior Business Intelligence Developer
- Period
- March 2013 – June 2013
- Organisation
- LC Waikiki
- Stack
- SSAS Multidimensional, F5 load balancer, SQL Server
- Status
- Completed
In short
F5 load-balanced OLAP cluster for multidimensional cubes with 24/7 high availability: sub-second average response for 4,800 users, zero downtime.
Context
LC Waikiki’s multidimensional OLAP cubes, first built in 2011, give business units centralized access to analytics across merchandising, supply chain, finance and HR. The SSAS estate grew to 15 TB.
Role: designed the load-balanced OLAP cluster, as Senior Business Intelligence Developer.
Problem
The cube estate had to stay available around the clock and answer quickly for 4,800 users.
Approach
- Scale-out cluster: an OLAP cluster for the multidimensional cubes, load-balanced with F5.
- High availability: 24/7 operation with Active-Active and Active-Passive set-ups.
Architecture and technology
An F5 load balancer in front of a cluster of SSAS multidimensional servers, serving a 15 TB SSAS estate.
Outcome
- Sub-second average query response for 4,800 users.
- Zero downtime.
- 24/7 high availability.
Lessons
- The measures were the ones users feel: sub-second average response and zero downtime for 4,800 users.
- High availability was part of the design from the start, with Active-Active and Active-Passive set-ups behind one load balancer.