Determining the need for a load balancer: Concepts

Load balancing distributes work across multiple systems and across several services at once. For example, in an IMPAX cluster, the distribution of HTTP(S) requests to multiple Application Server, the distribution of study arrival, and the distribution of study retrieval is required. Load balancing each of the these services has different configuration parameters and issues. The choice of load balancer options can effect the stability of the load-balancing environment.

Why implement load balancing?

Load balancing should be considered when any of the following criteria are met:

The site in question has a high volume of data that must be stored/manipulated.
The workload will overload a single Application Server.
Disruption of services cannot be tolerated. Outages of the IMPAX services will adversely impact hospital operation.

When should load balancing be implemented?

Implementation of load balancing can depend on several factors. The following list is not exhaustive but does illustrate some of the major points that indicate that load balancing should be instituted at a site. Implement a load balancer when:

Traffic volume increases and data/workflow is impacted by network operation (constriction).
The Application Server is non-responsive and/or usage patterns indicate the server is approaching saturation of critical resources (disk, memory, CPU).
Analysis of site environment warrants additional resources.
A duplicated server configuration is necessary to meet uptime and failover requirements.

What should be assessed before implementing a load balancer?

A site survey should be undertaken to assess the needs of the specific installation. The following areas must be understood for each site:

Network topology and infrastructure
Number and type of clients to be used
Number and type of modalities
- Image mix and size
- Study frequency
Amount of online/nearline/offline storage
Profile of the expected growth