All Barracuda Load Balancers, all versions of firmware.
Scheduling methods can be configured on the Basic > Services page by clicking the blue Edit link next to the VIP of the service you wish to configure. This controls the algorithm that the Barracuda Load Balancer uses to determine where to send the next incoming request.
The Barracuda Load Balancer currently supports multiple scheduling algorithms including Weighted Least Connection (WLC) and Weighted Round Robin (WRR). WLC detects which server has the least open connections and sends new traffic there, while WRR distributes traffic to available servers in a sequential order with no consideration given to the number of existing connections to each server. They can be weighted toward one server or another by either user-specified weights or Adaptive Scheduling.
Users may apply weights to servers that determine how many connections each server receives. This is useful in situations where Real Servers use different hardware or operate at different levels of efficiency.
Adaptive Scheduling allows the Service Monitor to dynamically modify the weights of real servers based on a resource utilization of a user's choosing, including CPU and Windows Terminal sessions. The Barracuda Load Balancer can even read a load value from a URL on each real server, giving a user complete control over how the Barracuda interprets a server's load.
Firmware version 3.0 introduces default scheduling policy changes for Layer 4 and Layer 7 services.
- Layer 4 - Weighted Least Connections and Weighted Round Robin still apply.
- Layer 7 can now be applied two ways, Layer 7 HTTP and Layer 7 RDP (enabled when service is setup on port 3389).
- Layer 7 HTTP - Includes Weighted Round Robin, now includes Weighted Least Requests. In 3.0, Layer 7 HTTP connections are recorded by actual HTTP requests and not connections, since a single client connection can have multiple requests (i.e. each image on a page can be multiple requests).
- Layer 7 RDP - Same policy as standard Layer 4, Weighted Least Connections or Weighted Least Requests.
Link to This Page: