Common Causes of Uneven Load Balancing on NetScaler

Common Causes of Uneven Load Balancing on NetScaler

book

Article ID: CTX120542

calendar_today

Updated On:

Description

Verify the following items when diagnosing and explaining uneven load balancing on NetScaler platform:

  • Slow Start – The system performs slow start to avoid overloading the physical servers. During the slow start phase, the system distributes requests by round robin, regardless of the actual load balancing method configured on the virtual server. It honors the configured weight on bindings even during round robin.

    A slow start occurs when one of the following conditions exist:

    • The load balancing method changes.
    • A new server is added to the virtual server.
    • An existing server binding is removed from the virtual server.
    • A server changes its status from DOWN to UP.
  • Persistence – When enabled, persistence (by definition) creates uneven loading because the NetScaler appliance must direct requests from a specific client to a specific server. Only unique or expired clients get load balanced according to the load balancing method.

  • Inconsistent Server Performance – Performance across a set of servers is seldom consistent. Therefore, some servers which serve requests faster and based on the configured load balancing method might cause an uneven balance. The degree of imbalance increases as the degree of inconsistency between the servers increases.

It is common to see uneven load balancing because of slow start when there is a server that is fluctuating status between UP and DOWN on a frequent basis.

Additional Resources

CTX108886 – Citrix NetScaler Load Balancing - Slow Start Mode

Issue/Introduction

This article provides a list of some of the more common causes of uneven load balancing when using the NetScaler platform.