Virtual Servers, Pools, Nodes, and Load Balancing Methods
Domain 2 - Virtual Servers, Pools, Nodes, and Load Balancing Methods
This domain tests object design and the load-balancing decision: which eligible member receives a new connection.
Object rules
A virtual server defines the listener and references a pool. A pool contains pool members; each member is a node IP plus a port. A node can back multiple members on different ports. Health monitors set member state; only up members are eligible unless the pool uses priority groups. Virtual server types include Standard, Performance (L4), Forwarding (IP), Stateless, and Reject. Standard is the usual L7 proxy; Performance L4 reduces inspection; Forwarding IP sends traffic based on routing; Reject resets traffic.
Methods
Round Robin rotates. Least Connections picks the fewest active connections. Weighted Least Connections uses member weights. Observed uses a performance score; Predictive uses a trend; Ratio uses configured ratios; Fastest uses response time; Least Sessions uses persistence entries; Dynamic Ratio uses a monitor-supplied ratio. Priority Group Activation makes lower-priority members standby until the active group falls below a threshold.
Worked example
A pool has A weight 3, B weight 1, and C weight 1, method Weighted Least Connections, all up. Current counts are A=30, B=10, C=10. Divide each count by its weight: A 30/3=10, B 10/1=10, C 10/1=10. The effective load is tied, so tie-breaking sends to one eligible member, then counts change. If A were 33, A's score would be 11 while B and C remain 10, so B or C is preferred. If A and B are priority group 1 and C is group 2 with an activation threshold of 1, C stays standby while at least one group-1 member is up.
Traps
- Using a ratio method without setting ratios; equal defaults are not a weighting scheme.
- Misreading priority group activation: the threshold is the minimum active members before standby members are used.
- Treating a disabled member as merely down; disabled and forced offline are administrative states that remove eligibility.
- Assuming persistence is a load-balancing method; it overrides selection for matching flows.
- Confusing node monitors with member monitors, which can change availability scope.
- Forgetting that virtual server type changes how much traffic LTM inspects.