Load balancing and traffic management
What this domain tests
Load balancing and traffic management is the largest operational area: virtual servers, services and service groups, monitors, LB algorithms, persistence, and content switching. The exam expects you to choose the right object and predict which backend receives a request.
Core rules
- A load balancing virtual server (LB vserver) owns the VIP and listens on a protocol/port. It is bound to services or a service group; members carry the backend IP and port.
- Monitors determine service state. A monitor can be bound at the service level or at the service group level. Default monitors are protocol-based; HTTP-ECV checks a response body.
- LB methods include round robin, least connection, least response time, least bandwidth, source IP hash, and custom. Least connection is dynamic and ignores weight unless configured.
- Persistence keeps a client on the same backend: source IP, cookie insert, SSL session ID, or token. Persistence timeout and backup persistence matter.
- Content switching uses a content-switching vserver with policies that evaluate HTTP host, path, or other expressions and forward to target LB vservers.
- Backup vservers and overflow policies handle primary capacity or health failures.
Worked example
Two HTTP services sit behind one LB vserver using least connection. Service A has 10 active connections and service B has 2. A new request arrives. Least connection selects B because it has fewer active connections. If B's monitor marks it DOWN, the ADC removes it from selection and sends new requests to A; existing connections may persist depending on persistence and surge protection. If a cookie persistence rule is bound, a returning client with a valid cookie goes to its original service even if another has fewer connections.
Traps that cost marks
A monitor bound to a service group member overrides the group monitor; check both. An LB vserver can be UP while all services are DOWN-users still fail. Round robin does not consider load. Content-switching policies are evaluated in priority order, and a policy with no target or a missing binding silently sends traffic to the default. Do not assume persistence survives a service being disabled unless the persistence entry remains valid.