Networking and Traffic Management (ADC)
This domain covers the configuration that determines whether a published service is reachable, sticky, and healthy: load-balancing methods, persistence, monitors, content switching, and global server load balancing. Questions often describe a symptom and ask which ADC object or setting explains it.
Load balancing and persistence
- Least Connection is the default method. Round Robin, Least Response Time, Least Bandwidth, Custom Load, and hash methods such as Source IP and URL are alternatives.
- Persistence types include SOURCEIP, COOKIEINSERT, SSL Session ID, and RULE. Persistence timeout is not the same as client idle timeout; an expired or stale cookie can pin a user to a failed service unless a backup vserver exists.
- Service groups simplify bulk membership changes; monitors can bind at group level and be overridden per member.
Monitors
- The monitor type must match the application. A TCP monitor only proves the port is open; use HTTP/HTTPS with a receive rule for real application health.
- Detection time is roughly interval multiplied by retries. Set down-time and raise thresholds deliberately to avoid flapping.
Content switching and GSLB
- A content-switching vserver evaluates policies top-down. Always define a default action, or unmatched traffic is dropped.
- GSLB methods include failover, round robin, static proximity, and dynamic RTT. Site persistence can use cookie domain or source IP.
Worked example
Users must stay on the same backend for a 30-minute Java session, the app sets JSESSIONID, and two sites are active-active with automatic failover. Bind an HTTP monitor with a receive rule matching the app's health string, enable COOKIEINSERT persistence on the load-balancing vserver, and place the vservers behind a GSLB parent using the failover method with site persistence. Reasoning: cookie persistence handles intra-site stickiness, while GSLB failover handles site loss; a TCP monitor would keep sending traffic to a broken app.
Traps that cost marks
- Using a TCP monitor and declaring the service healthy when the app returns 500 errors.
- Confusing persistence timeout with the HTTP/SSL session timeout.
- Omitting the default content-switching action.
- Forgetting that a GSLB service needs a correct site IP and metric exchange configuration.
- Mixing RNAT and SNIP assumptions when backends need to initiate connections.