Plan CUBE high availability: box-to-box active/standby, cluster load-share, or DNS-based failover. Calculate expected failover time, call preservation, and carrier coordination needs.
CUBE has three HA models, each with different failover times, call preservation semantics, and cost. Enter your uptime target, call volume, and carrier topology — get the right model, the config skeleton, and the testing plan.
Set inputs, click Plan CUBE HA.
Box-to-box, cluster, or DNS.
Interchassis redundancy group.
Failover coordination.
Documented failover test.
Ops playbook at failure.
Box-to-box (Inter-Chassis Redundancy, ICR): two physical CUBE units share a virtual IP via RG (Redundancy Group). Signaling state synced. Active carries all calls; standby takes over on failure. Failover time: 3-15 seconds. Preserves new call invites, doesn't always preserve mid-call. Requires dedicated control-link (crossover or VLAN) between boxes. Cost: 2× hardware + 2× licenses (standby licensed at same CCS). Suited for: offices where all calls go through one SBC, ~100-500 CCS, single-site.
Cluster (Load-Share via DNS SRV or SIP proxy): 2-4 CUBE units all active, calls distributed. Each unit handles its share; if one dies, traffic redistributes to survivors. Failover time: 0-30s depending on OPTIONS keepalive. No shared state — each call is on one box, if that box dies the call drops (new calls route fine). Cost: N× hardware, N× licenses. Suited for: large deployments (500+ CCS), contact centers, active/active.
DNS-based (SRV multi-record): CUCM and carrier use DNS SRV to locate CUBE. Multiple SRV records point to cluster members with equal or weighted priority. Failover time: until TTL expires (typically 60s) + SIP retry (5-15s). Simple, no shared state, very reliable. Downside: active calls on the failed unit drop. Cost: still N× hardware but simpler config — no ICR setup. Suited for: cloud deployments, multi-site, moderate SLA targets.
Box-to-box: 3-15s. Cluster: depends on OPTIONS. DNS: 60s + retry.
Yes — box-to-box pair at each site, then multi-site behind DNS SRV.
Yes — Cisco licenses the feature set including the standby.
For box-to-box: they see one VIP. For DNS: they see multiple SRV records.
During a rolling upgrade, cluster model is best — active calls stay on un-upgraded nodes until drained.
200 CCS.
400 CCS.
800 CCS, zero loss.
Our UC Modernization Sprint brings up inter-chassis HA or cluster, documents failover, and runs the first real-world test.