Zero-Downtime Architecture Engineered for Five-Nines Uptime Reliability.สถาปัตยกรรม Zero-Downtime สำหรับ Uptime ระดับ Five-Nines

The most common cause of MVNO service outage is a control-plane link failure — the SS7 or Diameter connection to the host MNO drops and subscribers lose access to services that require real-time authorization: voice calls, data sessions, SMS.

Cipher Telecom eliminates this failure mode through multihoming: every SCTP association and Diameter peer connection is bound to multiple local IP addresses residing on different physical network interfaces, connected to different upstream switches. If a cable fails, a NIC fails, or an upstream switch fails, the SCTP stack detects the path failure in under one second and instantly shifts all traffic to the surviving paths — without tearing down the association and without interrupting any active user session or in-flight billing event.

How SCTP multihoming works

• Each SCTP endpoint advertises 2–4 local IP addresses at association setup.
• The remote peer (host MNO STP) monitors all paths via HEARTBEAT chunks.
• If the primary path becomes unreachable, SCTP promotes the next available path without dropping the association.
• No TCP connection teardown, no BGP reconvergence wait, no session state loss.
• Active user sessions continue without interruption — the failover is transparent to the application layer.

Every network element in Cipher Telecom's stack — OCS, PCEF, SMSC, USSD gateway, signaling gateways — is deployed across two redundant virtualized server environments: Node A and Node B, resident in the same data centre but on separate physical hosts, separate power feeds, and separate network uplinks.

The cluster operates in Active/Active mode: both nodes handle live traffic simultaneously. Subscriber session state is synchronously replicated between nodes so that if Node A fails, Node B has a complete and current view of every active session — no sessions are dropped, no billing records are lost.

Cluster topology per network element

Every Cipher Telecom network element runs as a Virtual Machine (VM) or container on a hypervisor-based infrastructure layer. This provides three operational advantages over bare-metal proprietary appliances:

• Elastic scaling — additional VM instances can be provisioned in minutes as subscriber load grows, without hardware procurement lead times.
• Snapshot-based backup — complete system state, including subscriber database, can be snapshotted and replicated off-site. Recovery from a catastrophic failure is a restore operation, not a rebuild.
• Resource isolation — each NE runs in its own isolated VM with dedicated vCPU and memory allocation; a resource spike in one element cannot starve another.
• On-premise or co-location flexibility — the same VM images run on customer-provided data-centre hardware or Cipher-managed co-lo infrastructure.