Distributed Computing Network Laboratory Reliability; Report on Models and Features for E2E Reliability ETSI GS REL 003 양현식

Distributed Computing Network Laboratory Reliability / availability methods

Distributed Computing Network Laboratory Reliability / availability methods Overview  NFV, architecture models  Network services elements Network function  Introduction Fault management cycle Protection schemes(2N, N+M, N-way)  NVFI and NFV-MANO support for VNF reliability and availability Fault management cycle phase review Non-redundant / on-demand redundant VNFC configuration Active standby Active Active

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Non-redundant / On-demand redundant VNFC configuration-(stateless/stateful) Placement – free anti-affinity / there is no concept of VNFCI protection in this mode. State protection: the VNFC state protection is not applicable. Fault detection: NFVI / Fault localization: NFVI and NFV-MANO Fault containment: containment may include powering off the failed nodes and/or network reconfiguration actions. (by NFVI and MANO) Fault remediation: NFV-MANO performs VM re-instantiation on failure. Supplementary actions may be required, e.g., network and/or storage association reconfiguration. (Fault recovery)

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Non-redundant / On-demand redundant VNFC configuration-(stateful-external) Placement – free anti-affinity / there is no concept of VNFCI protection in this mode. State protection: VNFC state protection is the responsibility of the VNFCI. State protection is done by an externalised entity, which may be (or utilises) storage service provided by the NFVI, or another VNFC that is provided by application (or combination thereof). Fault containment: containment may include powering off the failed nodes and/or network reconfiguration actions. (by NFVI and MANO) Fault detection: NFVI / Fault localization: NFVI and NFV-MANO Fault remediation: NFV-MANO performs VM re-instantiation on failure. Supplementary actions may be required, e.g., network and/or storage association reconfiguration. (Fault recovery)

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Active–Standby VNFC redundancy configurations- stateless Placement – VNFCs of the redundant pair need to be placed on different hardware servers with no or limited common failure modes. State protection: VNFC state protection is not applicable. Fault detection: NFVI / Fault localization: NFVI and NFV-MANO Fault containment: containment may include powering off the failed nodes and/or network reconfiguration actions. (by NFVI and MANO)

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Active–Standby VNFC redundancy configurations-stateless Fault remediation: NFVI performs VM failover on the hypervisor layer. Supplementary actions may be the responsibility of NFV-MANO (e.g., network reconfiguration). Fault recovery: NFV-MANO assigns the replacement of the failed node from the cloud resource pool as a new standby entity. NFV-MANO is then responsible for the on-demand diagnosis of the candidate failed entities, and initiation of any subsequent physical recovery request actions for entities with confirmed persistent faults.

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Active–Standby VNFC redundancy configurations-stateful-external Placement: VNFCs of the redundant pair need to be placed on different hardware servers with no or limited common failure modes. Depending on the nature of the externalized state repository, its placement may be subject to explicit or implicit anti-affinity requirements with respect to the VNFCI placement. State protection: VNFCI performs partial VM state replication for its critical state to external state replica repository. This state replication may be VNFC vendor proprietary or utilize 3 rd party or open source middleware services. Fault detection: NFVI / Fault localization: NFVI and NFV-MANO

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Active–Standby VNFC redundancy configurations-stateful-external Fault containment: containment may include powering off the failed nodes and/or network reconfiguration actions. (by NFVI and MANO) Fault remediation: VNFCI performs VM failover to standby. Stateful fault remediation requires that the standby node is brought to the state that is consistent with the state of the external state repository. (Reactive / proactive) / NFV-MANO (e.g., network reconfiguration). Fault recovery: NFV-MANO assigns the replacement of the failed node from the cloud resource pool as a new standby entity. NFV-MANO is then responsible for the on-demand diagnosis of the candidate failed entities, and initiation of any subsequent physical recovery request actions for entities with confirmed persistent faults.

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Active–Standby VNFC redundancy configurations-stateful-external Placement: VNFCs of the redundant pair need to be placed on different hardware servers with no or limited common failure modes. Depending on the nature of the externalized state repository, its placement may be subject to explicit or implicit anti-affinity requirements with respect to the VNFCI placement. State protection: VNFCI performs partial VM state replication for its critical state to external state replica repository. This state replication may be VNFC vendor proprietary or utilize 3 rd party or open source middleware services. Fault detection: NFVI / Fault localization: NFVI and NFV-MANO

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Active–Standby VNFC redundancy configurations-stateful-partial Fault containment: NFVI and NFV-MANO perform required containment actions. Depending on the specific failure mode and its associated scope, containment may include powering off failed nodes and/or network reconfiguration actions. Fault remediation: VNFM or VNF performs VM failover initiated by application. Remediation actions can also be split between the NFVI and VNF, e.g., NFVI may be fully responsible for the network fault remediation, while VNF may use network APIs at the NFV-MANO layer to request network reconfiguration as part of specific VNFC failure remediation cases. VNF/VNFM is responsible for starting the application in the state reflecting the replicated protected state.

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Active–Standby VNFC redundancy configurations-stateful-partial Fault recovery: NFV-MANO assigns the replacement of the failed node from the cloud resource pool as a new standby entity. VNFC is responsible for the state replication to bring the new standby up to date with the active state, which restores the redundancy configuration. NFV- MANO is then responsible for the on-demand diagnosis of the candidate failed entities, and initiation of any subsequent physical recovery request actions for entities with confirmed persistent faults.

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Active–Standby VNFC redundancy configurations-stateful-full Placement: VNFCs of the redundant pair need to be placed on different hardware servers with no or limited common failure modes. State protection: NFVI (specifically hypervisor) performs full VM state replication, including full VM execution state replication as a platform service. Fault detection: NFVI / Fault localization: NFVI and NFV-MANO

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Active–Standby VNFC redundancy configurations-stateful-full Fault containment: NFVI and NFV-MANO perform required containment actions. Depending on the specific failure mode and its associated scope, containment may include powering off failed nodes and/or network reconfiguration actions. Fault remediation: NFVI performs VM failover by hypervisor layer. Supplementary actions may be the responsibility of NFV-MANO (e.g., network reconfiguration).

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability Active–Standby VNFC redundancy configurations-stateful-full Fault recovery: NFV-MANO assigns the replacement of the failed node from the cloud resource pool as a new standby entity. The NFVI layer is responsible for the state replication to bring the new standby up to date with the active state, which restores the redundancy configuration. NFV-MANO is then responsible for the on-demand diagnosis of the candidate failed entities, and initiation of any subsequent physical recovery request actions for entities with confirmed persistent faults.

Distributed Computing Network Laboratory NFVI and NFV-MANO support for VNF reliability and availability MethodPlacementState protection Fault detection Fault localization Fault containment Fault remediationFault recovery Non- redundant StatelessFreeNot applicableNFVI detect NFVI & MANO NFVI & MANO VM re- instantiation (MANO) nothing Stateful (ext) Freethe responsibility of the VNFCI. NFVI detect NFVI & MANO NFVI & MANO VM re- instantiation (MANO) nothing Active / standby StatelessDifferent node Not applicableNFVI detect NFVI & MANO NFVI & MANO Hypervisor layer(NFVI) Cloud resource pool Stateful (ext) Different node External state replica repository NFVI detect NFVI & MANO NFVI & MANO VM failover (VNFCI) Cloud resource pool Active / standby Partial CP (stateful) Different node VNFC vendorNFVI detect NFVI & MANO NFVI & MANO VM failover (VNFM/VNF) by application Cloud resource pool Full VMCP (stateful) Different node NFVI (hypervisor) NFVI detect NFVI & MANO NFVI & MANO Hypervisor layer(NFVI) Cloud resource pool