Availability Management with Stateful Fault Tolerance
Ali Kafel, STRATUS Ali.Kafel@stratus.com
NFV World Congress May 6-8 in San Jose, CA
This PoC will demonstrate the concept of Service Accessibility and Service Continuity using Software Defined Availability(SDA), a dynamic way of providing fully automated selectable levels of resiliency, including stateful Fault Tolerance (FT), without the need for any application awareness or code changes. With this approach, multiple VNFs from multiple vendors can be deployed in an OpenStack NFVI environment with an Automated Virtualized Resilience Layer. This means different levels of availability can be enabled by the software infrastructure, based on what is already built in the application and what is needed.
The SDA approach enables operators to deploy any VNF in one of three availability modes: Fault Tolerant (FT), High Availability (HA) and General Availability (GA).
- With FT mode, each VNF is deployed with redundant VM instances and full-VM state checking (statepointing) is automatically performed by the NFVI, insuring full state protection, without application awareness.
- In the HA mode, no VM instance redundancy or VM state checkpointing is performed. However, fault detection and automatic recovery with re-instantiation is performed by the NFVI
- In GA mode, no VM instance redundancy, no VM state checkpointing and no automatic recovery are performed. However, other Fault Management functions such as detection and cleanup are performed by the NFVI
The VNFs that will be demonstrated as part of this multi-vendor PoC will be a Brocade Vyatta virtual router / firewall (vFW) configured as stateful inspection firewall VNF and Allot Communications’ virtual service protection system (DDoS) composed of two VNFs (virtual Service Protection Sensor vSPS and virtual Service Protection Controller vSPC). An Aeroflex TeraVM virtual test infrastructure is used as a primary test tool for the datapath functionality provided by the vFW and service protector functions.
These VNFs from multiple vendors will be instantiated onto a simple static NFV Forwarding Graph (emulating the configuration that is analogous to instantiation of corresponding physical NFs onto functionally equivalent service chain).
The PoC will demonstrate that fault tolerant (FT) VNFs will always have redundant VM states, and therefore a failure to one of the nodes will not have negative impact on the QoS or SLA of the service function. HA VNFs will show that when the primary node fails, the service will be interrupted only for a short period of time and will restart automatically, insuring service accessibility.
Final PoC Report (31/07/2015) NFVTST(15)000119_POC__35_Final_Report.docx