|
|
|
| [Participants](participants) | [Scope](themes) | [Publications](publications) | [Contact](contact) |
|
|
|
|
|----|----|----|----|
|
|
|
|
|
|
|
|
# SDN and NFV composition, resilience, and recovery (SDN/NFV)
|
|
|
|
|
|
|
|
When considering the use of network equipment (e.g. routers, servers and links) through network virtualization technologies, there is a risk of failure involved (either hardware or software). Obviously risks are inherent to physical infrastructure because its hardware is prone to failure so does the respective software infrastructure. Currently, network infrastructures tend to have more software components with the use of SDN and NFV. One big obstacles in this trend is the system availability. In general, software are much more error-prone than hardware components. Hence, it is necessary to take into account attributes of dependability in order to mitigate faults. Ignoring this matter may cause a high rate of failures in virtual requests, causing a low quality of service and even more causing poor Quality of Experience to the end users..
|
|
|
|
|
|
|
|
In a previous work UFPR has proposed a NFV implementation of the classical distributed systems abstraction known as unreliable failure detectors. A failure detector is a distributed oracle that provides information about the state of the processes of a distributed system. In general, the failure detector outputs the list of processes that are suspected to have crashed. Our implementation is called NFV-FD (Network Function Virtualization- based Failure Detector) and relies on information obtained from a underlying SDN based on OpenFlow 1. NFV-FD uses information obtained from an OpenFlow controller to monitor processes and determine their state. In addition, NFV-FD also provides information about the state of communication links and reachability of processes. NFV-FD ensures a consistent view among all available processes of which processes are suspected to have failed. A distributed application can access this view in order to take decisions. In other words, NFV-FD allows the implementation of fault-tolerant distributed algorithms. An example of this kind of application is reliable broadcast. UFPR has implemented this algorithm on top of NFV-FD so did UFPE in another project i.e. the reliable broadcast is performed based on information obtained through NFV-FD.
|
|
|
|
|
|
|
|
In the VNE project, the aim is to bring the expertise of different partners to addresses challenges related to service composition, resilience and dependability issues using SDN and NFV in the project context. Some heuristics like Greedy Randomized Adaptive Search Procedures (GRASP), Simulated Annealing (SA), Genetic algorithms and Local Search will be analyzed in different scenarios. When possible, techniques based on exact methods (e.g. Linear Programming) can be used as well. The aim is also to propose innovative techniques for faults recovery in a time frame that is inline with the services constraints.
|
|
|
|
|
|
|
|
|