Software-Defined Networking Layers presentation
- 1. SDN LAYERS S O F T WA R E - D E F I N E D N E T W O R K I N G L AY E R S BY : A B D U L L A H I B R A H I M A H M A D
- 2. OVERVIEW • Current State in Networking • What is SDN? • SDN Abstractions • SDN Architecture & Layers • Cross-Layer Issues
- 3. STATE OF QUO IN NETWORKING • Planes of functionality: – Management (Define the network policy) – Control (Enforce the policy) – Data (Execute the policy) • Control & data planes are tightly coupled – Difficult to add new functionality • Decentralized structure – Network resilience – Complex and Static Architecture
- 4. WHAT IS SDN? Network Architecture with four characteristics: 1. Control and data planes are decoupled 2. Forwarding decisions are flow based instead of destination based 3. Control logic is moved to SDN controller or Network Operating System 4. Network is programmable through software applications
- 5. SDN ABSTRACTIONS • What do we do when dealing with complex problems? – Decompose it to simpler problems – Define an abstraction for each component • SDN Abstractions: – Forwarding – Distribution – Specification
- 7. SDN LAYERS
- 8. NETWORK INFRASTRUCTURE • Switches, routers, … • No embedded control software • Include open and standard interfaces (e.g. OpenFlow, POF, …) • A data plane device is a hardware or software element specialized in packet forwarding based on a pipeline of flow tables
- 9. SDN DATA PLANE DEVICES
- 10. SOUTHBOUND INTERFACE • APIs connecting and separating control and forwarding elements • Openflow is the most widely accepted • Openflow provides three information sources for NOS: • Event-based messages when a port or link changes • Flow statistics • Packet-in messages when forwarding device doesn’t know what to do
- 11. NETWORK HYPERVISOR • Network-wide software layer • Under network control applications • On top of distributed networking devices • Multiplex, demuiltiplex and monitor • Implemented via distriputed system • Distribute networks states and loads • Logically centralized (huge difference) • Partition resources through multiple contexts • Distribute logical context over multiple physical devices
- 12. NETWORK OPERATING SYSTEM Operating System Model
- 13. NETWORK OPERATING SYSTEM SDN Model
- 14. NETWORK OPERATING SYSTEM Types of SDN Controllers(NOSs) • Existing controllers can be categorized based on many aspects • Centralized vs Distributed • Centralized • Single point of failure • Scaling limitations • Can be highly parallelized to overcome above limitations • Distributed • Scalable • Fault tolerant • May offer weak consistency
- 16. SDN CONTROLLER PARTS: CORE SERVICES • Topology • Statistics • Notifications and device management • Shortest path forwarding • Security mechanisms
- 17. SDN CONTROLLER PARTS: SOUTH AND NORTHBOUND • Southbound: – Common interface for upper layers while allowing different southbound APIs – Can be seen as device drivers • Northbound: – Ad hoc APIs – RESTful APIs – File systems
- 18. SDN CONTROLLER PARTS: WEST/EASTBOUND • Only in distributed controllers • Import/export data between controllers • Algorithms for data consistency models • Monitoring/notificatio n capabilities
- 19. ARCHITECTURE AND DESIGN OF SDN CONTROLLERS
- 20. NORTHBOUND INTERFACE • Mostly a software ecosystem • Can be compared to POSIX standard in operating systems • No de facto standard as of right now • Each controller defines its own northbound APIs • NOSIX is an attempt in this direction
- 21. LANGUAGE-BASED VIRTUALIZATION • Capability of expressing modularity • Allowing different levels of abstractions while still guaranteeing desired properties • Allow different views of a single physical infrastructure • One virtual “big switch” could represent a combination of several underlying forwarding devices • Simplifies the task of application developers • See the network as a simple “big switch” • Simplify the development and deployment of complex network applications
- 22. PROGRAMMING LANGUAGES • Current state in network programming languages: • Openflow: same as Assembly language • Mimic hardware • Too much low-level details • No modular code • No code reuse • Thus we are moving to higher level programming languages • FatTire (functional): uses reg exp to describe network paths • FML (dataflow, reactive): high level policy description language • Procera (functional, reactive): high level abstractions to describe reactive and temporal behaviors
- 23. NETWORK APPLICATIONS • “Network brains” • Implement control-logic which dictate the forwarding device behavior • Traffic engineering • Routing, load balancing, scheduling, … • Mobility and wireless • Interference management, wireless network modeling, … • Measurement and monitoring • Measuring link utilization, traffic monitoring, … • Security • Attack detection, access control, flow- rule enforcement • Data center networking • Optimizing network utilization, predict application workloads, …
- 24. CROSS-LAYER ISSUES • Debugging and troubleshooting – Runtime debugging • Ndb(same as gdb): breakpoints, watch, back-trace, … – Post-mortem analysis • Record and replay network events • Testing and verification – Verification • Connectivity, loop-freedom, access control – Testing • Generate streams of packets and test as many events as possible • Simulation and emulation – Mininet: prototype and evaluate SDN protocols and applications
- 25. REFERENCES • Kreutz, D., Ramos, F. M., Verissimo, P. E., Rothenberg, C. E., Azodolmolky, S., & Uhlig, S. (2015). Software-defined networking: A comprehensive survey. Proceedings of the IEEE, 103(1), 14-76.