SDN and NFV - SKKUmonet.skku.edu/wp-content/uploads/2018/08/W15.SDN_NFV.pdf · 2018-12-12 ·...
Transcript of SDN and NFV - SKKUmonet.skku.edu/wp-content/uploads/2018/08/W15.SDN_NFV.pdf · 2018-12-12 ·...
Networking Laboratory 1/78
Sungkyunkwan University
Copyright 2000-2018 Networking Laboratory
SDN and NFV
Prepared by P. Thorat, R. Challa, S. M. Raza and H. Choo
Networking Laboratory 2/78
Presentation Outline
Software Defined Networking (SDN)
Case Studies of SDN
Role of SDN in Different Networks
Knowledge Defined Networking
Network Function Virtualization (NFV)
NFV Architecture Framework
RouteFlow: NFV Case Study
SDN and NFV Role and Challenges
SDN Current Activities and Standardization
Networking Laboratory 3/78
SDN Introduction Video
Content
► Definition of a Software Defined Network (SDN)
► The control and data components in a SDN and how they work with one
another
► Connections establishment in a scalable enterprise architecture
► The features and benefits of SDN
Duration
► 5 minutes and 21 seconds
VIDEO
Networking Laboratory 4/78
SDN Concept-Past, Present and Future
(1/2)
Past► Combined control and data plan
► Limited performance and capability
► Restricted pace of innovation
Present► Decoupled control and data plans
► High throughput
► More Scalable
► Pace of innovation limited to pace of network
element innovation
► Restricted by hardware and silicon innovation
Future► Separate network and control element
► Powerful computers accelerate control plan
innovation
► Very simple data plan
Networking Laboratory 5/78
SDN Concept-Past, Present and Future
(2/2)
SDN concept
► Separates the data plane from the control plane
► Network Operating System in the controller layer controls the data plane
► Different applications run on top of the Network Operating System
Networking Laboratory 6/78
SDN Architecture
Networking Laboratory 7/78
Case Study: G-Scale (1/3)
General Description
Google’s global user based services (Google Web Search,
Google+, Gmail, YouTube, Google Maps, etc.) require significant
amount of data to be moved from one region to another, making
these applications and services very WAN-intensive
Google maintains two separate networks because they have
different requirements
Google’s WAN is organized as two backbones
► Internet facing (I-scale) network that carries user traffic
► Internal (G-scale) network that carries traffic between datacenters
Networking Laboratory 8/78
Case Study: G-Scale (2/3)
Operational Mode
Thousands of individual applications runing across B4 target
network, Google categorizes them into three basic classes
► User data copies (e.g., email, documents, audio/video files) to remote
datacenters for availability/durability
► Remote storage access for computation over inherently distributed data
sources
► Large-scale data push synchronizing state across multiple datacenters
Networking Laboratory 9/78
Case Study: G-Scale (3/3)
Deployed Equipment
Built from the merchant silicon
► 100s of port with non blockling 10 GE
support
► Custom hardware running Linux
Openflow support
Open source routing stacks
► Quagga BGP stack
► ISIS/BGP for internal connectivity
Features
► Fault tolerant
► Scale to multiple Tbps
► Support programming for traffic
engineering
Networking Laboratory 10/78
SDN Controllers (1/4)
OpenDayLight (ODL)
OpenDaylight is an open platform by Linux foundation for
network programmability to enable SDN and create a solid
foundation for NFV for networks at any size and scale
OpenDaylight software is a combination of components including
a fully pluggable controller, interfaces, protocol plugins and
applications
With this common platform both customers and vendors can
innovate and collaborate in order to commercialize SDN- and
NFV-based solutions
Networking Laboratory 11/78
SDN Controllers (2/4)
ONOS
The Open Network Operating System (ONOS) is the first open
source SDN network operating system targeted specifically at the
Service Provider and mission critical networks
The major goals of ONOS are:
► Carrier grade features (scalability, availability, and performance) in the
control plane
► Web style agility
► Migration of existing networks to white boxes
► Innovation in both network hardware and software, independent of their own
time scales
Networking Laboratory 12/78
SDN Controllers (3/4)
RYU
Ryu means "flow" in Japanese
Ryu is a component-based software defined networking
framework
Ryu supports various protocols for managing network devices,
such as OpenFlow, Netconf, OF-config, etc.
Ryu fully supports OpenFlow 1.0, 1.2, 1.3, 1.4, 1.5 and Nicira
Extensions
The Ryu Controller is supported by NTT and is deployed in NTT
cloud data centers as well
Ryu uses Python for development
Networking Laboratory 13/78
SDN Controllers (4/4)
Floodlight
The Floodlight is an open enterprise-class, Java-based
OpenFlow Controller
Floodlight was introduced by Big Switch Networks, and is still
backed by the engineers in Big Switch networks
Floodlight offers a module loading system that make it simple to
extend and enhance
Floodlight can handle mixed OpenFlow and non-OpenFlow
networks – it can manage multiple “islands” of OpenFlow
hardware switches
Floodlight is designed to be high-performance – is multithreaded
from the ground up
Networking Laboratory 14/78
OpenFlow (2/4)
Flow Table and Flow Entries (1/3)
An entry in the Flow-Table has three fields:► The rule
Defines the flow, and rule consists of mainly field in the packet header
► The action
Defines how the packets should be processed
► Statistics
Count the number of packets and bytes for each flow, and the time since
the last packet matched the flow
Networking Laboratory 15/78
OpenFlow (1/4)
OpenFlow is an open API that provides a standard interface for
programming the data plane switches
Provides open interface to “black box” networking node (ie.
Routers, L2/L3 switch) to enable visibility and openness in
network
Networking Laboratory 16/78
OpenFlow (3/4)
Flow Table and Flow Entries (2/3)
Flow table pipeline in OpenFlow switch, where packets are matched against multiple tables in the pipeline
Per-table packet processing
Networking Laboratory 17/78
OpenFlow (4/4)
Flow Table and Flow Entries (3/3)
Networking Laboratory 18/78
How OpenFlow works
Upon arrival on the OpenFlow switch, the packet will be match by
a Flow Entry in the Flow Table, so the action will be executed if
the header field is matched, and the counter is updated
If the packet doesn’t match any flow entry, it will be sent to the
controller over the secure channel
Networking Laboratory 19/78
Introduction to Mininet (1/3)
A virtual network environment that can run on a single PC
► Mininet uses lightweight virtualization to make a single system look like a
complete network
► Runs real kernel, switch, and application code on a single machine
► Internally uses Linux containers to emulate hosts, switches, and links
Command-line, UI, Python interfaces
Many OpenFlow features are built-in
► Useful: developing, deploying, and sharing
Networking Laboratory 20/78
Introduction of Mininet (2/3)
Platform Advantages Disadvantages
Hardware Testbed
realityaccurate: "ground truth"
expensiveshared resource?hard to reconfigurehard to change
Simulator inexpensive, flexibledetailed (or abstract!)virtual time (can be "faster" than reality)
may require app changesmight not run OS codemay not be "believable"may be slow/non-interactive
Emulator inexpensive, flexiblereal codereasonably accuratefast/interactive usage
slower than hardwareexperiments may not fitpossible inaccuracy
Networking Laboratory 21/78
Introduction of Mininet (3/3)
Why Use Mininet
► Fast
► Possible to create custom topologies
► Can run real programs (anything that can run on Linux can run on a Mininet
host)
► Programmable OpenFlow switches
► Easy to use
► Open source
Alternatives to Mininet and their limitations
► Real system: Pain to configure
► Networked VMs: Scalability
► Simulator: No path to hardware deployment
Networking Laboratory 22/78
SDN Testbeds (1/3)
GENI
GENI is an open infrastructure for at-scale networking and
distributed systems research and education that spans the US
Networking Laboratory 23/78
SDN Testbeds (2/3)
OFELIA
OFELIA – Pan-European OpenFlow Testbed
• Berlin, Germany (TUB)
• Ghent, Belgium (IBBT)
• Zurich, Switzerland (ETH)
• Barcelona, Spain (i2CAT)
• Bristol, UK (UNIVBRIS)
• Catania, Italy (CNIT)
• Rome, Italy (CNIT)
• Trento, Italy (CREATE-NET)
• Pisa, Italy (CNIT, 2 locations)
• Uberlândia, Brazil (UFU)
• Castelldefels, Spain (CTTC)
Networking Laboratory 24/78
SDN Testbeds (3/3)
KOREN
A multi purpose Korean
backbone network,
connecting 8 metropolitan
areas (Seoul, Suwon,
Daejeon, Gwangju,
Daegu, Busan, Gangwon,
Jeju) from 10Gbps to
160Gbps
Networking Laboratory 25/78
Network Virtualization with SDN (1/3)
Network Virtualization
► The goal of network virtualization is to improve speed, automation and
network management by adding new software elements
► It divides a network into multiple segments or create software-only networks
between virtual machines (VMs)
► It's clear that network virtualization and SDN technologies share common
elements
► Whether network virtualization is a subset of SDN or SDN is a subset of
network virtualization, the two have overlapping sets of technologies with
similar goals
Virtualization Platform
Physical Network
Virtual Networks
Networking Laboratory 26/78
Network Virtualization with SDN (2/3)
OpenVirteX (OVX) (1/2)
Network virtualization allows multiple tenants to occupy the same
network infrastructure
Each tenant has an illusion that they have a full network to their
own disposal
OVX achieves this by giving each tenant access to a virtualized
network topology
OVX implements network virtualization as a proxy that sits
between the network and the tenant’s Controller
OVX rewrites OpenFlow messages to translate between what a
tenant’s Controller sends and receives from its virtual network
Networking Laboratory 27/78
Network Virtualization with SDN (3/3)
OpenVirteX (OVX) (2/2)
Mapping between
them and the
physical and
virtual network
Virtual topologies
Physical switches
Networking Laboratory 28/78
SoftRAN [Gudipati et al.’13] – Improved
RAN (Radio Access Network) Design.
► Improved radio resource management while
maintaining connectivity for users
► Abstracts all radio resources of a geographical
area (cell)
► Suggests moving base station closer to mobile
client
SDN in Cellular Networks (1/3)
CellSDN [Li et al.’12] and SoftCell [Jin et
al.’13] are efforts to improve the cellular
core network using SDN.
► Current cellular networks consist of Base
Stations (BS)
► BS connects to mobility manager – Serving-
GW(S-GW)
Networking Laboratory 29/78
SDN in Cellular Networks (2/3)
SDN in Cellular Networks, SDN in 5G
► MobileFlow1, CellSDN2 – Software Defined
Mobile Networks (Architecture/Framework)1
► The authors of 3 have proposed to build and
deploy an open—but backward
compatible—wireless network infrastructure
that can be easily deployed on college
campuses worldwide.
► Radio resource sharing between different
cellular operators as well as cellular and
WiFi operators4.
1 Pentikousis K et al, "Mobileflow: Toward software-defined mobile networks", IEEE Communication Magazine 2013.2 L. Li et al, “Towards Software Defined Cellular Networks”, EWSDN 2012.3 K. Yap et al., Blueprint for Introducing Innovation into Wireless Mobile Networks, ACM VISA, 2010.4 Toktam Mahmoodi et al, “Radio resource sharing as a service in 5G: SDN approach”, in IEEE Trans. NSM 2015 (under review).
Networking Laboratory 30/78
SDN in Cellular Networks (3/3)
Future Trends - SDN in Cellular Networks, SDN in 5G
► SDN based control plane in 5G.
► Handling of Increased Mobile traffic volume using SDN paradigm.
► SDN assisted data offloading for 5G networks.
Networking Laboratory 31/78
OpenFlow Wireless [Yap et al.’10] envisage a future where end
user is free from worrying about the details as to which wireless
network is serving him/her.
► Open Research problem: The design of a mobility manager capable of
servicing each customer.
OpenFlow in Wireless Networks (1/3)
Networking Laboratory 32/78
[FlowVisor - Sherwood et al.‘10] Another important feature of
SDN enabled WLANs is Slicing.
► Separates traffic flow into individual subspaces / slices. Each slice
managed by a separate controller.
OpenFlow in Wireless Networks (2/3)
Networking Laboratory 33/78
OpenFlow in Wireless Networks (3/3)
Aeroflux [Schulz-Zander et al. ‘14] improvised Odin1 by proposing
a two tier control plane
► Near-Sight Controller – Frequently occurred events (latency critical tasks)
► Global Controller – Network monitoring or load balancing
► Radio Agent exposes all functionalities
Light Virtual AP (LVAP)1 abstraction
Handling WiFi data path entries
Provide interface for statistic collection
1 Lalith Suresh et al. “Towards Programmable Enterprise WLANs with Odin”
2012, HotSDN
Networking Laboratory 34/78
Mesh Networks► [Dely et al. 2011] – Feasibility study of
Openflow in Wireless Mesh Network
► Primary challenge is frequently changing
topology
► Proposed – All nodes are Openflow
enabled mesh routers. Each wireless
interface is used for both control and data
traffic
► Plane Separation is achieved by using
different SSIDs (Service Set IDs)
Multi-Hop Wireless Networks (1/2)
Networking Laboratory 35/78
MCC [Chen and Krishnamachari 2011] – A multi channel time-
scheduled protocol for real-time data collection in WSNs.
► MCC design features centralized channel allocation and time scheduling to
address co-channel interference, and routes from sources to the sink are
centrally computed
Multi-Hop Wireless Networks (2/2)
MCC – Multi Channel Collection Protocol uses improvised balanced
routing tree formation
► There is a centralized control
plane which obtains information
about the network topology in
order to configure the
transmission and routing
decisions for all nodes
Networking Laboratory 36/78
Knowledge-Defined NetworkingMotivation
Rapid development of communication technologies lead to the
dramatic growth on complexity, heterogeneity, and scale of
networks
The need of an autonomous network that have ability
► To perform well in the partial or conflicting context
► To recognize and mediate conflicts in policies and goals
► To respond to problems and attacks as fast as possible
► To perform optimizations in high-dimensional environments that are too
complicated to be addressed by humans or any other analytical approaches
► To operate automatically or partly on behalf of human to minimize the role
of human in the network management
Machine Learning (ML) offers the most suitable approach for
implementing that autonomous network
Networking Laboratory 37/78
Knowledge-Defined NetworkingDefinition
Most ML algorithms are not designed to work in a highly
distributed context like networking
Networks are inherently distributed systems where each node
(i.e. router), has only a partial view and control over the system
SDN provides a logically centralized control and view of the
whole network for ML algorithms to gather knowledge about the
network, and exploit that knowledge to control the network
The combination of ML and SDN leads to a new research topic
called Knowledge-Defined Networking (KDN)
Networking Laboratory 38/78
A ML algorithm is an algorithm that is able to learn from data
There are 3 types of ML algorithms (as shown in the figure)
► Supervised learning: the dataset consists of “input, output” pairs and the
learning algorithm have to find out the function that map input to output
► Unsupervised learning: the dataset contains only input and the learning
algorithm has to find out the relation between samples in the dataset
► Reinforcement learning: a learning algorithm that interacts with an
environment and act in a suitable way to maximize received rewards from
that environment
Knowledge-Defined NetworkingOverview of Machine Learning (1/2)
Machine Learning
Reinforcement
LearningUnsupervised
Learning
Supervised
Learning
Networking Laboratory 39/78
ML has been gaining a lot of attention from many computer
science areas and has achieved remarkable successes
► Computer vision (object recognition, object detection, etc.)
► Natural language processing (text generation, machine translation, etc.)
► Robotics (self-driving car, humanoid, etc.)
► Social network (search engine, recommendation system, etc.)
► Network security (intrusion detection, etc.)
However, the deployment of ML in network related research is
still in the early phase
Knowledge-Defined NetworkingOverview of Machine Learning (2/2)
Networking Laboratory 40/78
Knowledge-Defined NetworkingOverview of KDN system (1/3)
Albert Mestres et al., “Knowledge-Defined Networking”, ACM SIGCOMM 2017
KDN operational loop
Networking Laboratory 41/78
Knowledge-Defined NetworkingOverview of KDN system (2/3)
Analytics platform
► Gathers information from forwarding elements to provide a complete view of
the network
► Obtains control and management state from the controller
► Keeps a historical record of the collected data
► Feeds data to ML algorithm in the Knowledge Plane (KP)
SDN controller
► Receives the decisions from the knowledge plane
► Generates policies based on these decisions to control the forwarding
elements from the data plane
Forwarding elements
► Contains routers, switches, etc.
Networking Laboratory 42/78
Knowledge-Defined NetworkingOverview of KDN system (3/3)
Machine learning
► In closed loop, the knowledge plane makes decisions on behalf of the
network operator
► In open loop, the network operator is in charge of making the decisions with
the assistance of the knowledge plane
Closed loop Open loop
Supervised Automation Optimization
ValidationEstimation What-if analysis
Unsupervised Improvement Recommendation
Reinforcement Automation Optimization
N/A
Networking Laboratory 43/78
Knowledge-Defined NetworkingBenefits
Fault diagnosis and mitigation
► Offers a better interaction between users and the network
► Exchanges information and predict when failures happen in the network
► Performs suitable action to mitigate the failure
Automatic (re)configuration
► Automatically or partly (re)configures when the condition changes
Performance management
► Continually monitors and performs prediction mechanism to reduce the
amount of traffic when the overhead occurs
Security management
► Detects abnormal behaviors and act accordingly
Networking Laboratory 44/78
Knowledge-Defined NetworkingChallenges
Representative of the data
► The success of an ML algorithm much relies on the representative of data
Ground truth
► The difficulty of providing true labels for ML dataset
Routing of knowledge and ML techniques for networking
► The problem of how can we express knowledge from a KP to others
Requirements of new ML mechanism
► There is still no specific ML for networking
Security of ML
► The problem of how to avoid malicious components that poison the dataset
Networking Laboratory 45/78
Network Function Virtualization (NFV) (1/3)
Network Functions Virtualization (NFV) is an initiative to
virtualize network functions, previously carried out by dedicated
hardware.
Networking Laboratory 46/78
Network Function Virtualization (NFV) (2/3)
▪ Reduced Power
▪ Lower CapEx and OpEx
▪ Reduced Time To Market
▪ Open Market to SW Suppliers
▪ Flexible
▪ Standard Hardware
Networking Laboratory 47/78
Network Function Virtualization (NFV) (3/3)
Source: Network Function Virtualization: State-of-the-Art and Research Challenges.
A practical example of NFV.
Traditional CPE implementations Possible CPE implementation with NFV
Networking Laboratory 48/78
NFV Architectural Framework (1/3)
NFV Management and Orchestration (MANO): Responsibility for
management VNF and service.
• Orchestrator
• VNF Manager (VNFM)
• VIM
NFV Infrastructure (NFVI)
Networking Laboratory 49/78
NFV Architectural Framework (3/3)
NFV Orchestrator
► Generates, maintains and tears down network services of VNF
themselves. If there are multiple VNFs, orchestrator will enable
creation of end to end service over multiple VNFs.
► Responsible for global resource management of NFVI resources.
For example managing the NFVI resources i.e. compute, storage
and networking resources among multiple VIMs in network.
► Performs its functions by NOT talking directly to VNFs but through
VNFM and VIM.
Networking Laboratory 50/78
NFV Architectural Framework (2/3)
VNF Manager
► Manages a VNF or multiple VNFs
► Life cycle management of VNF instances
► Can do the same functions as EMS but through open
interface/reference point (Ve-Vnfm).
VIM (Virtualized Infrastructure Manager)
► The management system for NFVI.
► Responsible for controlling and managing the NFVI compute,
network and storage resources within one operator’s infrastructure
domain
► Responsible for collection of performance measurements and
events.
Networking Laboratory 51/78
SDDC Introduction Video
Content
► Why SDDC?
► Features and Benefits of SDDC
► How to transform to SDDC?
Duration
► 2 minutes and 24 seconds
VIDEO
Networking Laboratory 52/78
RouteFlow (RF) (1/2)
Provides virtualized IP routing service over OpenFlow (OF)
enabled HW:
► OF HW needs Flow tables to make forwarding decisions
► Linux based routing engines populate the Forwarding Information Base
(FIB), used for destination lookup
► Provided mechanism to convert the FIB entries to OF Flow table entries
► https://www.youtube.com/watch?v=YduxuBTyjEw
Networking Laboratory 53/78
RouteFlow (RF) (2/2)
Virtual network environment:
► Each VM maps to one OF HW
► Server virtualization technique used to create Virtual Machine (VM) to
reproduce the connectivity of a physical infrastructure
► Each VM runs one instance of IP routing engine (e.g. Linux-based
Quagga)
► All state information (e.g. routing tables) are exchanged in the virtual
network by the routing engines
► Linux Containers (LXC)1 used as the virtualization mechanism as it is a
lightweight mechanism.
1
http://rubenlaguna.com/blog/2014/09/15/comparing-hypervisors/
http://www.itworld.com/article/2915530/virtualization/containers-vs-virtual-machines-how-to-tell-which-is-the-right-choice-for-your-
enterprise.html
https://www.flockport.com/lxc-vs-docker/
https://www.scriptrock.com/articles/docker-vs-vagrant
Info. taken from https://www.clear.rice.edu/comp529/www/papers/tutorial_4.pdf
Networking Laboratory 54/78
Route Flow Introduction Video
Content
► Introduction of Route Flow
► Introduction of different components in Route Flow
► Working of Route Flow
Duration
► 4 minutes and 39 seconds
VIDEO
Networking Laboratory 55/78
RouteFlow (RF)- Architecture(1/4)
Key Features:
► Separation of data and control planes
► Loosely coupled architecture
► Three RouteFlow (RF) Components
Controller, Server and Slaves
► Unmodified routing protocol stacks
► Portable to multiple controllers
RF-Controller acts as “Proxy” app
► Multi-virtualization technologies
► Multi-vendor data plane hardware
Taken from
http://changeofelia.info.ucl.ac.be/pmwiki/uploads/SummerSchool/Program/session_007.pdf.
Networking Laboratory 56/78
RouteFlow (RF) - Architecture(2/4)
RF Server:
► The “brain” of RF
► Manages available Virtual Machine (VM)
► Configures the virtual environment
► Receives events from RF-Controller
► Associates VMs and OF switches
► Determines packet delivery from/to VMs
► Requests flow installation / modification
in OF switches
Taken from
http://changeofelia.info.ucl.ac.be/pmwiki/uploads/SummerSchool/Program/session_007.pdf.
Networking Laboratory 57/78
RouteFlow (RF) - Architecture(3/4)
RF-Controller:
► App on NOX controller
► Sends packets and events it receives
from OF- Switch and OF HW to RF
Server
► Sends packets it receives from RF
Server to OF-Switch (control packets) or
OF HW (to deal with flow additions /
deletions)
► Provides and interface to the rest of the
framework to the OF HW
Taken from
http://changeofelia.info.ucl.ac.be/pmwiki/uploads/SummerSchool/Program/session_007.pdf.
Networking Laboratory 58/78
RouteFlow (RF) - Architecture(4/4)
RF-Slave:
► Runs as a daemon in Linux-based VM
► Registers the VM with RF-Server
► Configures the VM (e.g. interfaces)
► Listens to ARP and IP table updates via
Linux Netlink events
► Translates routing updates into flow
rules
► Translates ARP entries into flow rules
► Sends flow update commands to RF-
Server
► Agnostic to the Routing Engine
Taken from
http://changeofelia.info.ucl.ac.be/pmwiki/uploads/SummerSchool/Program/session_007.pdf.
Networking Laboratory 59/78
SDN and NFV Introduction Video
Content
► Why and how to combine SDN and NFV
► Features and benefits of SDN and NFV
► How it works
Duration
► 4 minutes and 46 seconds
VIDEO
Networking Laboratory 60/78
SDN and NFV (1/2)
Networking Laboratory 61/78
SDN and NFV (2/2)
Research Areas / Challenges:
► Two major aspects of SDN that may need to be improved in order to meet
the requirements of NFV: the Southbound API (mainly OpenFlow), and
controller designs
► OpenFlow supports only L2-L4 flow handling. L5-L7 are required to
support NFV
► Improve SDN by considering distributed architectures
► Controller design should address/include distributed network management
and scalability.
Networking Laboratory 62/78
OpenFlow provides a capability which can be used to route
flows based on any arbitrary values in the match field of the
flow entries
[Agarwal INFOCOM’13] showed how the SDN can be used
effectively for traffic engineering during deployment stage
Load balancing is a classical application of SDN traffic
engineering
► [Costanzo12] proposed energy-aware routing. By measuring the energy
output, the controller could easily route (in low traffic situation)
► Controller gauges the usage levels of devices by analyzing the statistics
► If crossing threshold, controller could propagate a new rule to relieve
heavily used node/link and distributing it to other devices
► As a result, the SDN controller balances the load in wireless network
Traffic Engineering in SDN Based Wireless
Networks
Information taken from paper – “A Survey of Software-Defined Wireless Networks” by Adam Drescher.
Networking Laboratory 63/78
In Nutshell: Research Areas in SDN (1/4)
Controller and switch design
► SDN has significant scalability, performance, robustness, and security
challenges.
Scalability and performance in SDN
► Decentralized, like the Internet, call for a control plane that is logically
distributed (in order to boost Scalability).
Controller-service interfacing (North Bound Interfaces)
► Various controllers exist but their application interfaces are still in the early
stages and independent from each other. (Frenetic, Procera, Nettle, …).
Networking Laboratory 64/78
In Nutshell: Research Areas in SDN (2/4)
Virtualization and Cloud service applications
► High demand for virtualization and cloud services. The challenges include
rapid provisioning and efficient resource management.
Handoffs
► Seamless handoff between service providers (Vertical Handover)
Monitoring and Status Report
► Estimating the different wireless channels status (delay, loss rate)
► Topology Discovery, including close access points identification
Networking Laboratory 65/78
SDN-based Cloud Introduction Video
Content
► Why SDN and Cloud
► Features and benefits of SDN-based Cloud
► How it works
Duration
► 2 minutes and 3 seconds
VIDEO
Networking Laboratory 66/78
In NutShell: Research Areas in SDN (3/4)
Information Centric Networking (ICN)
► A new paradigm proposed for the future architecture of the Internet, which
aims to increase the efficiency of content delivery and content availability.
► The driving motivation is that the current Internet is information-driven, yet
networking technology is still focused on the idea of location-based
addressing and host-to-host communication. Shift from “named hosts”
based architecture to “named data”.
► How to combine ICN with SDN towards “Software-Defined Information-
Centric Networks”?
Networking Laboratory 67/78
In NutShell: Research Areas in SDN (4/4)
Enabling heterogeneous networking with SDN
► A major challenge is efficient utilization of resources (Capacity, Radio).
► The heterogeneous characteristics of the underlying networks (e.g.,
physical medium, topology, stability) and nodes (e.g., buffer size, power
limitations, mobility) also add another important factor when considering
routing and resource allocation.
► SDN techniques largely target infrastructure–based networks.
Centralized control mechanism that is ill–suited to the level of
decentralization, disruption, and delay present in infrastructure-less
environments.
Networking Laboratory 68/78
Case study: Autonomic Computing
Attributes (1/4)
The function of any autonomic capability is a control loop that
collects details from the system and acts accordingly
Autonomic systems is a self-managing autonomous and
ubiquitous computing environment that completely hides its
complexity and provides the user with an interface that exactly
meets his needs
Four aspects of self-management often cited by IBM are,
► Self-Configuring: The ability to readjust itself “on-the fly”
► Self-Protecting: Anticipate, detect, identify, and protect itself from attacks
► Self-Optimizing: Maximize resource utilization to meet end-user needs
► Self-Healing: Discover, diagnose, and react to disruptions
Networking Laboratory 69/78
Case study: Autonomic Computing Attributes (2/4)
Origin of Self-healing Networks
It is described as a combination of self-diagnosing and self-
repairing with the capabilities to diagnose and recover from
malfunctions
Research on self-healing systems has its origin in fault-tolerant
and self-stabilizing systems research
Fault-tolerant systems handle transient and mask permanent
failures in order to return to a valid state
These systems have two distinct properties.
► The system is guaranteed to return to a legal state in a finite amount of time
regardless of interferences
► Once in legal state it tries to remain in the same
Networking Laboratory 70/78
Case study: Autonomic Computing Attributes (3/4)
Properties of Self-healing
Networking Laboratory 71/78
Case study: Autonomic Computing Attributes (4/4)
Architecture for Self-healing in SDN
We propose a self-healing system for SDN which is capable of
optimally handling the failures in SDN
SDN Controller (NOX)
Rapid
Recovery
Module
Flow
Management
Action
Management
Resource
Management
Co
ntr
ol P
lan
e
Da
ta
Pla
ne
Optimized
Self-healing
Management
Topology
Discovery &
Management
Policy
Management
Load
Balancing
Routing
Management
Netw
ork
Ma
nag
em
en
t
Ap
plic
atio
n
Sw
itch
Lev
el
Mg
mt.
Network
Statistics
Management
Forwarding
Information
Base
OpenFlow
Notification
Module
Networking Laboratory 72/78
References ONF – [https://www.opennetworking.org]
OMG – [http://sdn.omg.org]
ITU-T – [http://itu.int/en/ITU-T/Pages/default.aspx]
TTA – [http://www.tta.or.kr/English/index.jsp]
IEEE – [http://sdn.ieee.org/standardization]
BBF – [https://www.broadband-forum.org]
MEF – [https://www.mef.net/ ]
ODCA – [http://www.opendatacenteralliance.org]
OIF – [http://www.oiforum.com]
3GPP – [http://www.3gpp.org]
ODL – [http://www.opendaylight.org]
ONOS – [http://onosproject.org]
Floodlight – [http://www.projectfloodlight.org/floodlight]
OpenStack – [http://www.openstack.org]
Networking Laboratory 73/78
Standardization Activities (1/4)
Open Networking Foundation (ONF)
► A user-driven organization for SDN standardization. ONF working groups
analyze SDN requirements, evolve the OpenFlow Standard and research
new SDN standards.
Object Management Group (OMG)
► SDN Working Group for vendor neutral Global Standard.
Telecommunication Technology Association (TTA)
► A non-profit organization provides certification services to establish new
standards for the ICT industry in Korea.
Networking Laboratory 74/78
Standardization Activities (2/4)
ITU-T Study Groups (SG) work
► SG11- develops signaling requirements and protocols for SDN.
► SG13- develops the SDN framework, as baseline of all ITU-T SDN
standardization.
► SG15- lays recommendations for "Architecture for SDN control of
Transport Networks".
► SG16- governs the OpenFlow based Virtual Content Delivery Networks.
► SG17- security aspects of SDN.
ITU-T International Telecommunication Union - Telecom Standard Sector
Networking Laboratory 75/78
Standardization Activities (3/4)
The Metro Ethernet Forum (MEF)
► This forum focuses on defining service orchestration with APIs for existing
networks.
ODCA
► An organization working on unifying data center in the migration to cloud
computing environments through interoperable solutions.
OIF
► Goal is to describe the features / functionalities needed to support the
deployment of SDN capabilities in carrier transport networks.
ODCA Open Data Center Alliance
OIF Optical Internetworking Forum
Networking Laboratory 76/78
Standardization Activities (4/4)
3GPP
► The mobile networking industry 3GPP consortium is studying the
management of virtualized networks, an effort aligned with the ETSI NFV
architecture to leverage SDN
IEEE (P802.1CF project)
► SDN standardization project to drive the functionality and interoperability
of SDN.
The Broadband Forum (BBF)
► Objective is to release recommendations for supporting SDN in multi-
service broadband networks.
IEEE Institute of Electrical and Electronics Engineers
3GPP 3rd Generation Partnership Project
Networking Laboratory 77/78
SDN Current Projects (1/2)
OpenDayLight (ODL)
► A collaborative, open source project to advance SDN. ODL is a
community-led, open and industry-supported framework for Software-
Defined Networking.
OpenStack
► An open source software project that aims to produce an open source
cloud operating system. It provides multi-tenant IaaS, and aims to meets
the needs of public and private clouds regardless of size, by being simple
to implement and massively scalable.
Networking Laboratory 78/78
SDN Current Projects (2/2)
Open Network Operating System (ONOS)
► The SDN network operating system for service providers architected for
performance, high availability, scale-out and well-defined northbound and
southbound abstractions and interfaces.
Floodlight
► An Open SDN Controller is an enterprise-class, high performance
OpenFlow Controller which can handle mixed OpenFlow and non-
OpenFlow networks. It is the core of a commercial produce from Big
Switch Networks.