Various Alternatives to achieve SDN Dhruv Dhody, Sr. System - - PowerPoint PPT Presentation

Various Alternatives to achieve SDN

Dhruv Dhody, Sr. System Architect, Huawei Technologies

Who?

Huawei

• A multinational

networking and telecommunications equipment and services company headquartered in Shenzhen.

• We are the largest

telecommunications equipment maker in the world.

• Over 140,000

employees and 21 R&D institutes.

India R&D

• First and biggest R&D

center outside of China.

• High end

communication software platforms/components/ applications.

• Over 2600 engineers.

Dhruv Dhody

• 11+ years in Huawei

(Bangalore, Beijing, Santa Clara)

• Network OS dept with

specialization in Traffic Engineering and Path Computation

• Lead a Research,

Standards & Prototypes team

• 3 RFC / 8 WG

documents / 13 Patents

SDN, Lets Recap!

Why SDN?

Computing Trends are Driving Network Change

Changing traffic patterns The “consumerization

• f IT” (BYOD)

The rise of cloud services “Big data” means more bandwidth

Complexity that leads to stasis

• Add/Move devices
• Implementing network wide

policy

• Time consuming & Manual
• Prone to errors

Inability to scale

• Link oversubscription to

provision scalability Vendor dependence

• Lengthy vendor equipment

product cycles

• Lack of standard, open

interfaces

Lim Limit itation of

• f Curr

rrent Networks

Software-defined networking (SDN) is an approach to computer networking that allows network administrators to manage network services through abstraction of lower-level functionality.

SDN - Definition

The physical separation of the network control plane from the forwarding plane, and where a control plane controls several devices.

Directly programmable: Network control is directly programmable because it is decoupled from forwarding functions. Agile: Abstracting control from forwarding lets administrators dynamically adjust network-wide traffic flow to meet changing needs. Centrally managed: Network intelligence is (logically) centralized in software-based SDN controllers that maintain a global view of the network. Programmatically configured: SDN lets network managers configure, manage, secure, and optimize network resources very quickly via dynamic, automated SDN programs. Open standards- based and vendor- neutral: When implemented through open standards, SDN simplifies network design and

• peration because

instructions are provided by SDN controllers instead

• f multiple,

vendor-specific devices and protocols.

Dynamic Manageable Cost Effective Adaptable

SDN @ ONF

• Programmatically

communicate via NBI SDN Applications

• Abstract view of network
• Controlling SDN Datapaths

SDN Controller

• Network Device

SDN Datapath

• Programmatic Control
• Capability and reporting
• Events

SDN CDPI

• Provide abstract network

views

• Direct expression of

network behavior (intent) SDN NBIs

Multi-Layer SDN

SDN's logically centralized network intelligence and ability to leverage cloud computing for almost unlimited compute power enables it to evaluate all layers of the network concurrently to determine where best to send traffic.. Today, a service is typically transported at a single layer. With multi-layer SDN, a network can transport services over the most efficient technology, not just the predefined transport technology.

• Ex. If bandwidth at a particular layer

is exhausted in some portion of the network, multi-layer SDN can evaluate options and dynamically add bandwidth from a lower layer

• r reroute traffic from upper layers

around the point of congestion.

E2E SDN

Orchestration of E2E service delivery across all network domains Distributed control planes with multiple cooperating controllers

• ƒ

EWBI (east-west) for controller to controller communications across domains

• Enhanced SBI for support of specific network

technologies and types

• Enhanced NBI for support of

customer/network applications

Network virtualization and control capability Cross layer coordination (e.g., IP routing + Optical Transport)

SDN – Alternatives

(looking beyond OpenFlow)

PCE ALTO I2RS ACTN SR

PCE – Path Computation Element

PCE is an entity that is capable of computing a network path or route based on a network graph and applying computational constraints Specializes in complex path computation across various domains on behalf of its path computation client (PCC) with enhanced scalability. Stateless PCE provides mechanisms to perform path computations in response to PCC requests.

• It utilize only the TE link information database to do this computation (TEDB).

Stateful PCE: Along with network state (TEDB), it also stores the state

• f all the computed paths or LSPs and their resources (LSPDB).
• Enhanced algorithms at stateful PCE

PCE-Initiated: Setup, maintenance and teardown of PCE-initiated LSPs from a central PCE server.

Stateful PCE

Stateful PCE as evolutionary approach to SDN.

In SP network with existing investment in IP/MPLS devices, active stateful PCE can

• ffer centralized control over the LSPs as a

simple evolutionary approach for SDN.

PCC

PCEP Protocol

PCC

Stateful PCE Server

Abstraction & Algorithms over Stateful PCE

Stateful PCE

OF-Based SDN Stateful PCE

PCE capable to compute, initiate, control and maintain the LSP. PCE Server acts as the SDN controller. Continue to use existing signaling mechanism (RSVP)

• r use segment routing (SR)

Only software update for edge router (PCC). Controlling devices using a central orchestrator Opens up the control of data flows to customizable software. New hardware might be needed New protocol - OF

ALTO – Application-Layer Traffic Optimization

Protocol provides simple mechanism to provide basic, abstract but useful network information to applications. Application can make use of this information to use network efficiently. This protocol developed on top

• f existing HTTP

(REST-ful) using JSON Usecases

• Peer to Peer File

sharing

• CDN
• Real-time

Communication

• Live media streaming

Extension to ALTO are proposed for

Abstract network topology graph Traffic Engineering Service aware parameters Calendering …

• Net-Arbiter uses stateful

PCE to query and obtain the network status

PCE

• Initially for P2P file sharing (torrents)

& CDN to get simple network cost map

• HTTP/Restful/JSON
• ALTO extensions to support DC and

network costs and events

ALTO

• Joint optimized DC application

resources along with network resources.

• DC Migration, Schedule Backup, DC

& Network Events

CSO

DC 1 DC 2 DC 3 DC 4 DC 5 Application Stratum Network Stratum Arbitration Layer

APP Arbiter NET Arbiter + PCE ALTO

PCE+ALTO for Data Centre Interconnect

PCEP

Cross Stratum Optimization (CSO) - Optimization of Datacenter and Network resources - which can only be achieved via joint effort & information exchange - to cross

• ptimize between stratums.
• Application DC resources Optimization
• Network resource Optimization

Abstract Graph

Application Controller Network Controller HYD (User region) DLI (Data Center) BLR BOM

Abstract Topology

Abstracted graph

Access to an Abstract network topology, could allow an Application to understand network in a much better way – find bottleneck, make dynamic decision with network conditions in mind…

I2RS: Interface to Routing System

• Switch programming (cross-connects)
• Forwarding (FIB)

SDN focuses on programming the data plane

• Control of routers
• Control of routing protocols
• Management of the “routing system”

There are many functions and features not covered

• Using CLI to achieve these functions is very frustrating
• Expensive, time-consuming, error-prone, risky

Existing techniques are non-standard

• Strong desire for a simple and standard approach

Need for a standard approach

I2RS

Usecases for I2RS

Programming and managing the RIB BGP use cases Traffic steering and classification DDoS mitigation Topology reading, monitoring, and control Service chaining

I2RS to use Netconf/Restconf/Yang as the base…

Netconf / Restconf / Yang

Netconf

• Network

Management Protocol

• Remote primitives

to view/manipulate data

• Encoding data as

per the data model

• Think SNMP
• Transaction based

(network wide) Yang

• Data Model
• Explicit precise

structure, syntax and semantics of the (externally visible) data

• Think MIB
• Configuration data

as well as state

• Also events

Restconf

• A REST protocol over HTTP
• accessing data defined in YANG using

datastores defined in NETCONF. Netconf/Restconf/Yang as a base to manipulate state on the device!! (I2RS)

ACTN - Abstraction & Control

• f Transport Networks
• Multi-layered multi-domain Network
• Technology, administrative or vendor islands
• Interoperability for dealing with different domains is a perpetual

problem for operators.

• New service introduction with connections that traverse multiple

domains

• Need significant planning
• Manual operations to interface different vendor equipment and technology

across IP and Optical layers.

• Facilitate virtual network operations
• Creation of a virtualized environment allowing operators to view and control

multi-subnet multi-technology networks into a single virtualized network.

• Accelerate rapid service deployment of new services
• including more dynamic and elastic services
• improve overall network operations and scaling of existing services.
• Hierarchy of controllers

ACTN

Customer B Control

Multi Domain Service Coordinator

Customer C Control Customer A Control

4

5

1

A.1 A.2 A.3 B.1 B.2 B.3 C.2 C.3

2

3

PNC

Creates abstraction topology per application/client need

4 1 6 5 2 1 1 3 6 4 3 8

A.1

A.2 A.3

2 1 4 5 3 2 3 6

C.1 C.2 C.3

Multi-domain network topology

PNC

Physical Network Control like PCE

Segment Routing (SR)

Simplicity - Less numbers of protocols and interactions and automated FRR for any topology Scale - Avoid thousands of labels and TE LSPs in the network Leverage all services supported over MPLS today (L3/L2 VPN, TE, IPv6) Bring the network closer to the applications

Segment Routing

• the source

chooses a path and encodes it in the packet header as an

• rdered list
• f segments

Segment

• an identifier

for any type

• f instruction
• Service
• Context
• Locator
• …
• Distributed

by IGP How

• MPLS
• Label Stack
• IPv6
• List of

segment in routing extension header Applications

• Through PCE

get optimized path

• Applications

can interact and make changes through PCE

• Only the

ingress needs to change the segment-list!

SR & SDN

SR supports a simple but efficient capacity planning process based on centralized

• ptimization

SR optimizes network resources by providing a very simple support for ECMP- based shortest-path flows SR provides much better scaling SR provides guaranteed-FRR for any topology SR provides ultimate virtualization as the network does not contain any application state. The state is in the packet. It is encoded as a list of segments SR provides very frequent transaction-based application as the network does not hold any state for the SR-encoded flows Application aware routing using SR

Conclusion…

There is more than

• ne way to look at

SDN.. SDN in Enterprise and Datacenter is different from SDN in carriers network Re-using existing building blocks need to be looked into…

• Retain some of the investment

made in current network

• As well as leverage benefit of

SDN

• PCE, BGP, Netconf, Yang, ALTO…
• Most Vendor companies are

exploring these options….

Inline with ODL architecture…

Support Multiple Southbound/Data-Plane Elements

OpenFlow, NetConf, SNMP, BGP-LS, PCEP, LISP etc Support IETF protocols to control physical and virtual routers and switches

• Border Gateway Protocol – Link State

(BGP-LS)

• Path Computation Element Protocol

(PCEP)

Thank You!