Решения WANDL и NorthStarдля операторов

Александр Беспалов

системный инженер

abespalov@juniper.net

Juniper’s Automation Horizon Network automation is a journey not a destination

Human Programmed Workflow,

Machine Executing Optimization

• ONLINE design, planning and monitoring

• Near real-time analytics and integrated topology

discovery

• Continuous network self-optimization

• Automated provisioning

• Rest APIs to integrate applications

Complete Human Executed Workflow

• OFFLINE design and planning

• Pull analytics, pull topology, pull configuration

• Simulation, optimization, and network trending

• Provisioning templates

Strategic Planning & modeling

AutonomousNetwork optimization

Network Design and Planner Self Aware Network

WANDL IP/MPLS View

About WANDL

• Founded in 1986 by former ATT Bell Labs R&D researchers. Headquartered in Piscataway, New Jersey

• Juniper acquisition of WANDL closed January 7, 2014.

• Main driver for the acquisition: technology transfer to optimize RSVP based networks (via NorthStar) and to monitor and design multi-vendor Layer 3 network (via IP/MPLSView)

• WANDL Software Solutions used by a number of Service Providers in EMEA

Automated Network CollectionBuild up an accurate network asset and service inventory

• Layer 3 Auto-Discovery

• Multi-vendor router discovery via IGP or TED database

• Network Data Collection

• Up-to-date Configuration, Interface, Tunnel Paths, Equipment, and other data to copy and paste production network

• Layer 2 Network Collection also supported

• Collects switches and build physical topology via Spanning Trees Reverse Engineering

1

2

3 4

5

6

Collect Network

ConfigurationData

Load Network

View NetworkTopology

View Network Data

Access NetworkDetails

GenerateReports

Architecture OverviewNetwork Management Solutions

Hardware requirements• Linux/Solaris for servers

• Windows/Mac/Linux for clients

Connectivity requirements• With network elements: ports

22,161,162

• Between server and client: 7000, 3389, 8091,…

SouthBound interface: CLI & SNMP polling and traps reception

Distributed Collectors (optional)

Central

processing serverDatabase server

(optional)

Remote Java clients

for end users

Architecture OverviewNetwork Planning Solutions

• First option• Multi-user environment sharing inputs/reports

• Ideal for large Service Providers

• Easier to maintain

• Second option• Standalone usage of the software

• Server is running on Virtual Machine on the same laptop with client

• Ideal for people on the road or power users

Central

processing server Remote Java clients

IP/MPLS View Data Model

• IP/MPLS View Data model is based on a number of tables: • Project file (mandatory)

• Router table (mandatory)

• Link table (mandatory)

• Traffic matrix table (mandatory)

• Interface table (optional)

• Router type table (optional)

• Tunnel table (optional)

• …

• Most IP/MPLS View input and output files are either CSV or text files• However traffic information stored in MySQL database for the Performance Management module

• No “strict” API for Southbound or Northbound interfaces

• Every set of IP/MPLS View input files has a unique runcode, aka extension. All files for a given network normally share the same runcode

Various Network Topology ViewsSelect Filter > protocol

Select Network > Maps > Map(BGP View) Select Network > Services > VPN

Select windows > Map(standard)

Detailed Network InformationNode, Link, LSP

Traffic Matrix Solver (T-Solve)

• Simple interface traffic measurements are NOT sufficient for planning.

• T-Solve intelligently estimates a plausible set of end-to-end traffic flows• From a combination of partial available measured information

• interface traffic, LSP, LDP, NetFlow,

• Combined with customer’s own estimates and knowledge of traffic patterns

An End-to-End Traffic Matrix is crucial for

accurate network planning and simulation:

• Failure Simulation

• Capacity Planning

• Traffic Projection

• Network Design

3D SKU Selection ProcessFeature set

Standard

S-MPLSV-S

Premium

S-MPLSV-P

Deluxe

S-MPLSV-D

Super Deluxe

S-MPLSV-SD

Light weight package for entrance

level to collect configuration & build up

topology

Typical user with added traffic

statistics, security and performance

management, & modeling simulation

Typical power user with added L3

Multicast and fault management

Most advanced user with all

capabilities

• Management Essentials

• Configuration Management

• Network Health Audit

• Compliance Assessment

• Management Essentials

• Configuration Management

• Network Health Audit

• Compliance Assessment

• Performance Management

• Security Management

• Modeling Essentials

• MPLS TE

• VPN

• BGP

• T-Solve

• Device Inventory

• Management Essentials

• Configuration Management

• Network Health Audit

• Compliance Assessment

• Performance Management

• Security Management

• Modeling Essentials

• MPLS TE

• VPN

• BGP

• T-Solve

• Device Inventory

• PM-TCA

• Fault Management

• L3 Multicast

• Management Essentials

• Configuration Management

• Network Health Audit

• Compliance Assessment

• Performance Management

• Security Management

• Modeling Essentials

• MPLS TE

• VPN

• BGP

• T-Solve

• Device Inventory

• PM-TCA

• Fault Management

• PM-SLA

• PM-MIB

• MIND

• Template Based Configuration

• Adv: MPLS TE LSP

• High Availability

3D SKU Selection ProcessCapacity

Network Node Size Unit

-50 Additive up to 50 nodes

-500 Additive up to 500 nodes

Admin & Concurrent Viewer

Normal Permits 2 Admin users & 5 concurrent viewers

Large Permits 4 Admin users & 20 concurrent viewers

2nd dimension, Number of Network Nodes

3rd dimension, Number of Admin Users & Concurrent Viewers

WANDL IP/MPLS ViewUse Case Scenarios

Network Inventory• Automate the equipment inventory reporting by importing “show

commands” e.g. show chassis hardware• Does not require to have purchased the network management modules as third party configuration software

can be used for input purposes

• Multi-vendor support: Juniper, Cisco, ALU, Huawei, Redback, Extreme, Brocade,…

• Very detailed reporting of all equipment components including Part Number/Serial Number• SFP/XFP

• Fan, Power Supply, PICs,..

• Visualization of equipment layout

• Provide Reports about Equipment Usage• Number of ports used per card

• Available number of ports per card

Compliance Assessment Tool

• Conformance checking of configuration files

• User-Defined Templates (rules) to check conformance• Support wildcards and regular expressions

Network Failure Simulation

Network Failure Simulation (results)

Interactive Failure Scenario

Interactive Failure Scenario (results)

Forecast and Capacity Planning Assessment

• Validation of capacity planning for next 12-24 months

• Ability to prune and add links in methodical way

• View impact of converged network

Customer Benefits

• Optimized topology

• SOW for implementation of Juniper Solution

• On-site workshop with customer reviewing results

Deliverables

High Availability Assessment

• Validation of network redundancy and consistency of configuration

• Understand impact under failure conditions

• Single node failure

• Single link failure

• Network outage Simulation

• Identify Single points of failure

• Potential Congestion points

• Performance of QoS under congestion scenarios

Customer Benefits

• Report on topological redundancy and capacity under simulated

failure conditions

• SOW for implementation of Juniper Solution

• On-site workshop with customer reviewing results

Deliverables

NorthStar

NorthStar WAN ControllerWANDL + Junos Integration

IP/MPLS

LayerTransport

Layer

SOLUTIONNorthStar combines the path computation expertise of

WANDL with the dynamic IP control plane of JUNOS.

Network

Modeler (IP /

MPLS / MIND) Static

Centralized

GUI

IP control

plane (JUNOS) Dynamic

Distributed

CLI

Active Stateful

PCE (NorthStar ) Dynamic Path

Computation

Optimized IP /

MPLS routing

Information Available for Path CalculationMore information enables optimized path calculation

Shortest Path(IGP, LDP)

Distributed TE(constrained SPF)

+

Reservable

bandwidthTopology

+

Centralized TE +

Reservable

bandwidthTopology

+

Complete view of

packet demands

Fragmented view

of packet demands

Topology

Centralized Packet Traffic Engineering

Application driven

routing constraints (latency, bandwidth, cost,

QoS, etc..)

Optical and packet

topology, as well as

reservable bandwidth

Centralized

path

computation(Maximize objective

function)

Complete view of

packet demands

Application-aware

traffic engineering

LSP Blocking Problem

20/100

0/50 0/50

PE1

PE2

PE3

X/Y:

X=reserved bw

Y=total link bw

Suppose we also want an LSP from PE2 to PE3 with BW=90. It is blocked by

presence of green LSP.

Centralised path computation would solve the problem.

bw=20

Live-Live Multicast Data Delivery

Source, S1 Source, S2

Ingress PE Ingress PE

Egress PEsEgress PEs

ReceiverReceiver

The two P2MP LSPs must not have any nodes or links in common

PCE: A Standards-Based Approach

• PCE: Path Computation

Element (RFC 4655)An entity that can calculate paths in the

network. Can be a network node, a server, an

application, etc.

•PCE: Path Computation ElementComputes the path

•PCC: Path Computation Client Receives the path. Signals LSP using RSVP

•PCEP: PCE protocol (RFC 5440)For PCE/PCC communication

•Path signaling & definition in the networkStatic, RSVP, IGP-Labels

ComponentsWhat is it?

PCEPCEP

PCC PCC

PCC

Stateless vs. Stateful PCE

PCE

Stateless PCEPCE has no memory of LSPs

previously setup in network

PCE

Stateful PCEPCE remembers LSPs

Passive Stateful vs. Active Stateful PCE

PCE

Passive stateful PCEPCE only computes new ERO for

an LSP if asked by PCC.

PCE

Active stateful PCEOnce PCC delegates control of LSP to

PCE, PCE can proactively modify ERO

whenever it sees fit.

Please

compute a new

ERO for this

LSP.

Use this new

ERO for this

existing LSP.

Delegated Control vs. PCE-Initiated LSPs

PCE

Delegating LSPs

PCE

PCE-initiated LSPs

Here are the

characteristics of this LSP.

From now on, if you

update it, I will re-signal it

with new properties.

Feel free to

initiate new

LSPs.

OK. Signal an

LSP with these

properties.

Different LSP TypesWhen using active stateful PCE

“Vanilla” LSP• Configured on ingress router (i.e. exists in router config file)

• Path computed by ingress router using CSPF

• Existence of LSP, and associated parameters, reported to PCE for PCE’s info only – PCE cannot modify the LSP

Delegated LSP• Configured on ingress router (i.e. exists in router config file)

• When first set up, path is computed by ingress router using CSPF

• Delegated to PCE – means that PCE can modify parameters from time to time e.g. bandwidth, ERO

PCE-initiated LSP• PCE decides to create LSP

• Sends set-up request to ingress router via PCEP, with ERO, bandwidth etc.

• Automatically delegated to PCE –means that PCE can modify parameters over time

• Is ephemeral – does not exist in router’s config file

Advantages of PCEP over CLI

• PCEP makes it easier to make changes to the network:• CLI requires commit to modify LSP state whereas PCEP creates/modifies

LSP state without commit.

• PCE GUI makes it easier to keep an overview of LSPs in the network, prevents a lot of “history” to accumulate in the network

• PCEP is a standardized interface:• PCEP interface is standardized by the IETF. Any PCE can talk to any PCC.

• CLI varies across vendors, making it more difficult to control a multi-vendor network

• Northbound interface of PCE allows easier integration with OSS, etc…

BGP Link StateNetwork-wide learning of topology and TE info

??

Area 0

(Level 2)

Area 20

(Level 1)

Area 10

(Level 1)

? ?

TE

Controller

NorthStar ControllerSouthbound interfaces for multi-vendor interoperability

SOFTWARE-DRIVEN POLICY

Topology Discovery Path Computation Path Installation

PCEP Install traffic engineered LSP

One session per head node

Netconf/YANG (may include: BGP, XML, OpenFlow, I2RS) Other state injection methods

ANALYZE OPTIMIZE VIRTUALIZE

PCEP Netconf/YANGPCEPCSPF Algorithms

RSVP Signaling

RESTful APINorthStar Controller

IGP / BGP-LS

PCEP LSP discovery

IGP (ISIS, OSPF) One session per domain

BGP-LS

Juniper NorthStar ControllerHardware platform (CSE2000)

Hardware

• 2U chassis (NEBS / ETSI compliant)

• Dual CPUs with multi-core and multi-thread computing platform

• 2x10GE IO per computing blade (hardware capable of extending to 6x10GE)

• Future 64 x1/10GE switch blade (dual rate)

Performance and Services Scale

• Multi-appliance tethering to routers (PTX first)

• High resiliency and flexible clustering between the CSE2000s to future-proof traffic growth and the next-gen platforms

Unified Cluster View

• Single admin point

• Operational simplicity

Higher Feature Velocity

• Provides network-level service infrastructure support such as JFLOW, IPFIX, and CGNAT, Virtual Route Reflector, and PCE Server

• Natively running third-party applications

How to Setup a PCE-Computed LSPSignaled vs. pushed end-to-end paths

NL

A

M

BY

C

D

Z

PCEP

RSVP

LSP

NL

A

M

BY

C

D

Z

OpenFlow

Packet path

Path

setup

Centralized

path

computation

Path

setup

Centralized

path

computation

RSVP path signaling OpenFlow path insertion

Why Use PCE for Network Programmability?

PCE allows an evolutionary approach towards centralized control of network

infrastructure:

• No need to replace entire control and forwarding infrastructure (as with Openflow) - only the

edge nodes need to support PCEP

• Can continue to use the same protocols for signaling (RSVP-TE) and same schemes for

mapping traffic to paths at the edges

• Only needs a SW upgrade to enable the client functionality

• Local repair capability in case of network failures using the distributed control plane and with

established protocols, this is difficult with a fully centralized approach

• Can easily work in a hybrid architecture where part of the traffic is managed via a central

controller and part of the traffic uses the conventional distributed approach

OpenDaylight High-Level Architecture

Relevant for

carrier SDN

NorthStarUse Case Scenarios

Offline Network Planning Together With WANDL

Dynamic topology / LSP acquisition through BGP-LS / PCEP allows NorthStar to build a network model for

visualization and off-line planning:

• Export of the topology to WANDL for network modeling and capacity planning

• ‘What if’ analysis in planning phase when new links / nodes are deployed in the network

• Exhaustive failure analysis to verify network resilience and worse-case failures. Enabling network optimization

based on maximum load under worse-case failures

NorthStar

Release 1

Year 1

Year 3

Year 5

Deployed Extension

WANDL

NorthStar

PCEP

BGP-LS

LSP Path Diversity (1/2)

What is the benefit from LSP path diversity:

• Eliminates risk that a primary and secondary link are both broad down by a single failure in the network. This is key to guarantee strict SLAs for business services

• Eliminate the risk of Shared Risk Link Groups (SRLGs) on the transport layer

• Path diversity allows for differentiated services by offering customers the choice between 1:1, 1+1 and/or 1:N protection options

London

Paris

Warning!Shared Risk

Shared RiskEliminated

Primary Link

Secondary Link

Primary Link

Secondary Link

London

Paris

NorthStar

Release 1

LSP Path Diversity (2/2)

NorthStar will support the creation and deletion of LSP tunnels that are disjoint as part of diverse path computation:

• Two or more wholly disjoint LSPs (only 2 disjoint paths in NorthStar 1.0)

• Set of diverse LSPs with the same end-points or with different end-points

• Site, Link or SRLG disjointness

• Can be used as back-up path and/or for multiple active load-balancing paths

NorthStar

Release 1

CE

CE

LSP-1 LSP-2

P2MP Path Diversity

The two P2MP LSPs have different ingress PEs and must not have any nodes or links in common

• No way of ensuring path separation when ingress nodes calculate the paths

• However, a central controller can ensure separation when calculating multi-point trees as well as ensuring the paths are optimized (minimize capacity, latency, etc...)

NorthStar

Future Release

Source, S1 Source, S2

Ingress PE1 Ingress PE2

Egress PEsEgress PEs

ReceiverReceiver

Maintenance-Mode Re-Routing

Automate re-routing of traffic before a scheduled maintenance window:

• Eliminate risk that services are mistakenly affected when a node / link goes into maintenance mode – important to guarantee strict SLAs

• Automatically path restoration after maintenance has been completed

NorthStar

Release 1

Node tagged as going into maintenance mode, affected LSPs are identified by NorthStar.1

1

3

XX

X

2

2LSPs are re-computed and re-signaled around the affected node through a PCEP update (all

make-before-break). Node is subsequently moved into maintenance mode.

3 Node is tagged as available after maintenance window finishes. Optimum LSP paths are

restored through PCEP update and re-signaling.

Programmable Path Optimization

Application-aware routing: optimum path is decided based on application specific requirements

• Today cost metrics are most often based on number of hops (minimize port cost) or distance (minimize transport cost and latency)

• NorthStar can use additional ‘cost’ functions for path computation, which are stored in a topology file, to decided the optimum path for each LSP

NorthStar

Release 1

Lowest latency path

Lowest number of hops path

Latency based

metricsInterface cost

based metrics

5

5 3

2

32

2

3

4

3

4

3 4

5

AZ

1

1 1

1

1

1

11

1

1

1

1

1

1

ZA

Enhanced Traffic Engineering (TE++)

NorthStar can enhance traffic engineering as it provides a network wide visibility to traffic demands:

• Centralized path computation can therefore improve ‘bin packing’, i.e. how to place LSPs such that the available bandwidth is used in the most optimal way

• NorthStar leverages TE++ for enhanced traffic engineering. This is a JUNOS feature that can dynamically create/tear down LSPs based on the traffic load. NorthStar especially improves visibility and manageability of LSP optimization

NorthStar

Release 1

(Cost = 1, BW

= 10)

(1,10)

(1,10)

(1,10)

(10,5)Demand = 5G

Demand = 10G

CSPF failure capacity upgrade / EROs?

(Cost = 1, BW

= 10)

(1,10)

(1,10)(1,10)

(10,5)

5

5

Demand = 5G

Demand = 10G 55

A

B

C

D

E

A

B

C

D

E

Bandwidth Calendaring

Bandwidth calendaring allows network operators to reserve resources up-front or for a dedicated period of time:

• Bandwidth calendaring enables highly accurate usage based charging for bandwidth

• Reduces the need for on-side configuration at customer premises, for example when upgrading bandwidth

NorthStar

Release 1

London

2.5G ELINE

Paris

London

1G ELINE

NorthStar 3 Saturday from 1am to 3:30amRepeat Every Week

4

1

Operator Load Calendar Event

2

Paris

Customer places an order to increase their E-line Service from 1G to 2.5G this upcoming Saturday1

3 The Network Operation Group will execute the Service Request through the NorthStar GUI or North Bound REST API

4 NorthStar will increase the customer’s E-Line service to 2.5G on Saturday at 3:30am

2 The Service Order is converted into a Service Request, which can be implemented through a customer portal.

Hierarchical LSPs (H-LSPs)

Full-mesh of RSVP LSPs scales with O(N2) in RSVP state, resulting in possible scaling problems for large networks with 100’s or 1000’s of nodes:

• Hierarchical LSPs (H-LSP) are an efficient approach to reduce RSVP state in the core

• Well suited for the server layer in a multi-layer network, for example between edge and core or in a SuperCorearchitecture that interconnects multiple networks

H-LSPs have long since been natively supported by Junos:

• Manual configuration of H-LSPs and mapping of PE-PE LSPs into H-LSPs is operationally complex

NorthStar can compute and configure H-LSPs automatically:

• Mapping of PE-PE LSPs into H-LSPs happens automatically

• Improves both the scalability and manageability of RSVP

NorthStar

Future Release

NorthStar

LSR1LSR3

Design & Planning Tool

• Topology Construction (Configuration parsing)

• Device Inventory Management

• Network Health Audit

• Compliance Assessment

• Traffic Management & Monitoring

• Health Monitoring

• Fault Management

• Protocol Simulation

• What If analysis

• MPLS TE LSP Provisioning (CLI)

• P2MP-TE

• Template Based Provisioning (Limited)

SDN Network Controller

• Topology Construction (IGP, BGP-LS)

• MPLS TE LSP Provisioning (PCE)• Many Use Cases

• Restful/API

• Netconf/Yang

• Futures (planned)• Multi-layer Optical Integration

• Third party portal – Integration with SAD

• Web GUI

• Flow mapping

• SPRING

Juniper WANDL IP/MPLS View Juniper NorthStar

IP/MPLS View & NorthStar

Спасибо