from virtual to high end hw routing for the adult

Copyright © 2015 Juniper Networks, Inc. 1

From virtual to high end HW routing for the adult

Sławomir Janukowicz Juniper System Engineer


Agenda

• MX Family

• High End boxes

• Virtual MX

• Use cases

• Family news


MX Family


MX family


High End boxes


The MX2020 3d universal edge router Scales to 80Tbps - the industry’s most powerful edge router 20 I/O slots – Scales to 2Tbps FD per slot Standard 19”Rack fit (45RU, 4 post rack) Support for all MX services & applications Support for all MPC’s - providing full investment protection 800Gbps per slot of fabric capacity

• Backplane future proofed to support 2x higher

Fully redundant design for all common components• Fully redundant power feeds, and power modules• DC and AC power support• Redundant Routing Engines based on Quad-Core RE-S-1800x4• Scalable, Redundant Fabric architecture

Optimized, scalable thermal design


The MX2010 3d universal edge router

• Exact same architecture as MX2020• Scales to 40Tbps• 10 I/O slots – Scales to 2Tbps FD per slot• Standard 19”Rack fit (34RU, 4 post rack)• Support for all MX services & applications• Support for all MPC’s - providing full investment protection• 800Gbps per slot of fabric capacity

• Backplane future proofed to support 2x higher

• Fully redundant design for all common components• Fully redundant power feeds, and power modules• DC and AC power support• Redundant Routing Engines based on Quad-Core RE-S-1800x4• Scalable, Redundant Fabric architecture

• Optimized, scalable thermal design


Clock Gating

Reduces Dynamic power dissipation by disabling unused logic circuitry

Voltage Scaling

Increase/Decrease voltage to conserve power

Memory Sleep Mode

Unused memory put in sleep mode

Power 1W/G by 2015 MX2020 will consume 1W

to forward 1G of traffic. Compare to 4.2W/G in 2011

Multiple levels of power redundancy

Zone, Supply, Distribution, Feed

Chassis common power amortization

Switch Fabric, RE, Fan power amortized over 20 slots

Active Power Monitoring Hotter FRUs are shut down

before it affects system performance

Fan Speed based on FRU temperature

At ambient temperature, fans consume only 20% of max power

Flexible Provisioned Power

Can provision less power if installation is guaranteed to operate at <25C or <40C

Dynamic Power Management

Drawing additional power only if a particular MIC is inserted

Chassis doesn’t pay the penalty for supporting a high powered MIC

SILICON

MX2000 Power DesignSYSTEM SOFTWARE


MPC5EShipping

24x10G SFP+ OR 6x40GE QSFP+ 2x100G CFP2 + 4x10G SFP+

240G/slot with 5-level HQoS

1M Queues per slot

128K subscribers per slot

G.709 termination

SyncE / IEEE 1588v2 (OC, BC)

Supported on all MX chassis

Non queuing SKUs also

available

Add on license for non Q SKUs


MPC6EShipping

480Gbps/slot capacity

Modular MIC Design

Port queuing mode

Per vlan queuing also

available

G.709 termination

SyncE / IEEE 1588v2 (OC, BC)

Supported on MX2K chassis

only

4 x 100GE w/ OTN

8 x 100GE OS (2:1)

48 x 10GE w/ OTN

48 x 10GE

Flexible Options


Virtual MX


VMX a Scale-Out Virtual RouterScale-out (Virtual MX)Scale-up (Physical MX)

• Optimize for density in a single instance of the platform.

• Innovate in ASIC, power and cooling technologies to drive density and most efficient power footprint.

• Virtualized platforms not optimized to compete with physical routers with regards to capacity per instance.

• Each instance is a router with its own dedicated control-plane and data-plane. Allows for a smaller footprint deployment with administrative separation per instance.

• Innovate in orchestration and management capabilities to easily deploy and manage a scale-out solution.


Virtual & Physical MX Comparison

VMX

MX

Forwarding Plane Control Plane

Trio ASIC

Trio μCode

Trio μCode compiled as

x86 instructions

= + +

= + +

x86 processor

This model enables Virtual MX to be feature parity with Physical MX


vMX Overview

Efficient separation of control and data-plane• Data packets are switched within vTRIO• Multi-threaded SMP implementation allows core elasticity• Only control packets forwarded to Junos• Feature parity with Junos (CLI, interface model, service configuration)• NIC interfaces (eth0) are mapped to Junos interfaces (ge-0/0/0)

Guest OS (Linux) Guest OS (JUNOS)

Hypervisor

x86 Hardware

CHAS

SISD

RPD

LC- K

erne

l

DCD

SNM

P

Virtual TRIO

VFP VCP

Intel DPDK


Reference Server

CPU Intel Xeon 3.1GHz

Cores Min 10

RAM 20GB

Host OS Ubuntu 14.04 LTSKernel: Linux 3.13.0-32-generic

NICs Intel 82599EB (for 10G)

QEMU-KVM Version 2.0


vMX Performance

• Up to 80G bi-directional (160G uni-directional) performance per VMX instance • 1 VCP instance and 1 VFP instance @ 1500 bytes

vMX

Tester

Test setup

• Single instance of vMX with 8 ports of 10GbE sending bidirectional traffic

• 16 cores (8 I/O, 8 packet processing)


Bandwidth License SKUs

• Bandwidth based licenses for each application package for the following processing capacity limits: 100M, 250M, 500M, 1G, 5G, 10G, 40G. Note for 100M, 250M and 500M there is a combined SKU with all applications included.

100M 250M 500M

1G BASE

1G ADV

1G PRM

5G BASE

5G ADV

5G PRM

10G BASE

10G ADV

10G PRM

40G BASE

40G ADV

40G PRM

BASE

ADVANCE

PREMIUM

• Application tiers are additive i.e ADV tier encompasses BASE functionality


Application package functionality mapping

Application package Functionality Use cases

BASE • IP routing with 32K IP routes in FIB• Basic L2 functionality: L2 Bridging and

switching• No VPN capabilities: No L2VPN, VPLS,

EVPN and L3VPN

• Low end CPE or Layer3 Gateway

ADVANCE (-IR) • Full IP FIB• Full L2 capabilities includes L2VPN,

VPLS, L2Circuit• VXLAN• EVPN• IP Multicast

• L2vPE• Full IP vPE• Virtual DC GW

PREMIUM (-R) • BASE• L3VPN for IP and Multicast

• L3VPN vPE• Virtual Private Cloud

GW


Use cases


Interdomain Carrier-of-Carriers

L2.5 Internetworking Muliticast

BUSINESS SERVICES – L3VPNs, L2VPNs, Internet Access

Leading Implementation

Robust IP/MPLS packet transport infrastructure Seamless MPLS architecture for scale and service independence Universal L2/L3 business service edge Multi-level hierarchical QoS for advanced business edge QoE

SLAs Consistent BGP-based framework for v4/v6, unicast/multicast

and inter-domain services Multicast MPLS optimizations for IPVPN and VPLS Intra-chassis redundancy for service edge resiliency VPN-aware NAT, Stateful FW, crypto Third-party validated performance and scale

Service-agnostic Any L2/L3 services at high scale Optimized for Ethernet and MPLS access Decoupled service and network architecture

Technology Innovations

E2E service restoration with 50msec guarantees Common BGP signaling for all services Stateful inter-system edge redundancy schemes Next-generation Multicast VPN over P2MP LSP

L2VPNE-Line, E-LAN, E-Tree

L3VPNIPv4/v6, VPNv4/v6, ISO-VPN

High-Touch ServicesNAT, CGNAT, FW, DPI, Video

Business Edge services MX 3D

ü

ü

ü

FR/ATM Private IPTDM E-line Elan

Legacy Ethernet IP

Deploy All VPN Services Simultaneously

Any L1/L2

MPLS

VPLS IPVPN (v4/v6)

E-Tree

IPsec

VPWS


Broadband RESIDENTIAL Edge

L2/L3 Wholesale

IPTV / Video

Broadband Internet

Residential Services MX 3D

ü

ü

ü

Subscriber Scale (via Trio Chipset)

64K / 128K subscribers per slot/chassis Up to 1M queues per line module (MPC) Multicast scale and performance

Rich Broadband Edge Feature Set

Advanced PPP & DHCP services Dynamic policy & service activation per sub Hierarchical QoS (port, VLAN, user, application) Advanced PPP and DHCP, Radius/AAA services Full IPv4/IPv6 subscriber mgt feature set Subscriber-based CALEA/Lawful Intercept

Service Innovation and Scale

Rich set of L2-L7 services DAA, sFW, NAT, CGN, DPI, IDP Inline Video Quality Monitoring Per subscriber service control.

Mobile

Residential

Business

Backbone

Service Provider’s Network

Universal Edge

L4 /L7 Services

Content Services ü

ü

In-Line Services

Jflow, NAT, L2TP (LAC and LNS)


WAN (MPLS, IP)

MX : UNIVERSAL SDN GATEWAY

VMWARE NSX BASED POD CONTRAIL SDN BASED POD LEGACY, VLAN BASED POD

L2: EVPN, VPLSL3: L3VPN, NG-MVPN

Industry leading L2-L3 LAN-WAN-Overlay Gateway

Standards based, multivendor solutions

Highly scalable, virtualized, multitenant connectivity

Vmware (VXLAN) POD

Contrail (MPLS, VXLAN) POD VLAN POD

Any to any gateway Universal SDN Gateway

Building on proven track record in major DC and SP deployments

With extensible, future proof platform capabilities

GW GW GW


ETHERNET VPN (EVPN)

A new standards based protocol to inter-connects L2 domains

Juniper leading the multi-vendor industry wide initiative

Improves network efficiency Ideally suited for Datacenter Interconnectivity Allows L2 multi-tenancy in IP fabric DC

Standards driven Datacenter Interconnect solution Support Virtualized environment and seamless inter Datacenter MAC moves Support multi-tenancy

Ethernet VPN service for enterprise customers Natively support last redundancy and dual homing active-active configuration

MARKET REQUIREMENT

LAG

BGP based state

exchange

EVPN router

EVPN router

LAN

WAN


Family news


NG-MPC OVERVIEW • Latest Gen Trio chipset with upgraded

CPU

• Full feature parity with MPC1/2/3

• HQoS upgrade SKUs available

• Universal MPC supporting all MICs

• New –IR bundle for Peering &

Aggregation roles

• Dynamic Power ManagementMODEL BANDWIDTH HQOS

MPC2E-3D-NG 80Gbps No

MPC2E-3D-NG-Q 80Gbps Yes

MPC3E-3D-NG 130Gbps No

MPC3E-3D-NG-Q 130Gbps Yes

TARGETFRS

Mainline-15.1R1JAM-14.1R4


JAM

JUNOS Agile Deployment Methodology Plug and play approach to new hardware adoption Non-disruptive introduction of new hardware Consistent features across old and new hardware


ADOPTION OF NEW HARDWARE

Test the new JUNOS release

Test the new hardware

OSS/BSS integration of

new OS

OSS/BSS integration of new hardware

Deployment of new OS(whole network) and new hardware

Stabilize

HW Released

14.2 15.1


ADOPTION OF NEW HARDWARE – JAM

Test the new hardware OSS/BSS integration of new hardware Deployment of new hardware and JAM package

HW Released

14.1 14.1


EASY STEPS TO PRODUCTION WITH JAM

Download JAM

package for 14.1R2

from Juniper.ne

t

Install JAM

package

Add instance(s) of the new HW

to the router

New H/W is

operational

STARTING STATE

• MX Router running 14.1R2• No JAM SW or HW

New H/W

NO REBOOT


Hypermode

• MPC3 MPC4 MPC5 MPC6

• L2 VPLS, L2VPN, L2circuit Layer 2 bridging and switching.

• L3 IPv4, IPv6 and MPLS forwarding L3VPN unicast and multicast

# of instructions executed/packet

Forwarding performance (PPS)

MX with universal edge feature-set i.e MX-VC, BNG, etc

MX with forwarding path optimized for infrastructure services

Hyper-mode streamlines forwarding path reducing instructions executed per packet hence increasing throughput and performance


JFLOW

JFLOW provides flow monitoring for traffic analysis, accounting and security

On MX routers, sampling can be done inline i.e. in the data path by the lookup chip PFE, services card is not required

Flow creation, Flow update, flow export done by PFE

Flow version industry standard IPFIX or v10 format

Inline JFLOW is supported for both ingress and egress flows

1:2 sampling – worst case, MPC5 1:1 MPC3 1.3 MPC4 1.8

Supported Line Cards

MPC1, MPC2, MPC3, MPC4, MPC5, MPC6


PERFORMANCE IMPROVMENTS – VPN LOCALIZATION

Improve next hop scaling to support larger number of customers and routes

Improve convergence

Core -facing FPC

CE-facing FPC 1

CE-facing FPC 2

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

CNH

Core -facing FPC

CE-facing FPC 1

CE-facing FPC 2

CNH

CNH

CNH

CNH

CNH

CNH

VRF1-TNH

VRF2-TNH

CNH

CNH

CNH

CNH

CNH

CNH

VRF2-TNH

VRF1-TNH

MODERN : TOPOLOGY AWARE FORWARDING STATE ON PFES

TRADITIONAL : SAME FORWARDING STATE ON ALL PFES

Software optimization that reduces composite next-hop footprint by 30% - improving scale

Infrastructure improvement to improve convergence Decoupling of the VPN labels/Prefixes from the LSP

label Network topology aware forwarding state on MPCs to

optimize memory - improve scaling

MARKET REQUIREMENT

VALUE PROPOSITION

Thank you

from virtual to high end hw routing for the adult

Technology

fan power

redundant power feeds

power modules dc

sleep mode power

dynamic power dissipation

redundant design

redundant fabric architecture

post rack support