Delivering
High- Performance
for Carrier Ethernet
Emil Gągała
SOA Collaboration Server Virtualization Virtual Desktops Unified Comms LAN Access Control
Network Administration Interface
Service Enablement Enterprise Network Architecture
Employee Guest Customer Partner Managed/Hosted Datacenter Datacenter Campus Branch Remote WAN SaaS
Enterprise Carrier Ethernet
WAN Core Network
Internet Peering
Metro Ethernet
Carrier Ethernet Requirements
Technology to meet carrier’s needs
•
Services
reliability
&
predictability
•
Network
Carrier-grade operations from core to metro
Scalability for Ethernet enabled services
•
As well as cost efficiency
Breaking the layer-2 low-margin lock-in
•
Ability to move from cost/bit to value/bit services
performance
and
optimization
scalability
and
stability
Multicast
Juniper’s Transport Services Approach
Any to Any connectivity - Any Service on Any
Port
•
Point-to-Point:
Legacy
, Ethernet
•
Point-to-Multipoint: Ethernet, IP, Broadcast/Multicast
•
Multipoint-to-Multipoint: Ethernet, IP
VPLS
TDM
FR/ATM
L2VPN
IP
IPVPN
MPLS
Any L1/L2
Ethernet
IP
Legacy
}
MPLS is
where the
services
are
E-TREEE-LAN
E-LINE
Multicast optimized:
Enhanced E-LAN
No MAC
learning
needed
MPLS Components
VLAN Components
“MPLS is too Hard” Objection Handling
Extending VLAN Knowledge to MPLS
VLAN segmentation is
localized and limited in scale
VLAN Tags (4 bytes)
•
16-bit PID, 3-bit Priority,
1-bit CFI, 12-bit VLAN ID
Layer 2 Segmentation
Spanning Tree Protocol
Active/Blocking
VLAN Trunking
VLAN ACLs
802.1p QoS Markings
Ethernet failures/repairs
Allows network-wide segmentation
with very large scale
MPLS Label stack (4 bytes)
•
20-bit Label, 3-bit QoS (EXP), 1-bit
bottom of stack flag, 8-bit TTL field
Layer 2 & Layer 3 Segmentation
OSPF/LDP
ECMP
LSP Switching
IP ACLs
DSCP/EXP QoS Markings
Juniper Innovation and Know-How
IQ2
Ethernet Services Engine
PICs
Multiplay Subscriber Edge
Terabit Core
Multi-service Business Edge
Carrier
Ethernet
Carrier
Ethernet
ASIC Expertise
Industry Leading Mad Scientists
(aka Protocol Developers)
The JUNOS Software Difference
Operationally deployed since 1998
•
First high-performance network Operating System
11+ years of innovation and development
•
4 releases per year
•
TL 9000 certification from QuEST Forum
•
Thousands of features address diverse needs
Serving most demanding customers
•
Installed in ‘top 40’ service providers as well as many
high-performance enterprise & public sector accounts
One OS
Q408 9.3 9.4 Q109 9.5 Q209One Release
Module XOne Architecture
A P IMX-Series Portfolio
10
GE
Su
pp
or
t
T
b
p
s
+
480G
240G
Optimized for Carrier Ethernet and Ethernet Edge Requirements
Optimized for Carrier Ethernet and Ethernet Edge Requirements
Medium PoP
960G
Large PoP
Small PoP
Sm/Med PoP
MX Development Strategy - Leverage
1.
Leverage the I chip (Packet Forwarding Engine)
For cost-optimization & density
2.
Leverage T-1600 fabric chip
For performance & density
3.
Leverage JUNOS
For Carrier Class features and stability
4.
Leverage “Ethernet Services Engine” (IQ2)
MX-series: architecture
+
IQ2
+
=
MX-series
Full-service IP/MPLS forwarding
Per-customer per-service QoS
Same mature and scalable JUNOS
MX in the Juniper Portfolio
T-series
Provider core
M-series
Multi-interface service edge
MX-series
Carrier Class Ethernet L3/L2
MX-series: JUNOS to Ethernet-centric Deployments
Access Network
CPE
1 2 3 4 5
0
MTU/CPE
Access
Aggregation
Metro Network
Provider
Edge (PE)
Core NetworkCore
M & E-series,
MX-series
M & E-series,
MX-series
MX-series
MX-series
T-series,
M-series,
MX-series
T-series,
M-series,
MX-series
J-series
J-series
EX-series
MX-series
EX-series
MX-series
J-series
EX-series
J-series
EX-series
platforms
The MX-series
16 RU/ (3 per rack) 8 RU/ (6 per rack) 5 RU (9 per rack)Physical dimensions
48 / 480 24 / 240 12 / 12010 GigE / Gig E ports
960 Gbps 480 Gbps 240 Gbps
Capacity
MX960 MX480 MX240MX Product Family – High Density, Ethernet-only platforms
•
Same DPCs, SCBs, REs
• Different chassis, fan-trays and power-supplies
•
Same JUNOS across M/T/MX
•
Same extensive VPN portfolio
L2 features for seamless operation
Adding L2 features to JUNOS & MX
Concurrently supported on a port
Virtual Switches
STP
MST
RSTP
PVST
VPNs:
E-LINE & E-LAN
Multicast optimized E-LAN
Layer2 to Layer3
IP-VPN
Multicast optimization
P2MP technology
Hardware-based mcast
Ethernet OAM
802.1ag
802.3ah
Distributed
MX960 Ethernet Services Router
14 Slot Chassis
Physical size
•
Height: 16RU (about 1/3 rack), Depth: <800mm deep
Dependable hardware
•
Passive Mid-Plane
•
Redundant Routing Engines
•
Redundant Switching Fabric (2+1)
•
Distributed Packet Forwarding Architecture
•
Redundant Fan & Power
Power and cooling
•
Front-to-back cooling with separate push-pull fan
assemblies
•
Holds up to 2 fan trays (1+1 redundancy)
•
Holds up to 4 power supplies (2+2 DC, 3+1 AC)
•
Rear-side power cabling
System capacity
•
14 slots - 2 for Fabric Cards / REs with the option of 1
additional SCB for redundancy
MX480 Ethernet Services Router
8 Slot Chassis (6+2)
Physical size
•
Height: 8RU (about 1/6 rack), Depth: <800mm
deep
Dependable hardware
•
Passive Mid-Plane
•
Redundant Routing Engines
•
Redundant Switching Fabric (1+1)
•
Distributed Packet Forwarding Architecture
•
Redundant Fan & Power
Power and cooling
•
Side to Side cooling
•
Holds single fan tray
•
Holds up to 4 power supplies (2+2 DC, 2+2 AC
240V, 3+1 AC 110V)
•
Rear-side power cabling
System capacity
•
8 slots - 2 for Fabric Cards / REs
MX240 Ethernet Services Router
4 Slot Chassis (2+2 or 3+1)
Physical size
•
Height: 5RU, Depth: <800mm deep
Dependable hardware
•
Passive Mid-Plane
•
Redundant Routing Engines (2+2 configuration)
•
Redundant Switching Fabric (1+1)
•
Distributed Packet Forwarding Architecture
•
Redundant Power
Power and cooling
•
Side to Side cooling Holds up to 2 fan trays
(1+1 redundancy)
•
Holds up to 4 power supplies (1+1 DC, 1+1 AC
200-240VAC, 2+2 AC 100-110VAC)
•
Rear-side power cabling
System capacity
•
4slots – 2 available for Fabric Cards / REs
•
Up to 120Gbps (full-duplex) from 3 line cards
•
System reuses existing SCBs, REs, and DPCs
– common across all MX platforms
MX-series Switch Fabric
SF chip: Next generation Juniper switch fabric ASIC
Design requirements:
•
High performance
•
Non blocking
•
Low latency
•
Fault tolerant
Simple, scalable architecture also used by the M120, M320,
T320, T640
10G
10G
10G
10G
10G
10G
10G
10G
PFE
PFE
PFE
PFE
Switch
Switch
Fabric
Fabric
MX system
SCB 1
Routing
Engine
SF
SF
SCB 0
Routing
Engine
SF
SF
Control Plane Architecture
Completely redundant
control network from each
RE to each DPC
GE connectivity for control
plane performance &
scaling
Switch fabric failover
completely independently of
REs
GE switch
I
I
I
I
DPI
I
I
I
DPI
I
I
I
DPI
I
I
I
DPRE options
1.3GHz processor
2GB memory
2.0Ghz processor
4GB memory
40GB drive
Not interchangeable
with M or T-series
High Density DPC Architecture
Dense Port Concentrator: SFPs or XFPs
Line rate connectivity to the switch fabric
4 packet forwarding engines (PFEs) per DPC
MX480
DPC
I
I
I
I
I
ESE
ESE
ESE
ESE
Optimizing for Two Different Ethernet Applications
MX-R
Optimized for
full L3 routing
and services
Full L2/L2.5
transport
MX-X
Optimized for
L2/L2.5
transport
Supports full
IP/MPLS
MX-Series
Common linecards
Common
Switch/Control boards
Common REs
MX-R
MX-X
Dense Port Concentrator Types
4 Port 10GE XFP
20 Port 10/100/1000 SFP
40 Port 1GE SFP
20 GE SFP & 2 10GE
XFP
40 Port 10/100/1000
Copper
2 Port 10GE XFP
----X
X
X
X
----R
R
R
R
----Q
Q
Q
Q
Density
MX240
Density
MX480
Density
MX960
120
ports
ports
240
ports
480
120
ports
ports
240
ports
480
60
ports
ports
120
ports
240
12
ports
ports
24
ports
48
60 & 6
Multi-Rate DPCs
Multi-Rate cards combine 10GE and 20
ports of 10/100/1000 Ethernet SFP
•
Reduces space and power requirements for smaller
configurations. Increases MX-series platform flexibility
and deployment options.
• Both CAPEX and OPEX savings
Three card versions:
•
Transport scalability (-X) 32K IGP & BGP routes
•
High Scalability (-R) 1 million routes
•
Enhanced Queuing (-EQ) 64k queues, Advanced QoS
MX-series Enhancements
New Services with Performance and Scale
Service Engine for the MX
•
MS-DPC is a dedicated hardware engine for stateful services
•
Service consistency with M-series MS-PIC
Enables Services, Maintains Performance
•
SBC Gateway
•
Dynamic Application Awareness
•
Intrusion Detection and Prevention
Non-Ethernet Interfaces for the MX
•
MX-FPC enables connectivity to traditional networks
•
Initially supports high speed OC-192 and OC48 PoS
Reduces Cost and Complexity
•
M/T PIC reuse saves money and simplifies sparing model
MS-DPC
MX-FPC
Adding Application Intelligence
MX Services DPC
Forwarding
Plane
Control
Plane
Service
Plane
PSDP – Services Toolkit
Application Identification
PSDP – Services Toolkit
3
rdParty
Apps
SFW
VQM
VoIP
BGF
IPSec
NAT
L2 Ethernet Scalability
512
N/A
STP Instances
4K
4K
Bridging Domains
Hardware is designed to support
higher scaling. Can Scale if tunnel
services on DPC used.
8K
2K
VPLS instances
Hardware is designed to support
higher scaling.
32K
16K
VLAN
Numbers assume 1M L3 routes on
Card. Can be greater if routes are
less.
1M
250K
MAC
Comments
Per System
Per Line Card
Feature
L3 Scalability
750k/750k
IPv6 RIB/FIB
1500 directed
2400 targeted
LDP Sessions
500
OSPF Sessions
500
IS-IS Sessions
4000
L3 VPN(VRFs)
4000
BGP Sessions
32K, 50K
LSP Head-Ends,
Transit LSPs
16M/1M
IPv4 RIB/FIB
Per System
Feature
Architecture Comparison
SRX
MX
Forwarding
Session based
Centralized forwarding performed by
SPU
Packet based
Distributed forwarding performed by DPCs
Services
SPU{1..N} acts as single service plane
with session based load balancing
Each SDPC is preconfigured to serve user
instances, i.e VRF, interface, user, etc.
SDPC{N} SDPC{N} SDPC{N} SDPC{N} SDPC{1} SDPC{1} SDPC{1} SDPC{1} DPC{1} DPC{1} DPC{1} DPC{1} DPC{N} DPC{N} DPC{N} DPC{N} SPU{N} SPU{N} SPU{N} SPU{N} SPU{1} SPU{1} SPU{1} SPU{1} IOC{1} IOC{1} IOC{1} IOC{1} IOC{N} IOC{N} IOC{N} IOC{N}
Every packet traverses SPU; Every packet traverses SPU; Every packet traverses SPU; Every packet traverses SPU;
Centralized forwarding, lookup based on session Centralized forwarding, lookup based on session Centralized forwarding, lookup based on session Centralized forwarding, lookup based on session
SPUs SPUs SPUs
SPUs act as a single service plane providingact as a single service plane providingact as a single service plane providingact as a single service plane providing session based load balancing
session based load balancing session based load balancing session based load balancing
Packets traverse SDPC only when service is required; Packets traverse SDPC only when service is required; Packets traverse SDPC only when service is required; Packets traverse SDPC only when service is required; Distributed forwarding, lookup based on packet
Distributed forwarding, lookup based on packet Distributed forwarding, lookup based on packet Distributed forwarding, lookup based on packet
Each SDPC is a single service plane configured Each SDPC is a single service plane configured Each SDPC is a single service plane configured Each SDPC is a single service plane configured for an user instance, i.e. VRF, interface
for an user instance, i.e. VRF, interface for an user instance, i.e. VRF, interface for an user instance, i.e. VRF, interface
SRX
SRX
SRX
SRX and MX Differentiation
SRX 5600/5800
MX 240/480/960
1
Product Positioning Dynamic Services Gateway Ethernet Services Router2
Market Integrated Security/FW Edge & Aggregation Router3
Target Deployments services, SP network securityEnterprise & SP data center, managedAggregation/core switching and routing,
WAN Gateway, Metro Ethernet & Edge
4
Architecture Session basedSession load balancing with central processingPacket based
Static (Hash based) load balancing with distributed processing
5
High Availability Inter chassis – Active/Passive Active/ActiveNSR/GRES
ISSU
6
Performance 36-120 Gbps Firewall/IDP/VPN, routing 240-960Gbps routing & switching 20Gbps Firewall per MS-DPC5Gbps IDP per MS-DPC
7
L2 Switching Not supported Supported8
L3 Routing IPv4 routingIPv6, Limited MPLS, multicast MPLSFull IPv4 & IPv6 routing, multicast,9
L4-7 Services Stateful Firewall, IDP, NAT, DoSIPSec VPNs NAT, DoS IPSec VPNs, SBC, IDP in 1Q09VRF/Application-aware Stateful Firewall,NOW NOW
Modular Ethernet Platforms Vital Statistics
FEATURE
EX 8208/8216
MX 240/480/960
TARGET DEPLOYMENTS LAN Aggregation & Core, Data Center Aggregation & Core WAN Gateway, Metro Ethernet Aggregation & Core, Ethernet Services Edge
GbE PORT DENSITY
EX8216: 768 ports (1536/rack) EX8208: 384 ports (1152/rack)
MX960: 480 ports (1440/rack) MX480: 240 ports (1440/rack) MX240: 120 ports (1080/rack)
10GbE PORT DENSITY
EX8216: 128 ports (256/rack) EX8208: 64 ports (192/rack)
MX960: 48 ports (144/rack) MX480: 24 ports (144/rack) MX240: 12 ports (108/rack)
L2 SCALE 160k MAC addr, 4k VLANs 1M MAC addr, 8k-16k VLANs, 64k Circuit IDs
L3 SCALE 400k IPv4 routes, 128 BGP peers 16M/1M IPv4 routes (RIB/FIB), 2k BGP peers
MULTICAST 4k multicast groups 256k multicast groups
MPLS flexible label push/pop, 6k VRFs
ACLs up to 64k entries up to 256k entries (list, range, & exception)
ENHANCED SERVICES
Future Application visibility HQoS for multiple customers / port or VLAN
VPLS hub for L2 circuit transport over MPLS
LIST PRICE $18K-$24K/10GbE port, $2K-$3K/GbE port
PRODUCT POSITIONING Ethernet Switch Ethernet Services Router
AVAILABILITY
$6K/10GbE port, $500/GbE port Future