Next Generation Networks Jornadas Técnicas Rediris Alcalá de

Скачать презентацию Next Generation Networks Jornadas Técnicas Rediris Alcalá de

d366fa138fc8a8039e5891fbd7bf0748.ppt

Количество слайдов: 68

Next Generation Networks Jornadas Técnicas Rediris Alcalá de Henares – Noviembre 2008 Javier Antich jantich@juniper. net Iberia SP SE Manager Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 1

Goals Understand which are todays challenges and context that determine the requirements for the next generation networks Describe the new technologies that will help addressing the challenges introduced. Highlight the growing relevance of Energy efficiency and IP&Optical transport convergence as techniques to reduce OPEX Present how Juniper is sensible with these challenges and how we can help addressing them. Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 2

Trend #1: The Grand Exodus • • Data/Apps are getting consolidated into a few Data centers People are getting scattered all over the world Data Center Work Force Globalization Branches & Campuses Network Data Center Consolidation RESULT AT DATA CENTER: Demand for • Massive performance/scale • Carrier-class reliability • Green designs • Virtualization of everything Copyright © 2008 Juniper Networks, Inc. RESULT AT BRANCH/CAMPUS: Demand for: • “All-in-one” integrated appliance • Remote deployment and management on a large scale Proprietary and Confidential www. juniper. net 5

Trend #2: The Blurring Work / Home • • People are taking work home People are bringing home-expectations to work Branches & Campuses Data Center RESULT AT BRANCHES & CAMPUSES: Demand for: • Securing corporate laptops even inside the “trusted” perimeter Network • Dual-mode WLAN or Enterprise Femto-cells RESULT AT END POINTS: Demand for: • A bewildering array of “unapproved” endpoint devices • Un-tethered mobility • Data Leakage Prevention Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential End Points www. juniper. net 6

Trend #3: The Blurring of Company / Cloud • Companies are putting their applications in the cloud (“Saa. S”) Data Center Branches & Campuses Network Content Service Provider RESULT AT CONTENT SP: Demand for • DPI for XML/SOAP • Heightened Qo. S and acceleration Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 7

Did you know this? § Assumptions • Single application = single connection § Underlying Functionality • Multiple connections are established to retried map segments • Segments are then pieced together to form a whole map § Infrastructure Requirements • Must support multiple connections at once Lack of NAT sessions • Network delays result in grey map areas until graphics are loaded Web. Pages § Reality • i. Tunes creates > 200 connections # of Sessions No Operation 5～ 10 Yahoo Top page/Google Map 10~20 200~250 i. Google 80~100 Youtube 50~80 Amazon Copyright © 2008 Juniper Networks, Inc. i. Tune ~80 Proprietary and Confidential www. juniper. net 8

IPv 4: The End of the Road Comes into View § Only 15% of IPv 4 space remains available § Depletion projected late 2010 Source: www. potaroo. net/tools/ipv 4/ www. tndh. net/~tony/ietf/ipv 4 -pool-combined-view. pdf Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 9

Three Trends in Networking 1. TDM is past its prime § § Built primarily for voice, and adapted reasonably successfully for leased lines, fine-grained TDM (PDH/SDH) is increasingly irrelevant for Next Generation Networks TDM is also very expensive on a cost/Gbps basis 2. Packet transport is on the rise § § There is recognition that transport must focus on packets, not bits There are multiple approaches, and a lot of confusion out there 3. Interest in the Packets+Photons Phenomenon is growing § § There is also recognition that the worlds of packets and of optical transport must come together Again, there are several approaches, and no clear way forward What Should Be Done? Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 10

Energy Savings § “The cost of power consumption by data centers doubled between 2000 and 2006, to $4. 5 billion, and could double again by 2011” according to the U. S. government. Bussiness. Week March 2008 Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 11

§ Joost § Zattoo § And many more… IP Video/Voice IP Data Traffic CAGR of 40% IP Video/Voice CAGR of 85% 2004 2005 Copyright © 2008 Juniper Networks, Inc. 2006 2007 2008 2009 Proprietary and Confidential www. juniper. net 13

Scalability and stability in large scale networks Absolut: Multi-chassis 25 Tbps System T 1600 #1 #16 1600 Gbps Switch Fabric Chassis 1600 Gbps #2 #3 #4 Copyright © 2008 Juniper Networks, Inc. #15 25. 6 Tbps Non-blocking #9 #14 #13 Proprietary and Confidential www. juniper. net 17

Scalability and Stability in Large Networks Control plane can become a bottleneck Shared Control Plane SVC 1 SVC 2 SVC 3 SVC n § Each service has diverse requirements (TE, QOS, security, growth rates) Control Plane Forwarding Plane § Requires multiple control planes Router Stability • Popular notion that convergence has happened is false. It only happened at the forwarding plane – not the control plane Processing Requirements § Since today’s equipment only supports one control plane, Service Providers are forced to roll out multiple subnets, or risk compromising scale, stability and/or security Scale § As more new services are introduced this leads to escalating Cap. Ex and Op. Ex SVC 1 SVC 2 Copyright © 2008 Juniper Networks, Inc. SVC 3 SVC n Proprietary and Confidential www. juniper. net 19

Scalability and Stability in Large Networks Control plane multiplicity § Control plane multiplicity changes that dynamic and fulfils the true promise of convergence • Shared infrastructure • Services are decoupled from network Independent Control Plane SVC 1 SVC 2 SVC 3 SVC n CP 1 CP 2 CP 3 CPn Juniper Control System • New services can be introduced without building a new subnet Forwarding Plane • Each services can be managed and controlled individually Router • Service introduction is swift and with reduced risk Scale Stability § Each service now runs on its own “Virtual Service Network” § Lower Cap. Ex, lower Op. Ex, Lower risk Processing Requirements SVC 1 Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential SVC 2 SVC 3 SVC n www. juniper. net 20

Virtualization Continuum Protected System Domain Shared hardware platform; Separate routing instances Shared hardware chassis; Dedicated routing resources P Logical Router Horizontal Consolidation PE Logical Router ü Isolates routing protocols & interfaces ü Enables hardware reuse – shared uplinks, efficient inter-LR forwarding ü Deployed for service separation, additional security, managed service, substitute for physical route Copyright © 2008 Juniper Networks, Inc. RE Pair Logical Routing PSD 1 PSD 2 Next Steps … Vertical Consolidation Delivered Safari ü Dedicates and isolates forwarding and control plane resources ü Run independent versions of JUNOS ü Share uplinks across virtual nodes ü No customer facing slots ü Flexibility and scalability of investment Proprietary and Confidential www. juniper. net 21

Scalability and Stability in Large Networks JCS 1200: A Radically New Architecture 2008 Juniper takes control plane architecture to the next level by physically decoupling the forwarding and control platforms T 1600 TX Matrix M 40 2004 1996 Juniper pioneers the separation of control and forwarding plane First multi-chassis routing system 2007 100 Gbps/slot Core IP/MPLS forwarding density LR_1 Service A LR_2 Service B 2003 Multiple control instances running on one router LR_3 Tier 2/3 ISPs RI_1: ISP A RI_2: ISP B Logical Routers Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 22

Example: Virtualized Routing System for Collapsed POP INTERNET NETWORK CORE PRIVATE PEERING Core Routers Internet Router Consolidated Router Peering Router Aggregation Router PRIVATE PEERING INTERNET Aggregation Router 20 -30% Cap. Ex Reduction Safari PSD 1: Core PSD 2: Aggregation PSD 3: Private Peering PSD 4: Route Reflection Edge Routers IP/MPLS CUSTOMERS Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 23

100 GE § Juniper is an active participant in the 100 GE standardization effort. § We are the only routing vendor to currently support 100 Gbps/slot of minimum packet sized Ethernet traffic and are working on support of 100 GE interfaces § Providing 100 GE in a timely fashion, commensurate with ratification of the technical details of the 100 GE standard, is a significant part of this effort within our product development team § Target delivery: 2010 Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 27

Nonstop Operation Nonstop Routing § Self-contained solution • No requirement for peers to support § No disruption of protocol adjacencies • Switchover is transparent to neighbors § Stateful replication of adjacency information on standby RE • Routing updates, hello messages, adjacency state, etc. § Dual active protocol sessions Primary Routing Engine Active Standby Routing Engine • Standby RE is fully active and can immediately take over sessions § Switchover is not dependent on stable topology • Topology changes can occur during switchover Continuous Systems Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 31

Daemon n Daemon 3 Daemon 2 Daemon 1 § What is our definition of ISSU? Routing Engine In-Service Software Upgrade (ISSU) Kernel Packet Forwarding JUNOS 9. 0 Physical Interfaces High-level Architecture View Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 32

Is this ISSU? Daemon n Daemon 3 Daemon 2 § NO: this is not true ISSU!! Daemon 1 § Upgrade of an individual module Routing Engine JUNOS 9. 2 Kernel Packet Forwarding JUNOS 9. 0 Physical Interfaces High-level Architecture View Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 33

Daemon n Daemon 3 Daemon 2 Daemon 1 § Upgrade of control plane software only § NO –this is not true ISSU! Routing Engine Is this ISSU? Kernel JUNOS 9. 2 Packet Forwarding JUNOS 9. 0 Physical Interfaces High-level Architecture View Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 34

Daemon n Daemon 3 Daemon 2 Daemon 1 § Upgrade within same major release § Example: 9. 0 R 1 to 9. 0 R 2 § Yes, this is possible with ISSU, but this is not always enough! Routing Engine Is this ISSU? Kernel Packet Forwarding 9. 0 R 1 JUNOS 9. 0 R 2 Physical Interfaces High-level Architecture View Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 35

§ …with minimal disruption to traffic § Can even go from one major release to another! JUNOS 9. 2 JUNOS 9. 0 Daemon n Daemon 3 Kernel Packet Forwarding Physical Interfaces Very comprehensive definition of ISSU! Copyright © 2008 Juniper Networks, Inc. Daemon 2 • Routing Engine • Packet Forwarding Engine • Physical Interfaces Daemon 1 § Our definition of ISSU: § Upgrade the entire code on the router… Routing Engine In-Service Software Upgrade (ISSU) Proprietary and Confidential High-level Architecture View www. juniper. net 36

Automated Operations Vision Advancing towards systems that proactively adapt to change and discover and mitigate problems • Error-resilient configuration, now with scripts to prevent procedural errors and to simplify common configurations • Confirmed adherence to business rules and policies • Auto-discovery and adaptation to network changes • Autonomic response to network conditions • Systematic implementation of diagnostics and repair to speed trouble response and resolution Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 37

JUNOScript Automation Commit Script • Enforce Configuration Rules • Automatic Configuration Generation Op Scripts • Build Custom Operational Commands • Build Powerful Troubleshooting Tools Event Scripts • Automate Diagnostics • Automate Change Detection Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 38

JUNOScript Automation Examples § Commit Script: [edit] admin@re 0 -ganimedes# commit [edit protocols ospf area 0. 0 interface fe-0/2/3. 0] 'interface fe-0/2/3. 0; ' warning: ATENCION: LDP no esta habilitado para este interface commit complete [edit] § Operational Script: admin@re 0 -ganimedes> op vecinos - OSPF: Hay 2 vecinos OSPF activos - ISIS: No hay vecinos ISIS activos - BGP: Hay 3 vecinos BGP activos - LDP: Hay 2 vecinos LDP activos - RSVP: No hay vecinos RSVP activos admin@re 0 -ganimedes> § Event Policy: [edit] admin@re 1 -leda# run file list detail /var/home/admin/: total 48 … -rw------- 1 admin field 209 Feb 23 12: 22 re 1 -leda_Event-LINK-UP-Script. txt_20080223_122233 1391 Feb 23 12: 22 re 1 -leda_Event-LINK-UP. txt_20080223_122231 [edit] Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 39

JUNOS™ Software – A Single-source Operating System One OS Routers One Release 8. 5 9. 0 9. 1 4 Q 07 1 Q 08 2 Q 08 One Architecture Module X Switches API Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 41

Why Care About Energy? 1. Electricity costs rose 88% in US since 2003 (US EIA data) Intl Energy Outlook ’ 07 predicts doubling energy generation by 2030, mostly via increasing the use of fossils Energy has become a non-trivial OPEX item 2. Worldwide legislation changes and public support for energy efficiency and climate control EMEA: reduce CO 2 by 20% by 2020 UK: reduce CO 2 by 20% by 2010 Japan: reduce CO 2 to 6% under 1990 level by 2010 § Carriers and businesses are setting new targets reduced energy consumption reduced heat dissipation reduced space requirements (volume footprint) Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 43

What Does This Mean for Data Networking? § Telecom facilities require power and cooling • Direct contributors to CO 2 emission • The cost of energy and space will rise § Data networking is still a growth industry • Global connectivity relies massively on routing and switching and this dependency increases • Significant increases in traffic are expected • This should NOT result in higher OPEX ► Vendors need to respond to the challenge Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 44

Energy-Efficient Routing Platforms – Basics § Energy efficiency must be built into design • Once the platform is designed and built, it is too late to speak of energy improvements § Consumed energy dissipates as heat • Heat is the major limit for building faster routers Building energy-efficient routers goes well along building the fastest routers § Energy savings must be verifiable • Absolute energy consumption makes little sense • Energy should be normalized to capacity Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 46

Energy-efficient router – Definition Energy-efficient router is the one that needs the least amount of energy (in joules) to transfer network data (in bits) Energy Consumption of Router (ECR) ECR = Σ C(i) T C is the power rating of a router’s component i Є I, I is the set of configured components T is the router’s effective capacity (full-duplex) ECR is normalized to Watts/10 Gbps Also we can use Energy Efficiency (EER), EER is expressed in Gigabits/KW Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential EER = 1 / ECR www. juniper. net 47

What can be done to improve energy metrics? § Today • Custom-designed silicon dies: No wasted blocks or gates • • • § Compare to commercial RISC CPU arrays (number of gates, clock) Compare to off-the-shelf NPUs (effective speed per feature set) Find fastest and simplest solution possible to do the job Use DRAM instead of power-hungry TCAM Shut elements when not in use (lookup cores, Ser. Des and memory) Tomorrow • • • Better integration, faster silicon and lower voltage Use of MCM (multi-chip modules) to unite several chips Possible use of CLI to monitor the real-time energy consumption Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 48

Energy Efficiency: Positive Impact § Energy efficiency is synergetic with higher speed • • § Efficient designs need fewer gates, allowing dense packaging Less energy means less heat dissipation, easier to scale up Promotes newer silicon fabrication technologies Promotes novel software and hardware structures Accelerated technology introduction • • • Promotes intensive scaling over extensive scaling (larger systems) Shortens effective silicon lifecycle in production networks Newer and better technologies deployed more frequently Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 49

Reference Data: Silicon in Progress Juniper M 40 M 160 T 640 T 1600 3. 0 10 40 100 System Capacity 40 Gbps 160 G 640 G 1600 G Technology 180 nm 130 nm 90 nm 1. 5 3. 15 6. 34 8. 21 50. 5 Gbps/KW 97. 5 Gbps/KW > 100 Gbps/KW 2002 2007 2010+ Slot Capacity, Gpbs Consumption, KW EER (Gbps/KW) FRS Copyright © 2008 Juniper Networks, Inc. 13 Gbps/KW 25 Gbps/KW 1998 2000 Proprietary and Confidential Next-gen 65 nm and < www. juniper. net 50

Juniper Experience: Technology Into Energy 1) Small power overhead for packet operations - Fully custom in-house packet processors - Unique ASIC expertise, very high gate utilization 2) Highest Integration Levels – fabrication and packaging - Full 10 Gbps datapath on a single IP 3 chip (includes lookup engine, memory controller and fabric interface) - Up to four forwarding engines on one blade (MX 960) - Industry’s only 100 G/slot core router in commercial use (T 1600) 3) Patented and Energy-Optimized Design - Stateless packet services without TCAM - Best-of-breed power converters Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 51

Synergy Between Vendors and Customers Efficient network design is extremely important § Data networking is very mature now § Many protocols and technologies were developed • some are obsolete But every time a network buildout is considered, it comes with a hefty list of features on RFP § Someone has to pay for all those features • Carved in silicon, unused gates and wasted power § Time to stop and think – • Which features are really needed and where? § Precise match of form and function is the best Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 52

Conclusions Environmental impact and energy efficiency are verifiable • Choose right platforms to satisfy energy requirements • Use normalized ECR/EER metrics for comparison • Design networks to minimize the energy and rack space usage We need to define and pursue aggressive energy goals • • Reduced energy consumption Reduced heat dissipation Reduced space requirements (compact footprint) This should be a joint effort between vendors, carriers and enterprises Energy efficiency stimulates the industry • New designs will increase EER and decrease environmental footprint • Fast networking and energy efficiency are not conflicting goals Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 53

Convergence from many different perspectives § Services Convergence. § Technology convergence. • Optical and IP § Topology / planes convergence. • POP consolidation. • Reduction of network layers. § Networks convergence • Fixed and mobile. Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 55

Transport Intelligence Substituting key SONET/SDH functions with G. 709 and GMPLS § Easy operations (OAM&P) • G. 709 overheads mimic SONET/SDH functions • GMPLS allows optical layer visibility into hard to detect failures • Integrated optics low-cost optical monitoring and provisioning § Fast protection • Integrated DWDM interfaces of a router enable fast triggers • Router-based fast reroute (FRR) may be more economical and as fast and reliable as SONET/SDH ring-based protection Sub-wavelength grooming • Not needed—router trunks can fill 10 G/40 G wavelengths • Manage bandwidth at the wavelength level using ROADMs Replacing SONET/SDH functions by MPLS + G. 709 + DWDM allows for a simpler, more scalable architecture Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 57

Transport Intelligence Optical integration Next steps Available today Fixed and Tunable Optics OTN Interfaces GMPLS interoperability 40 Gbps IPo. DWDM 10 GE Tunable Optics WAN PHY Single Transport-Service Control Plane Container Interfaces GMPLS Ethernet OAM § Simplifies core topologies § Offers flexibility in provisioning and response to topology changes § Enables on-demand services § Connecting Metro E over SONET Copyright © 2008 Juniper Networks, Inc. G. 709 OSS 1: 1 or 1: n protection GMPLS § Elegant, multilayer failover scenarios § End-to-end performance monitoring § Coordinates end-to-end restoration across optic and routing layers § Reduces bandwidth and interface requirements for redundancy Proprietary and Confidential www. juniper. net 58

Transport intelligence Cap. Ex and Op. Ex performance After Before Benefits Lower Cap. Ex § 66% optics reduction Lower Op. Ex § Fewer shelves (space, cooling, power, management), § Fewer interconnects Enhanced resiliency Router Transponder Mux/ROADM Copyright © 2008 Juniper Networks, Inc. Router Mux/ROADM Proprietary and Confidential § Fewer devices § Fewer active components § Fewer interconnects www. juniper. net 59

Transport Intelligence Management Options OSS Mgmt. Plane § Single intelligent IP control plane for delivering service flexibility and lower Op. Ex Router Management Transmission Management § Segmented or integrated management model for faster provisioning, reduced Op. Ex E. g. , SNMP Control Plane Juniper GMPLS CLI NETCONF JUNOScript WDM GMPLS End to end service view provided by transmission MGMT or other common OSS OTN Data Plane Copyright © 2008 Juniper Networks, Inc. Integrated Control Plane based provisioning Proprietary and Confidential § Integrated transponders lower Cap. Ex/Op. Ex, increase reliability § ROADMs eliminate OEO and minimize truck rolls for reliability, service flexibility, and lower Op. Ex www. juniper. net 60

New services, new ideas § The network must be open to the integration of new services, new capabilities. § Equipment vendors should no longer be the only source for the innovation. § Example: • Juniper PSPD. Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 64

Juniper Vision: An Ecosystem of Choice § For Customers Idea Developer User Existing Supply Models § For Partners • Reducing barriers to partnership • Integrate new technologies more quickly Copyright © 2008 Juniper Networks, Inc. Juniper Customer Customer Independent Vendor Customer ? • An ecosystem of choice • Build competitive differentiators internally • Bring new technologies to customers Juniper ? Customer New Supply Models Proprietary and Confidential www. juniper. net 65

DNA of the Next Generation Networks IP & Optical convergence GMPLS NSR, ISSU IPv 6 100 Gb. E Multichasis systems Open Networks for innovation. Energy Efficiency Operational Automation Flexible control plane virtualization Copyright © 2008 Juniper Networks, Inc. Proprietary and Confidential www. juniper. net 67