CATT Seminar on Networks ResearchPolytechnic UniversityMarch 27, .ppt

《CATT Seminar on Networks ResearchPolytechnic UniversityMarch 27, .ppt》由会员分享，可在线阅读，更多相关《CATT Seminar on Networks ResearchPolytechnic UniversityMarch 27, .ppt（38页珍藏版）》请在麦多课文档分享上搜索。

1、CATT Seminar on Networks Research Polytechnic University March 27, 1999 Next Generation Networks Richard D. Gitlin Chief Technical Officer and Data Networking Technology Vice President Data Networking Systems Lucent Technologies ,Next Generation Networks,Introduction The Network Revolution Technolog

2、y Trends Applications and Requirements Issues and Solutions Quality of Service Security Network Management High Reliability Intelligent Networking Example: Voice on the Next Generation Network Summary,This R/Evolution Is Fueled By Unparalleled Customer Demand (and by telecom deregulation and the Int

3、ernet),It took about a century to install the worlds first 700 million phone lines; an additional 700 million lines will be deployed over the next 15-20 years There are more than 200 million wireless subscribers in the world today; an additional 700 million more will be added over the next 15-20 yea

4、rs There are more than 200 million Cable TV subscribers in the world today; an additional 300 million more will be added over the next 15-20 years More than 100 million additional Internet users will come on-line by 2001 -the Net is experiencing a 1000% per year growth! If this trend continues, by 2

5、004 99% of the worlds bandwidth will be Net traffic -including computer-to-computer communications.,Global Internet Users,1994 1998 2001,Average Hold Times,Internet Session 20 - 30 minutes,Changing Traffic Patterns,Voice Call 3 minutes,Worldwide Access Lines,3B,2B,1B,Next Generation Networks (The Ne

6、w Public Network): Current situation,No longer any debate that wide-area networks based on packet technology will emerge as a compelling alternative to the PSTN The new public network will be optimized for IP-based applications and will become the platform for future voice and data service innovatio

7、ns-it will not be based on merging existing legacy voice and data frame relay, SMDS, IP, networks Carriers expect that the simpler new network will also reduce costs of operations, equipment and staff and will capitalize on the faster pace of networking element development Migration strategies, qual

8、ity of service (QoS), network management, security, rapid service creation, and reliability are the major concerns of the carrier -as well as the almost $1 Trillion invested in the PSTN Almost 80% of the service providers intend to build their multiservice network with an ATM core and about 20% base

9、d on IP Some principles for the new network Give customers access choices (DSL, cable, wireless, ISDN, ) Work hard to optimize IP switching (DiffServ, MPLS, RSVP, .) Separate service intelligence from the network transport -open interface between intelligent call control features and packet gear Bui

10、ld IP-based billing and management,A Networking Paradigm Shift Occurring(IP Becomes Dominant WAN and LAN Protocol),Next-generation data networking Excellent performance with IP QoS breakthroughs: wire speed and per flow control “Route once, switch often” Route at wire speed Distance transparency and

11、 distributed “computing” Policy driven network management Directory Enabled Broadband access Wireless and optical networking Silicon and software Data on voice (circuits) Voice on data (circuits) “80/20” Enterprise/WAN data traffic split “20/80” Networks Network of networks,More than moving voice ov

12、er the Internet Converged, multi-service networks reduce costs provide integrated services Voice over cell/packet solutions - VoATM and VoIP Virtual Private Networks - VPNs Quality of Service - QoS Accommodate multiple protocols (e.g., IP, ATM, frame relay) Provide at least todays voice services (e.

13、g., 3-way connections, hold, add, forward, toll free, 911) Interoperate with one another, the Internet and the Public Switched Telephone Network,“Convergence” Driving Change & QoS,The real challenge is to build converged networks that are as reliable, robust and scalable as voice networks,Convergenc

14、e of Communications Paradigms Leads to New Services and Requires New Technologies,Telecommunications,ConnectionsTightly CoupledCentralized ControlsHW Fault ToleranceFeatures At Call Set-UpObsession With QoSLow Latency,Data Communications,ConnectionlessLoosely Coupled Loose Controls,DistributedSW Fau

15、lt ToleranceFeatures During SessionLittle Attention To QoSHigh Latency,Applications,The Pace of Technology,Technology Trend,Silicon Chips X2 in density/speed every 18-24 months Optics X2 in transmission capacity every year Data/Web X2 Internet subscribers every 2-3 years X2 Internet hosts/servers ev

16、ery year Wireless X1000 in capacity in 5 years Power X2 MIPs/MW every 2 years (DSPs) Compression X2 in information density every 5 years,Disruptive Technologies and their Impact on Networking,Impact of Transmission Speeds on Networking,Available WAN bandwidth has been less than LAN bandwidth - this

17、situation is expected to change at the millennium (WANs no longer a bottleneck for leading edge customers) Fiber optic transmission speeds have increased by 50% per year since 1980 (x100 in 10 years) LAN bandwidth has increased at 25% per year and WAN bandwidth has remained expensive (shared) “Avail

18、able” curve purchased by leading-edge users (e.g., OC-3c); about 1% of WAN BW,1975,1980,1985,1990,1995,2000,LAN,Single Channel Fiber,Multi-Channel (WDM),Available,T1,T3,OC-3c,Ethernet,Mbps,102,10,103,104,105,Fast Ethernet,Gigabit Ethernet,Impact of Speeds of Fiber Transmission and Microprocessors on

19、 Networking,Speed gains for microprocessors have kept pace with fiber transmission speeds The number of instructions available to process an optically transported packet, using the “hottest” micro has remained constant,1975,1980,1985,1990,1995,2000,Microprocessor speed (Mhz),Single Channel Fiber,Mul

20、ti-Channel (WDM),Merced,PowerPC,486,386,286,Mbps or Mhz,102,10,103,104,105,Single Channel Fiber,Impact of DRAM Memory Size and Transmission Speeds on Networking,With increasing transmission speeds, more packets are “in flight” for a given round trip propagation time; common error recovery protocols

21、require that one round trip worth of data be stored e.g., NY-LA-NY round trip propagation time of 50 ms results in 1 MB for a 155 Mbps link Size of DRAM increasing 58% per year Effective BW of memory is increasing at about 40% Storage capacity and transmission speeds are increasing at the same rate,

22、 thus number of chips to hold one “window” of data has remained constant,1975,1980,1985,1990,1995,2000,DRAM Size,Multi-Channel (WDM),256 MB,64 MB,16 MB,4 MB,Much More Traffic (leads to much more traffic - Metcalfes Law),Metcalfes Law: the value of a network grows exponentially with the number of use

23、rs and connected sources and a “network of networks” becomes the organizing principle for most communications,US Businesses WAN Peak Capacity Will Need to Increase at Least 10X in Three Years,0.0,1.0,2.0,3.0,4.0,5.0,1997,1998,1999,2000,Tb/sec,Source: Estimated from projections of data port shipments

24、 (Dataquest, 12/97),56 Billion Year 2000,3 Year Growth of Email Messages,Source: email projections: Yankee Group,3.5 Billion 1997,Major Requirements for Next Generation Network Applications,The Leading Protocols for Transporting Information on Next Generation Networks Are ATM and IP,*Related Items,E

25、conomies of scale favor IP,Issues to Be Solved for Next Generation Networks: QoS,Application & Source,Performance Issues,(e.g., Latency,Jitter),Reduction of large frequently,encountered latency and,response time,Efficiency of network traffic,Caching,Network and Server Load Balancing,Efficient,Multic

26、asting,Mirroring,Firewall/Proxy Server Farms,Private Peering Agreements,How Will IP Networks Approach the Performance of ATM Networks?,Implementing wire speed switches Decreasing effect of IP packet variability and header size with transmission of higher speeds Selecting good designs and paths with

27、VPN Designer expert system Making IP connection oriented via MPLS, per flow queueing Implementing QoS infrastructure akin to PNNI Using policies and directories to enable QoS Exploiting ASICS for congestion control directly on flows Executing congestion control within core instead of at edge,Next Ge

28、neration Switches,Wire speed traffic classification and filtering No performance degradation when filtering or QoS is switched on Complete traffic isolation: Can meet Service Level Agreements without the need for over-provisioning Guaranteed minimum bandwidth based on source address, destination add

29、ress, protocol and/or TCP/UDP port numbers Hierarchical Weighted Fair Queuing,Situation Large IP packets cause longer delays than short ATM packets Variable IP packets create more jitter than fixed ATM packets 20 Byte IP header causes less economic efficiency than 5 Byte ATM header(Voice over ATM),D

30、ecreasing Effect of IP Packet Variability and Header Size (Example Application: Voice over ATM vs. Voice over IP),Natural Solution IP Performance and Economics Comparable at Speeds beyond OC-12,Allows QoS path optimization SLAs are easy to implement. Facilitates identifying individual flows Can be u

31、sed with IP, ATM, SONET, WDM, . Supports multi-vendor environments Complements Enterprise need for tunnelsWill require building QoS capabilities intoOSPF, LDP, RSVP protocols,Make IP Connection Oriented via MPLS.,IP With MPLS and IP Over ATM For IP QoS Guarantees,*ERLSP=Explicitly Routed Label Switc

32、hed Path,Both at same priority with no discrimination,Both at same priority with routers using flow isolation ( by VPN) and equal weights for the two VPNs,Benefit of Isolating Flows,Price of Not Isolating Flows,Congestion Control of Bad Behavers: Value of Isolating Flows in QoS Management,VPN1 And V

33、PN2 Have The Same Contract ( 0.4 of the DS1 capacity)VPN2 uses 0.52 of the capacity (i.e., 30% more than contract),Without flow isolation, all VPNs get unacceptable delay when one creates congestion With flow isolation, all well behaving VPNs get acceptable delay With flow isolation, misbehaving VPN

34、s can get acceptable delay only when other VPNswell below contracted load,Reducing Latency: Web Access With Next Generation Caching,,Client,,,L4 Request Trap,http,Load Balance Requests,Router,PULL,Multicast,Request,Reply,Request,Solution Principle:Move content closer to users much lower web access l

35、atency reduced network congestion higher content availabilityNext Steps pre-fetch “hot” objects multicast to cache sites load balance cache sites high level trap of cache request support “streaming” multimedia cache dynamic content support value-added services,Central Cache Control,Cache Sites,Curre

36、nt Situation High End-to-end latency High Network load High Server load High Cost for ISP and Enterprise,Deploy cache sites in:- NAP- Backbone network- Data center- ISP- POP- Enterprise,Reducing Latency With Multicasting,Current Situation Redundant traffic causing needless loading of network and ser

37、vers Results in unacceptable latency,Issues to be solved for Next Generation Networks: Security,Requirements for Access to VPNs,VPN Requirements Private Addressing: to allow access to corporate network resources (Tunneling and Network Address Translation) Security: authentication of users and privac

38、y of user data as it goes over the network (RADIUS/DIAMETER, Tunneling) Legacy Protocols: allow user to use non-IP protocols (e.g. IPX, AppleTalk) over an IP network (Tunneling) Performance: provide a level of performance comparable to that of private networks (QoS) Network Management: provide custo

39、mer management of the VPN (monitoring, reconfiguration,) Issue: Tunneling addresses many VPN requirements but makes QoS more difficult since flow information becomes hidden in the core,Evolving Tunneling Options,Issues to Be Solved for Next Generation Networks: Network Management,Current paradigm ha

40、s following problems: Individual Device management Device Manager per vendor Device Manager per product No unified configuration store Network Manager and Devicehave Client-Server model andare not peers,Historical Network Management/Policy Paradigm,Device Manager (NMS),Data store,Network Device,Agen

41、t,NVRAM,Network Device,Agent,NVRAM,SNMP,Device Manager (2),Data store,Network Device,Agent,NVRAM,Evolving to Next Generation Network Management,Current Situation Independent device and independent services management Table-driven device functions Client(NM)-Server(Device)architecture SNMP,The Future

42、 Distributed policy management Integrated services through policies Reactive agents added Complex & reactive policy capabilities,Near Term Directories drive data unification Central policy management on service basis Dynamic device functions Policy agents added,Complex Networks and New Dynamic Servi

43、ces Drive Changes to Policy Management and Infrastructure,Meta Directory Solution All directory changes are arbitrated through the Meta-DirectoryMeta-Directory maintains consistency between information in each physical directory/databaseAppearance of a single directory to Network Manager Single entr

44、y link to other directoriesMeta-Directory Is A Band-aid Does not resolve any overlapping schema issues,Meta-Directory,Directory Evolution: Near Future,Network Device,LDAP,Directory,Data store,Directory,Data store,Directory,Data store,Address Policy Server,QoS Policy Server,Directory Management Inter

45、face,Solution Policy scriptsDistributed by Policy ServerInterpreted by Network DevicesAlternative to COPS/DIAMETERNetwork Device uses Directory for configurationPolicy Server uses Directory for decision support and policy storagePolicy Server and Directory Access Client both manipulate device data s

46、tructures,Network Management (The Future): Supporting Complex and Reactive Policies,Policy Server,Directory Access Client,Policy Manager (PIP),Directory,Decision Support Info,Distribution,LDAP,Management & Decision Support,Configuration Activities,Policies Are Represented as Scripts,Example Voice ov

47、er IP Application: What is Required to Support VoIP With QoS?,Voice over IP (VoIP) Architecture Requirements,Todays products do not scale well. Need to separate signaling from media transport and control for large scalable networks Media Gateways 1000s Media Gateway Controllers/Gate Keepers 10s, Sig

48、naling Gateways 10Todays solutions do not interface with value added feature data bases or Signaling Control Points (SCPs). Voice feature support requires interaction with existing and future SCPs such as Local Number Portability (LNP), 800, SDN, .VoIP is growing much faster than multimedia over IP. Thus, focus on voice protocol simplification first.Commercial Success of VoIP (including VPNs) will require QoS Call Admission Media Transport,