Internet QoS : A Big Picture

Transcription

1 Internet QoS : A Big Picture Xipeng Xiao and Lionel M. Ni, M, Michigan State University IEEE Network, March/April 1999 Oct 25, 2006 Jaekyu Cho

2 Outline Introduction IntServ/RSVP DiffServ MPLS Traffic Engineering/CBR NGN QoS Conclusion 2/31

3 Introduction Current State of Internet Uses best-effort service model No guarantee of timeliness or delivery No service discrimination Bandwidth and network congestion problems Unpredictable network response time Today s Internet is plagued by sporadic poor performance This is getting worse, not better 3/31

4 Introduction What is QoS User point of view: Assurance of end-to-end service E.g. Guaranteed delay (VoIP), Guaranteed bandwidth (VPN) Relaxed definition: Service differentiation: different packets is treated differently end-to-end service guarantees may be achieved by provisioning e.g. only a small portion of high priority packets 4/31

5 Introduction QoS is not magic QoS will not alter the speed of light On an unloaded network, QoS mechanisms will not make the network any faster Indeed, it could make it slightly worse! QoS does not create nonexistent bandwidth Elevating the amount of resources available to one class of traffic decreases the amount available for other classes of traffic QoS cannot offer cures for a poorly performing network provided through relative prioritization of resource allocation to different flows/packets in the network 5/31

6 Introduction IETF has proposed many service models and mechanism to meet the demand for QoS Integrated Services/Resource Reservation Protocol (RSVP) model Differentiated Services (DS) model MultiProtocol Label Switching (MPLS) Traffic engineering Constrained Based Protocol 6/31

7 IntServ/RSVP Integrated services model is characterized by resource reservation For real-time service, before data are transmitted Apps must first set up paths and reserve resources Two service classes Guaranteed service:fixed delay bound Controlled-load service:reliable Four components signaling protocol admission control routine classifier packet scheduler 7/31

8 IntServ/RSVP sender 2) Determine bandwidth Receiver RESV RESV PATH DATA RESV PATH DATA PATH DATA 1) Traffic spec 3) Transmit Data 8/31

9 IntServ/RSVP The problems The architecture does not scale well in the Internet core Amount of state information increase Huge storage Process overhead The requirement on routers is high RSVP Admission control Classification Packet scheduling 9/31

10 DiffServ Differentiated Services (DS) are introduced because of the difficulty in implementing and deploying integrated services and RSVP DS is essentially a relative-priority scheme Per-flow service Per-aggregation service Process complexity core network edge network Network Boundary Traffic Conditioner DSCP field setting Traffic Policing DS Domain Per-hop Behavior Perform PHB process according to DSCP (Differentiated service codepoint) 10/31

11 DiffServ Internet Service Provider (ISP) decide which services to provide Service Level Agreement (SLA) specify the service classes supported the amount of traffic allowed in each class Static SLA Dynamic SLA must use a signaling protocol to request services. Many services can be provided. Premium service:low-delay and low-jitter Assured service:better reliability Olympic service:different quality 11/31

12 DiffServ Assured Service SLAs for assured service are usually static. It can be implemented as follows: Classification and policing are done at the ingress routers of the ISP networks. All packets, in and out, are put into an AQ. The queue is managed by a RED or RIO. RED drops packets randomly to prevent the queue fr om overflowing. RIO is a more advanced RED scheme, it drops out p ackets more aggressively. 12/31

13 DiffServ Premium Service The SLA specifies a peak bit-rate for a specific flow The customer is responsible for not exceeding the peak rate The ISP guarantees that the contracted bandwidth will be available to support both static and dynamic SLAs The exit routers of the customer domain may need to reshape the traffic At the provider side, the ingress routers will police the traffic Uneven distribution of traffic may cause a problem 13/31

14 DiffServ Service allocation Each host makes its own decision Bandwidth Broker (BB) makes decision Given the SLAs, ISPs must decide how to configure their boundary routers so that they know how to handle the incoming traffic. static SLAs, boundary routers can be manually configured. dynamic SLA, admission control decisions can be made by the boundary routers, or by a BB. 14/31

15 DiffServ Meter E.R Host B Classifier Marker Shaper DS Domain C.R E.R BB C.R C.R PHB E.R PHB PHB SLA Host A E.R BB SLA BB 15/31

16 MPLS MPLS: Multi Protocol Label Switching Incoming packets are assigned a label by edge router Packets are routed based only on the label through the network When a label switch router (LSR) receives a MPLS packet, it uses incoming label to find the next hop and the corresponding outgoing label In MPLS, the path a packet traverses is called label switched path (LSP) Network protocol independent 16/31

17 MPLS Label Short, fixed-length packet identifier Unstructured Link local significance Label (20-bits) EXP S TTL MPLS Header IP Packet 32-bits IP packet is encapsulated by ingress LSR IP packet is de-encapsulated by egress LSR 17/31

18 MPLS MPLS forwarding model Intf In Label In Dest Intf Out Label Out Intf In Label In Dest Intf Out Intf In Dest Intf Out Label Out Request: Mapping: Request: 47.1 Mapping: IP IP In tf In D e s t In tf O u t L a b e l O u t /31

19 MPLS MPLS is strategically significant Provides faster packet classification and forwarding Provides an efficient tunneling mechanism without encryption Moves processing to edges Core only forwarding, Scalable Priorities and QoS 19/31

20 Traffic Engineering/CBR The basic problem Given a network and traffic demands, how can traffic flows in the network be organized so that an optimization objective can be achieved Major goal To facilitate efficient and reliable network operation To optimize network resource utilization To efficiently manage bandwidth resources To optimize traffic performance To enhance QoS of traffic stream 20/31

21 Traffic Engineering/CBR Shortest Path problem C1 C3 C2 Path for C1 <> C3 Path for C2 <> C3 Longer paths may become under-utilized 21/31

22 Traffic Engineering/CBR Constrained Based Routing Each link a collection of attributes (performance, administrative) Constraint-based routing selects paths that satisfy one or more constraints Typical constraints include the available bandwidth along the path and administrative constraints. How different from IP routing CBR is source based whereas IP routing is distributed CBR needs explicit routing Support for distribution of link attributes 22/31

23 Traffic Engineering/CBR Constrained Based Routing C1 C3 C2 Path for C1 <> C3 Path for C2 <> C3 Longer paths may become utilized 23/31

24 Traffic Engineering/CBR Pros Meeting the needs for QoS requirement of flows Improved network utilization Cons Increased communication and computation overhead Increased routing table size Longer path may consume more resources Potential routing instability 24/31

25 NGN QoS 25/31

26 NGN QoS Service service VoIP Chatting Multimedia portal Audio/video Broad/multi casting Video on demand Dynamic file sharing characteristic 300Kbps ~ 1Mbps Bi-directional Video/Audio Stream Video clip, data, real time image High capacity, Ultrawide band Internet broadcasting service, IPTV Mpeg 2, Mpeg 4 Video Stream P2P, movie, music, entertainment 26/31

27 NGN QoS Requirement QoS VoIP FTP Broad casting ERP, VPN etc Bandwidth Low / Medium High / Medium High / Medium Low Loss rate sensitivity Medium Medium Medium High / Medium Delay sensitivity High Low High Low / Medium Jitter sensitivity High Low High Medium 27/31

28 NGN QoS Consideration (service) Service oriented QoS Quality based charge Usage End-to-end reliability Mobility Consideration (operation) Network Load balancing Resource optimization Minimization of communication Cost Aggregation/Segregation Reducing Complexity Guaranteeing Best Network Performance Delicate balance between good network design and engineering and QoS damage control 28/31

29 NGN QoS Open Problem How does the provider measure QoS? How does the customer measure QoS? How do you traffic, account, and bill for QoS? How will QoS work in a heterogeneous Internet? How to analyze the network level behaviors? Per-class, Per-service From/To Matrix QoS monitoring per class, per service Network Dimensioning Traffic characteristics Trial & Error approach based on Internet, PSTN, CDMA, mobile Internet, 29/31

30 Conclusion QoS is hotly debated issue Fibers and Wavelength division multiplexing (WDM) will make bandwidth so abundant and cheap QoS will be automatically delivered How much bandwidth the network provide, new application will be invented to consume Mechanism will be needed to provide QoS Many mechanisms are provide but they not solve QoS problem Quality of Service There is little hope for success But, Way to go 30/31

31 Q&A 31/31