Lic.(Tech.) Marko Luoma (1/23) Lic.(Tech.) Marko Luoma (2/23) Today’s Topic � This lecture is about control plane of Service(s) & Internet routers and especially Customers Service Level Agreement S−38.180 Palvelunlaatu Internetissä routing in Control plane [SLA] S−38.180 Quality of Service in Internet Service Architecture Service Level Resource Specification Reservation [SLS] Luento 5: QoS reititys Signaling Conditioning Actions Relay actions Lecture 3: QoS routing Routing Policy Control Policy Information Management Information Base [PIB] Base [MIB] Admission Control Network Device(s) Input Processors Output Processors Lic.(Tech.) Marko Luoma (3/23) Lic.(Tech.) Marko Luoma (4/23) Conventional IP routing Conventional IP routing Routing is a matter of finding a path (usually shortest possible) between any � � Nature of conventional shortest path algorithms cause traffic to be aggregated two networks in the whole Internet to lowest cost links � Finding a path means that mess of networks is organised in to tree like � Centralises traffic into hot spots in the network structure representing neccesary links to reach all possible networks from � Large amount of links are left to idle while few are overloaded the point of interest
Lic.(Tech.) Marko Luoma (5/23) Lic.(Tech.) Marko Luoma (6/23) Conventional IP routing Interior Gateway Protocols � Construction of routing tables is responsibility of routing protocols Possibility to full knowledge of domain characterstics � � Routing protocols can be divided based on their usage (scalability): � Capacities � Interior Gateway Protocols: Running inside one autonomous system � Delays � OSPF, IS−IS, RIP, IGRP ... � Offered traffic � Exterior Gateway Protocols: Running between autonomous systems � Preferences � BGP, IDPR Routing normally based on the shortest path � � Routing protocols implement neccesary optimization algorithms to find shortest � Least amount of hops between two end points paths between end points: � Distance vector (RIP, IGRP, BGP) OPTIMIZATION � Link−state (OSPF, IS−IS) Lic.(Tech.) Marko Luoma (7/23) Lic.(Tech.) Marko Luoma (8/23) Exterior Gateway Protocol OSPF Domain characteristics relatively unknown � � Operation goes through four phases: � Knowledge is based on agreements and policies � One: Neighbours are aquired and maintained in adjacency by hello packets � Real−time data is rarely distributed � Adress and cost information is gathered � Reachability information (distance vector features) � Heartbeat of particular link (failure detection) � Support for QoS ??? � Two: Link−state advertisement (LSA) packets are formed based on information gathered by hello packets � Three: LSA packets are flooded into the network and received from the network to construct topology database � Four: Least cost routes are calculated to every other router in the network POLICIES
Lic.(Tech.) Marko Luoma (9/23) Lic.(Tech.) Marko Luoma (10/23) OSPF OSPF � Link−state advertisement packet contains: Topology database is initially copied from one of the adjacent neighbours � LS Age Options LS Type Link State ID � Header part identifying � Updates to initial database are received and sent by flooding Advertising Router � Advertising router LS Sequence Number � Every adjacent neighbour receives flooded LSAs and process them to LS Checksum Length topology database. � LSA type Router Type 0 # Links Link ID � Certain LSA types may have � After processing LSAs are repacked and flooded ahead Link Data additional header information � Every router in the net receives a copy of original LSA Link Type # ToS Metric Link ID � LSA information part (depending on LSA type) � ’Full’ knowledge of network devices and links Link Data � Link information and metrics Link Type # ToS Metric Calculation of routes is based on Dijkstra algorithm and information in � . . . � Network information and attached topology database Link ID routers Link Data Link Type # ToS Metric Lic.(Tech.) Marko Luoma (11/23) Lic.(Tech.) Marko Luoma (12/23) OSPF Routing in general � Metric used in route computation is based on information received in LSAs Optimize Constrain � � � It set by � Find best possible solution to the � Find possible solution to the problem in hand problem in hand � Network administrator to indicate preference of particular link � Minimum cost � Delay less than X � Automatically as a form of computational intelligence in a router � Shortest path � Free capacity larger than Y 10.1.1.1 (1,1) � Maximum bandwidth � Usually suboptimal 10.1.1.6 10.1.1.4 (1,3) (4,1) (5,1) (1,5) � Optimal � Many possible options (2,1) (3,3) 10.1.1.2 (4,5) (2,5) 10.1.1.8 � One solution � Limited search (3,3) (1,5) (1,1) (2,1) � Full depth search (2,5) (1,1) (1,3) (5,1) (2,1) (2,3) (4,5) (4,1) 10.1.1.3 (3,1) (3,1) 10.1.1.5 10.1.1.7 (1,1)
Lic.(Tech.) Marko Luoma (13/23) Lic.(Tech.) Marko Luoma (14/23) Routing QoS Routing problems Unicast QoS routing � Conventional IP routing is based on � Integrated Services is based on Basic routing problems Composite routing problems Link optimization routing Link constrained link optimization routing connectionless network philosophy connection oriented network (bandwidth optimization) (bandwidth constrained − buffer optimization) philosophy � Each packet is independent and Link constrained routing Link constrained path optimization routing complete unit � Path is coupled into the packet (bandwidth constrained) (bandwidth constrained − least delay) Link constrains: streams through state � Routing is decoupled from the Multilink constrained routing � Capacity information in the routers (bandwidth and buffer constrained) packet streams � Buffer space Link constrained path constrained routing � Multivariable constraint problem � Pure optimization problem Path constrains: (bandwidth and delay constrained) � Multiprotocol label switching is Differentiated Services is based on � Delay � Path optimization routing Path constrained link optimization routing based on connection oriented � Cost connectionless network philosophy (least cost) (delay constrained bandwidth optimization) philosophy � Routing is decoupled from the Path constrained routing Path constrained path optimization routing � Path is coupled into packet (delay constrained) (delay constrained least cost) packet streams streams through state Multipath constrained routing NP � Multi variable constraint and (delay and delay jitter constrained) � Multivariable constraint problem optimization problem Lic.(Tech.) Marko Luoma (15/23) Lic.(Tech.) Marko Luoma (16/23) Routing Strategies Routing Strategies Source routing: Distributed routing: � � Hierachical routing: � Biggest problems: � � Centralized routing decission � Path computation is distributed � Aggregation causes imprecision � State base is shrinked with to all routers between source and which causes paths to be only clustering and aggregation � Source computes route semi−optimal destination (distance vectors) through the network � Network is partitioned to � Biggest problems: � Formation of aggregate metrics clusters reflecting areas of � Biggest problems: common policy is not straight forward � State change in the network � Knowledge of the global may cause loops which can � State of the clusters is state is only approximate not be easily solved aggregated at the boundaries (communication delay) � Construction of distributed � Approximates distributed source � Size of the state base is huge heuristics for multiple routing (all links and nodes and their attributes is not straight attributes) � Each cluster is individually forward source routed
Lic.(Tech.) Marko Luoma (17/23) Lic.(Tech.) Marko Luoma (18/23) Problems with multiple metrics Pruning � Metric aggregation: � Path selection: � Metric 1: Free capacity greater than � Metric 2: Delay less than Y ms X bps � E−>G is correct as largest � Link which do not qualify by bandwidth path is equal to link constraint should be pruned lowest delay path before optimization with path constraint � E−>F is incorrect as bandwidth and delay paths are not same D E F D.2 D.1 (3,1) F (3,1) (1,1) Links which do not have resources to fulfill constraints of the (3,2) metric are removed (pruned) from the graph E D (3,1) (3,1) (3,1) G G (3,1) (3,3) (3,1) D.3 D.4 Lic.(Tech.) Marko Luoma (19/23) Lic.(Tech.) Marko Luoma (20/23) Constraint based routing What is the difference � Optimization is used to find path � Optimization can be done straight � Pruning constraint 1: Capacity � Pruning constraint 1: Capacity from the reduced topology after pruning of first metric � Pruning constraint 2: Delay � Optimization with delay � Cost � Lowest delay path is searched Optimization with <delay> Sanity check � � � Delay Requires check whether delay � � Delay less than constraint 2 constraint is held Delay is path constraint which has very little meaning on link by link basis. Therefore it has to be broken down to link constraints. Easily NP complete problem...
Recommend
More recommend
