of 37
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

lecture 05: centralized control opportunities and challenges

Category:

Leadership & Management

Publish on:

Views: 3 | Pages: 37

Extension: PDF | Download: 0

Share
Description
lecture 05: centralized control opportunities and challenges 5590: software defined networking anduo wang, Temple University TTLMAN 402, R 17:30-20:00 some materials in this slide are based on lectures
Transcript
lecture 05: centralized control opportunities and challenges 5590: software defined networking anduo wang, Temple University TTLMAN 402, R 17:30-20:00 some materials in this slide are based on lectures by Jennifer Rexford RCP 3 BGP background BGP destination -de-facto inter-domain (inter-as) routing protocol ebgp session AS A AS B functionality partitioned across routing protocols W ibgp session 4 X 1 V 1 2 Y IGP link -ebgp -ibgp -IGP Z 4 BGP background BGP route-selection destination 1. highest local preference 2. lowest AS path length AS A AS B 3. lowest origin type ebgp session 4. lowest MED (with next W X Y hop) 4 1 V 2 IGP link 5. ebgp-learned over ibgplearned ibgp session 1 6. lowest path cost to egress Z 7. lower router ID 5 BGP: shortest path routing BGP route-selection destination 1. highest local preference 2. lowest AS path length AS A AS B 3. lowest origin type ebgp session 4. lowest MED (with next W X Y hop) 4 1 V 2 IGP link 5. ebgp-learned over ibgplearned ibgp session 1 6. lowest path cost to egress Z 7. lower router ID 6 BGP problem: oscillation BGP route-selection destination 1. highest local preference ebgp session AS A W AS B A prefers B, B prefers A X Y 2. lowest AS path length 3. lowest origin type 4. lowest MED (with next hop) 4 1 V 2 IGP link 5. ebgp-learned over ibgplearned ibgp session 1 Z 6. lowest path cost to egress 7. lower router ID 7 BGP problem: hot-potato BGP route-selection destination 1. highest local preference ebgp session AS A W X AS B Y 2. lowest AS path length 3. lowest origin type 4. lowest MED (with next hop) V 2 IGP link 5. ebgp-learned over ibgplearned ibgp session 1 Z 6. lowest path cost to egress (hot-potato, early-exit) 7. lower router ID 8 BGP problem: RR full-mesh ebgp Physical Peering ibgp 9 BGP problem: RR full-mesh ebgp Physical Peering ibgp A 1 2 B selected by router reflector (RR) RR 1 1 selected by full mesh C 9 BGP problems BGP is broken -converge slowly, sometimes not at all -routing loops -misconfigured frequently -traffic engineering is hard fixing BGP is hard -incremental fixes: even more complex -deployment of new inter-domain protocol almost impossible 10 solution: RCP RCP Inter-AS Protocol RCP ibgp RCP Physical Peering AS 2 AS 1 AS 3 use centralized controller to customize control -controller computes routes on behalf of routers -uses existing routing protocol for control traffic 11 Getting From Here to There ebgp% ibgp% Two issues RCP% AS%1% Backward compatibility Inter9AS%Protocol% ibgp% Physical% peering% RCP% AS%2% 3 phases to achieve -backward compatibility, deployment incentives 12 phase 1: control protocol interactions Phase 1: Control Protocol Interac Before:(convenXonal(iBGP( ebgp% ibgp% AKer:(RCP(gets( best ibgp(routes((and(igp(topology)( ebgp% RCP% ibgp% only one AS has to change 13 phase 2: AS-wide policy Before:(RCP(gets( best ibgp(routes((and(igp(topology)( ebgp% RCP% ibgp% AKer:(RCP(gets(all(eBGP(routes(from(neighbors( ebgp% RCP% ibgp% 14 phase 3: AS-wide policy Before:(RCP(gets( best ibgp(routes((and(igp(topology)( ebgp% RCP% ibgp% AKer:(RCP(gets(all(eBGP(routes(from(neighbors( ebgp% RCP% ibgp% 15 phase 3: all ASes have RCP Before:(RCP(gets(all(eBGP(routes(from(neighbor ebgp% RCP% ibgp% AKer:(ASes(exchange(routes(via(RCP( RCP% Inter9AS%Protocol% ibgp% RCP% AS%1% Physical% peering% AS%2% 16 RCP architecture P1, P2 Routing Control Platform (RCP) -IGP partitions Route Control Server (RCS) BGP Engine IGP Viewer s 17 RCP architecture Routing Control Platform (RCP) Route Control Server (RCS) BGP Engine IGP Viewer s 18 RCP architecture IGP viewer Routing Control Platform (RCP) Route Control Server (RCS) -maintains IGP topology -computes pairwise shortest paths with AS BGP Engine IGP Viewer s 18 RCP architecture IGP viewer Routing Control Platform (RCP) Route Control Server (RCS) -maintains IGP topology -computes pairwise shortest paths with AS benefit: scalability BGP Engine IGP Viewer -cluster routers -reduce # independent route computation s 18 RCP architecture Routing Control Platform (RCP) Route Control Server (RCS) BGP Engine IGP Viewer s 19 RCP architecture BGP engine Routing Control Platform (RCP) -communicates RCS decision to routers via ibgp Route Control Server (RCS) BGP Engine IGP Viewer s 19 RCP architecture BGP engine Routing Control Platform (RCP) Route Control Server (RCS) -communicates RCS decision to routers via ibgp benefit -backward-compatibility BGP Engine IGP Viewer s 19 RCP architecture Routing Control Platform (RCP) Route Control Server (RCS) BGP Engine IGP Viewer s 20 RCP architecture RCS Routing Control Platform (RCP) Route Control Server (RCS) BGP Engine IGP Viewer -computes BGP route assignments -obtain topology from IGP -disseminate decision via BGP engine s 20 scalability, efficiency, and reliability ebgp% RCP% ibgp% requirements -many routers ( ) -many destination prefixes (150, ,000) -converge quickly 21 reliability RCP! RCP! 22 reliability RCP! replicate RCP -multiple identical servers RCP! 22 reliability RCP! RCP! replicate RCP -multiple identical servers independent replicas -each receives same information, running the same routing algorithm -NO need for a consistency protocol if both replicas always see the same information 22 single RCP under partition RCP! ParXXon(1( ParXXon(2( only use state from routers partition to assign BGP route -ensure next-hop is reachable 23 multiple RCPs under partition RCP! RCP! ParXXon(1( ParXXon(2( ParXXon(3( 24 multiple RCPs under partition RCP! RCP! ParXXon(1( ParXXon(2( ParXXon(3( RCPs receive same state from each reachable partition -IGP offers complete visibility -only acts on partition with complete state 24 three continual challenges 25 three continual challenges scalability -large topology, huge volume of events, flow initiations 25 three continual challenges scalability -large topology, huge volume of events, flow initiations reliability -handle equipment (and other) failover gracefully 25 three continual challenges scalability -large topology, huge volume of events, flow initiations reliability -handle equipment (and other) failover gracefully performance -low control-plane latency 25
Search Related
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks