Building interdom ain traffic engineering tools Steve UHLIG suh@ info.ucl.ac.be http://www.info.ucl.ac.be/~suh/ Université catholique de Louvain, Belgium 1/12 Internet architectures W G, Cam bridge (M A), 26/1/05
Context 2/12 Internet architectures W G, Cam bridge (M A), 26/1/05
State-of-the-art ● Large-scale event-driven sim ulators: – SSFNet – JSim ● Route prediction: – CBGP – RCP 3/12 Internet architectures W G, Cam bridge (M A), 26/1/05
CBGP ● An efficient open-source BGP sim ulator written by Bruno Quoitin ● http://cbgp.info.ucl.ac.be ● Key features ● allows to m odel large network topologies ● allows to reproduce BGP routing policies such as ● im port and export filters, com m unities,... ● Perl and Python interfaces for interacting with CBGP ● Part of the TOTEM TE toolbox (http://totem .info.ucl.ac.be) 4/12 Internet architectures W G, Cam bridge (M A), 26/1/05
CBGP config exam ple 5/12 Internet architectures W G, Cam bridge (M A), 26/1/05
TE over CBGP (1) ● TE tool collects – IGP inform ation for topology (e.g. pyrt) – BGP routes (e.g. zebra) – Traffic statistics (e.g. netflow) ● TE tool – changes IGP weights inside CBGP – injects eBGP update m essages inside CBGP – keeps traffic info up to date 6/12 Internet architectures W G, Cam bridge (M A), 26/1/05
TE over CBGP (2) ● Based on traffic engineering objectives, TE tool determ ines the BGP route to be advertised to each border router via iBGP – by sending different BGP UPDATEs for each im portant prefix to each ingress routers, TE tool can influence the flow of the IP traffic – ingress routers send transit traffic inside tunnels (M PLS, GRE, ...) to egress routers or TE tool ensures consistency in IP forwarding if no tunnels are used 7/12 Internet architectures W G, Cam bridge (M A), 26/1/05
CBGP case studies ● Case studies perform ed up to now: – tweaking iBGP to lim it im pact of peering failure on traffic m atrix – assessing im pact of internal link and router failures on best route choice inside the AS – optim izing (on-line) a cost function defined on outbound traffic of a stub AS – com putation of interdom ain M PLS LSPs in PCE's 8/12 Internet architectures W G, Cam bridge (M A), 26/1/05
Purpose of interdom ain TE ? (for ISPs) ● reduce congestion ● reduce costs ● im prove delays ● better load-balance traffic ● im prove network robustness ● m ake TM m ore stable ● ... 9/12 Internet architectures W G, Cam bridge (M A), 26/1/05
Open questions (for ISPs) ● Is congestion on peering links im portant ? ● Do we need online or offline TE solutions ? ● Do we need to optim ise for delay ? ● Do we need to optim ise for bandwidth ? ● Is m ultipath routing useful ? ● Do we need interdom ain M PLS ? ● Do we need new policies ? 10/12 Internet architectures W G, Cam bridge (M A), 26/1/05
Open questions (for ISPs) ● Do we need to quickly reroute in case of failures ? If yes, how fast ? ● Billing issues ? ● W hat about interdom ain QoS ? 11/12 Internet architectures W G, Cam bridge (M A), 26/1/05
Relevance of interdom ain TE ● Does som eone really care about it ? or is it just about research ? ● Is interdom ain TE aim ed at providing solutions to operational problem s or just architectural guidelines ? ● Is interdom ain TE relevant for interdom ain architecture at all ? 12/12 Internet architectures W G, Cam bridge (M A), 26/1/05
Recommend
More recommend