120 likes | 412 Views
Flow Aware Transport of MPLS Pseudowires draft-bryant-filsfils-fat-pw IETF-74. {stbryant, cfilsfil}@cisco.com, Ulrich.Drafz@t-com.net, {vach.kompella. joe.regan}@alcatel-lucent.com, shane@castlepoint.net. Changes since last. Merged with draft-vkompella-pwe3-hash-label-00.txt
E N D
Flow Aware Transport of MPLS Pseudowiresdraft-bryant-filsfils-fat-pwIETF-74 {stbryant, cfilsfil}@cisco.com, Ulrich.Drafz@t-com.net, {vach.kompella. joe.regan}@alcatel-lucent.com, shane@castlepoint.net
Changes since last • Merged with draft-vkompella-pwe3-hash-label-00.txt • Major re-write with more technical detail • Significant introductory and applicability text discussing the benefits and issues associated with deployment. • Applies to ECMP and LAG case.
Stack • Only one mode label at bottom of stack +-------------------------------+ | MPLS Tunnel label(s) | n*4 octets (four octets per label) +-------------------------------+ | PW label | 4 octets +-------------------------------+ | Flow label | 4 octets +-------------------------------+ | Optional Control Word | 4 octets +-------------------------------+ | Payload | | | | | n octets | | +-------------------------------+ FL MUST be greater than 15
Signaling • Signal the presence of the flow label but not the flow label value. • No FL TLV means feature not supported (backwards compatibility) • Draft proposes that we optionally support FL from one end only (asymmetric config) – we propose to remove this from the next version.
OAM • Basic test – single VCCV session sufficient • VCCV per FL scales poorly • PSN failure will be addressed by IGP • LFIB can be tested using draft-ietf-mpls-lsr-self-test • Trouble shoot using RFC4378 & RFC4379
Router Alert +-------------------------------+ | MPLS Tunnel label(s) | n*4 octets (four octets per label) +-------------------------------+ | Router Alert label | 4 octets +-------------------------------+ | PW label | 4 octets +-------------------------------+ | Flow label | 4 octets +-------------------------------+ | Optional Control Word | 4 octets +-------------------------------+ | Payload | | | | | n octets | | +-------------------------------+
Applicability • Large divisible flows only (for example IIP traffic) • ECMP networks • LAGs
Single Large Flows • Five options to the operator • No nothing (the case today) • Tell the customer that the flow will be policed • Same flow label for ALL high b/w flows (only hits one path) • Random label to high b/w flows (out of order delivery, but less n/w pain) • Special (not reserved) label (out of scope)
Questions • Need to look at TTL text • Should we remove the option to run without a control word
Changes in the next version • It’s really Flow Aware Transport of Pseudowires over an MPLS PSN • A bunch of nits and typos • Revert to fully symmetric operation only.
WG Draft • There is significant interest in this technology. • Should not wait for for MPLS generalized version (the pkt format is the same, the signaling will be different) • Orthogonal to the PW-bonding proposal • The authors request that the draft be accepted as a WG draft.