Transcript Around the IETF Other IETF Work Of

Around the IETF
Other IETF Work Of-Interest to ICCRG
Wesley Eddy
MTI Systems
[email protected]
Internet Congestion Control Research Group (ICCRG)
IETF 77, Anaheim, March 21-26, 2010
Goals
• Raise awareness of work going on in the IETF which ICCRG
might be interested in
– Some WG’s products require congestion control mechanisms and they
can benefit from review & guidance
– Some WG’s products may interact with congestion control
mechanisms at other points in the stack or path in interesting or
unforeseen ways
• Not exhaustive; based just on personal review
• There are also several BoFs of interest, but the goals of those
are often fluffier, so I’ve left them out for now …
Applications Area
• A lot of things that simply use TCP … probably not so
interesting to ICCRG
• Application-Layer Traffic Optimization (ALTO) WG
– From charter: “… provide applications with information to perform
better-than-random initial peer selection. ALTO services may take
different approaches at balancing factors such as maximum
bandwidth, minimum cross-domain traffic, lowest cost to the user,
etc.”
– Since P2P traffic is a large bandwidth consumer in many networks,
ALTO’s work may have an impact on large-scale traffic patterns and
network loading / congestion profiles
Internet Area
• Locator/ID Separation Protocol (lisp) WG – developing
experimental mechanisms for routing system scalability
– May cause interesting packet delays or reordering as mappings are
resolved for initial packets
– See “data probe” concept discussed in LISP+ALT
• Pseudowire Emulation Edge to Edge (pwe3) WG – developed
“pseudowire” technology for emulating a point-to-point or
point-to-multipoint link perceived by users as an unshared
link/circuit over an IP/MPLS network.
– ICCRG open-issues document discusses multi-domain pseudowires
– There is a pseudowire congestion control framework:
• See: draft-ietf-pwe3-congestion-frmwk-02
Operations & Management Area
• Nothing in particular …
• Interesting high-level questions ICCRG might think about
could include how congestion is detected, measured, and
responded to by network management (having different
visibility, timescales, and mechanisms available than a
transport protocol has, for instance)
Real-time Applications & Infrastructure Area
• The question of “overload protection” or “overload control”,
which differs from what congestion control classically
considers, is relevant to some of the RAI WGs.
Routing Area
• The Bidirectional Forwarding Detection (BFD) WG defined a
protocol that signals liveness; the specification requires
congestion control for use over multihop paths:
– See draft-ietf-bfd-base-11
• Congestion control for MANET has been a popular subject in
the research realm
Security Area
• Nothing that I know of …
Transport Area
• Datagram Congestion Control Protocol (DCCP)
– Maintains TFRC and the DCCP protocol capable of incorporating
new/alternative congestion control mechanisms
• Low Extra Delay Background Transport (ledbat)
– Specifying a congestion control mechanism rather than a transport
protocol itself
– Goal includes a responsiveness to delays beyond a threshold; intended
use includes P2P file transfer, e.g. BitTorrent
– See: draft-ietf-ledbat-congestion-00
• Multipath TCP (mptcp) WG is working on experimental
mechanisms to allow TCP to use multiple paths
• Congestion & Pre-Congestion Notification (PCN) WG
– Focused on protecting QoS for flows within a DiffServ domain
– Detecting, signaling, ingesting, exporting, and reacting to congestion
information within the network
• TCP Maintenance & Minor Extensions (TCPM) WG