title: A Loss-Latency Trade-off Signal for the Mobile Network abbrev: Loss-Latency Tradeoff docname: draft-fossati-tsvwg-lola-latest date: category: std
ipr: trust200902 keyword: Internet-Draft
stand_alone: yes pi: [toc, sortrefs, symrefs]
ins: T. Fossati
name: Thomas Fossati
org: Nokia
email: [email protected]
- ins: G. Fairhurst name: Gorry Fairhurst org: University of Aberdeen email: [email protected]
- ins: P. A. Aranda Gutierrez name: Pedro A. Aranda Gutierrez org: Universidad Carlos III de Madrid email: [email protected]
- ins: M. Kuehlewind name: Mirja Kuehlewind org: ETH Zurich email: [email protected]
normative: I-D.white-tsvwg-nqb:
informative: Podlesny: title: "Rd Network Services: Differentiation Through Performance Incentives" author: - ins: M. Podlesny - ins: S. Gorinsky seriesinfo: SIGCOMM date: 2008 Custura: title: "Exploring DSCP modification pathologies in mobile edge networks" author: - ins: A. Custura - ins: A. Venne - ins: G. Fairhurst seriesinfo: TMA date: 2017 Shbair: title: "A Multi-Level Framework to Identify HTTPS Services" author: - ins: W. M. Shbair - ins: T. Cholez - ins: J. Francois - ins: I. Chrisment seriesinfo: NOMS date: 2016 Fossati: title: "Loss Latency Tradeoff and the Mobile Network" target: "https://github.com/mami-project/roadshows/blob/master/ietf-103-lola/hotrfc/build/main.pdf" author: - ins: T. Fossati - ins: P. A. Aranda Gutierrez - ins: M. Kuehlewind - ins: D. R. Lopez date: 2018
--- abstract
This document proposes a marking scheme for tagging low-latency flows (for example: interactive voice and video, gaming, machine to machine applications) that is safe to use by the mobile network for matching such flows to suitable per-hop behaviors (EPS bearers defined by 3GPP) in its core and radio segments. The suggested scheme re-uses NQB, a DiffServ-based signalling scheme with compatible rate-delay trade-off semantics that has been recently introduced in the context of fixed access to allow differential treatment of non-queue building vs queue building flows.
--- middle
Today's mobile networks are configured to bundle all flows to and from the Internet into a single "default" EPS bearer whose buffering characteristics are not compatible with low-latency traffic. The established behaviour is partly rooted in the desire to prioritise operators' voice services over competing over-the-top services. Of late, said business consideration seems to have lost momentum and the incentives might now be aligned towards allowing a more suitable treatment of Internet real-time flows. However, a couple of preconditions need to be satisfied before we can move on from the status quo. First, the real-time flows must be efficiently identified so that they can be quickly assigned to the "right" EPS bearer. This is especially important with the rising popularity of encrypted and multiplexed transports, which has the potential of increasing the cost/accuracy ratio of Multi-Field (MF) based classification over the acceptable threshold. Second, the signal must be such that malicious or badly configured nodes can't abuse it. Today's mobile networks take a rather extreme posture in this respect by actively discarding (remarking or bleaching {{Custura}}) DiffServ signalling coming from an interconnect. Therefore, the signal must be modelled in a way that the mobile network can either trust it or, even better, avoid the need of trusting it in the first place. The Rate-Delay trade-off {{Podlesny}} satisfies the above requirements and has been shown {{Fossati}} to integrate well with a simplified LTE QoS setup that uses one dedicated low-latency bearer in addition to the default.
This document suggests reusing the Non Queue Building (NQB) signalling protocol described in {{I-D.white-tsvwg-nqb}} as the method employed by endpoints to mark their real-time flows and by the LTE network to classify and route these flows via a suitable (low-latency) bearer through the LTE core network and E-UTRAN.
- DPI: Deep Packet Inspection
- EPS bearer: Evolved Packet System bearer, a virtual circuit with a given set of QoS attributes which spans the entire mobile network including the LTE core and E-UTRAN segments;
- GBR: Guaranteed Bit Rate. EPS bearers can be GBR, in which case they are guaranteed to not drop packets under congestion, or non-GBR, in which case no guarantee of delivery is made by the mobile network;
- LTE: 3GPP Long Term Evolution, aka 4G;
- E-UTRAN: LTE Radio Access Network;
- QCI: QoS Class Identifier. In LTE networks, EPS bearers are partitioned into equivalency classes modulo the QoS treatment they receive. QCI is an integer that labels a specific QoS class. Its semantics is consistently understood by all network elements involved in packet forwarding;
- UE: User Equipment, any device (e.g., smartphone, laptop, tablet) attached to an LTE network.
{::boilerplate bcp14}
Given the semantic equivalence of a Loss-Latency trade-off with the Non Queue Building (NQB) behaviour, this document reuses the NQB DSCP (Section 4 of {{I-D.white-tsvwg-nqb}}) as-is.
Endpoints SHOULD mark packets that belong to the Best Effort class and are latency sensitive by assigning the NQB DSCP value to the DS field.
The marking could also be added to other BE traffic if the default class could be reclassified by the network to use the NQB DSCP before the packet enters the mobile domain - for example by a classifier in the SGi-LAN or in an LTE router.
The Mobile network is configured to give UEs a dedicated, low-latency, non-GBR, EPS bearer with QCI 7 in addition to the default EPS bearer.
A packet carrying the NQB DSCP shall be routed through the dedicated low-latency EPS bearer. A packet that has no associated NQB marking shall be routed through the default EPS bearer.
NQB markings SHOULD be preserved when forwarding via an interconnect.
The NQB DSCP can have end-to-end semantics and this might benefit any NQB-marked traffic if utilised by other path elements (e.g. allowing DS treatment if a bottleneck link happens to be in the part of the path outisde the mobile access network segment).
This document makes no request to IANA.
Internet applications cannot benefit from wrongly indicating low-latency as they have to pay the expense of high loss as a trade-off. Hence there is no incentive for Internet applications to set the wrong code-point.
The NQB signal is not integrity protected and could be flipped by an on-path attacker. This might negatively affect the QoS of the tampered flow.
As described in {{Shbair}} state of art encrypted traffic analysis based machine learning can successfully identify the type of transported application (e.g., HTTPS, SMTP, P2P, VoIP, SSH, Skype) with good accuracy and without any need to access the clear-text. In this context, despite it being coarse grained, a 1-bit signal such as the one described in this document might be used to improve the precision of the classifier.
We would like to thank the authors of the "Latency Loss Tradeoff PHB Group" draft: Jianjie You, Michael Welzl, Brian Trammell and Kevin Smith. Big thanks to Chris Seal, Dan Druta, Diego Lopez, Shamit Bhat, Georg Mayer, Florin Baboescu, James Gruessing for the help.
This work is partially supported by the European Commission under Horizon 2020 grant agreement no. 688421 Measurement and Architecture for a Middleboxed Internet (MAMI), and by the Swiss State Secretariat for Education, Research, and Innovation under contract no. 15.0268. This support does not imply endorsement.