Network Working Group S. Shah
Internet-Draft P. Thubert
Intended status: Informational Cisco Systems
Expires: July 12, 2014 January 08, 2014
Deterministic Forwarding PHB
draft-svshah-tsvwg-deterministic-forwarding-00
Abstract
This document defines a Differentiated Services Per-Hop-Behavior
(PHB) Group called Deterministic Forwarding (DF). The document
describes the purpose and semantics of this PHB. It also describes
creation and forwarding treatment of the service class. The document
also describes how the code-point can be mapped into one of the
aggregated Diffserv service classes [RFC5127].
Status of this Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on July 12, 2014.
Copyright Notice
Copyright (c) 2014 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
Shah & Thubert Expires July 12, 2014 [Page 1]
Internet-Draft Deterministic Forwarding PHB January 2014
described in the Simplified BSD License.
This document may contain material from IETF Documents or IETF
Contributions published or made publicly available before November
10, 2008. The person(s) controlling the copyright in some of this
material may not have granted the IETF Trust the right to allow
modifications of such material outside the IETF Standards Process.
Without obtaining an adequate license from the person(s) controlling
the copyright in such materials, this document may not be modified
outside the IETF Standards Process, and derivative works of it may
not be created outside the IETF Standards Process, except to format
it for publication as an RFC or to translate it into languages other
than English.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1. Use-cases . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3. DF code-point Behavior . . . . . . . . . . . . . . . . . . . . 5
3.1. Potential implementation of DF scheduling . . . . . . . . . 6
3.2. Conditioning DF traffic at Enqueue . . . . . . . . . . . . 8
4. Diffserv behavior through non-DF DS domains . . . . . . . . . . 8
5. Updates to RFC4594 and RFC5127 . . . . . . . . . . . . . . . . 8
6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 8
7. Security Considerations . . . . . . . . . . . . . . . . . . . . 9
8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 9
9. References . . . . . . . . . . . . . . . . . . . . . . . . . . 9
9.1. Normative References . . . . . . . . . . . . . . . . . . . 9
9.2. Informative References . . . . . . . . . . . . . . . . . . 9
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 9
Shah & Thubert Expires July 12, 2014 [Page 2]
Internet-Draft Deterministic Forwarding PHB January 2014
1. Introduction
IP Networks typically implement Diffserv to provide differentiated
forwarding behavior to different class of traffic. Networks that
implement Diffserv relies on DSCP code-point in the IP header of a
packet to select PHB as a specific forwarding treatment for that
packet [RFC2474, RFC2475]. This document describes a particular PHB
called Deterministic Forwarding (DF). The proposed new code-point
defines a service class for the purpose of forwarding treatment of a
packet at determined/fixed scheduled time providing no jitter service
to the class of traffic (updates RFC4594 with the addition of a new
Service Class).
DF PHB can be used for the network services that require the
capability to ensure a predictable interaction between networked
systems and guarantee a very strict time scheduled services.
Applications of such networks may be able to absorb a loss but are
very sensitive to end to end latency and jitter. Examples of such
networks include Machine to Machine (M2M) control and monitoring
deployment with IP over varieties of Layer 2 networks.
The definition of Expedited Forwarding (EF) [RFC2598] PHB is low
latency and thus one can envision use of EF code-point for such
service. However, even though EF defines low latency and low jitter,
it does not guarantee deterministic/fixed scheduled time service.
Depending on co-existence of the other traffic in the network, EF
traffic may have more or less variance on jitter and thus not
suitable for the deterministic service. DF PHB thus is more suitable
for deterministic time sensitive traffic.
Typically for an application where end to end deterministic service
is important, relevant traffic should be provisioned through DF PHB
at every hop in that end to end path. However, in cases where
intermediate hops (or DS domains) either do not support DF PHB or
supports only aggregated service classes described in RFC5127, DF
traffic in those DS domains MUST be mapped to Real Time Treatment
class (EF PHB) defined in RFC5127. Traffic in such scenario MUST be
conditioned at the Edge before entering and after exiting such DS
domains. This is described further in later section.
1.1. Use-cases
With an introduction of machine to machine networks over IP, a new
set of applications are emerging. Traffic types from such
applications/networks are some-what different from the traditional
traffic types. Though most traffic types have characteristics
similar to that of traditional ones [LLN-DIFF], certain control
signals for some of the applications are extremely sensitive to
Shah & Thubert Expires July 12, 2014 [Page 3]
Internet-Draft Deterministic Forwarding PHB January 2014
latency and jitter. Such control signals demand much stricter
latency and jitter, at pretty much decisive time scheduled delivery,
end to end. Industrial automation, Smart cities and automobiles/
planes/trains built around such networks are examples of such use-
cases.
Machine to machine networks may be implemented on varieties of Layer
2 protocols. 802.3 and 802.15e [TiSCH] are examples of layer 2 that
are enhancing their capabilities to allow time scheduled delivery of
packets.
In a wireless sensor networks, that are implemented over IP, multiple
LLN (Low power and Lossy Networks) may be connected through Backbone.
---+------------------------
| Converged Campus Network
|
+-----+
| | Gateway
| |
+-----+
|
| Backbone
+--------------------+------------------+
| | |
+-----+ +-----+ +-----+
| | LLN border | | LLN border | | LLN border
o | | router | | router | | router
+-----+ +-----+ +-----+
o o o o
o o o o o o o o o o o
o o o o o o o o o o o o o o o o o o
o o o o o o o o o o o o o o o o
o o M o o o o o o o o o o o o o
o o o o o o o o o
o o o o o
LLN-1 LLN-2 LLN-3
As shown in the diagram, multiple LL Networks are connected to each
other via Backbone through LLN Border routers. Each LL Network
consist of many nodes. There are different types of traffic
forwarded through each LL node and from one LL Network to another.
Most LLN traffic types have characteristics similar to that of
traditional ones and thus can be supported through existing Diffserv
classes except time sensitive control signals. Without segregating
Shah & Thubert Expires July 12, 2014 [Page 4]
Internet-Draft Deterministic Forwarding PHB January 2014
such control signals to a specific Diffserv class would require
Intserv support for LLN traffic in such networks. All traffic would
be subject to flow classification to differentiate time sensitive
control signals which can be a big scale concern. Supporting time
sensitive control signals via newly proposed DF Diffserv class allows
implementation of Diffserv in LLN Networks.
2. Terminology
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC2119.
3. DF code-point Behavior
The DF PHB is to implement time scheduled forwarding treatment.
Provisioning of such a service has two parts,
1) Provisioning of the fixed/relative time for scheduling of such
service
2) Provisioning of the max size of the data to be transmitted at
each scheduled time
Provisioned scheduled time may be absolute or relative. For example,
a DF class may be provisioned to schedule packets (or bytes) at every
fixed time. Fixed time can be time of a day or any other absolute
definition. In a multi hop forwarding of DF traffic, absolute time
service provisioning at each hop may require to be dependent on the
clock synchronization (clock synchronization is not in the scope of
this specification). In relative time scheduling, packets to be
scheduled at every specific interval or it could be relative to any
other specific event/trigger. The definition of the time interval or
any other event is relevant to that specific provisioned node only.
The size of the data, to be transmitted at each scheduled time
service, provisioned can be in the unit of bytes or time. Once DF
PHB is provisioned and enabled, forwarding treatment MUST service
packets (bytes) from this class at the scheduled time for max
allowable data. Scheduling MUST pre-empt any other service,
including EF, during the schedule time service for the DF class. In
order to avoid incurred latency to EF class of traffic, it is
expected to carefully provision DF class to limit scheduled time
service to as minimal data transmission that would prevent larger
than expected delays to EF class of traffic.
Shah & Thubert Expires July 12, 2014 [Page 5]
Internet-Draft Deterministic Forwarding PHB January 2014
Provisioning can be done via any of multiple possible methods. It
could be via command interface, or could be via external provisioning
agents, or could be via some sort of signaling that may dynamically
pre-negotiate time window of transmission at each node in a network
path.
3.1. Potential implementation of DF scheduling
Following are examples of potential implementations. They are not
any form of guidelines or recommendations.
There are at least two ways to implement scheduling for DF traffic
class.
1) One queue to buffer and schedule all DF traffic (from all flows),
2) Multiple sub-queues for DF traffic class, one queue for each DF
provisioned flow
Flow here represents macro definition, it does not have to be only
5-tuple.
Any chosen DF scheduling implementation MUST run traffic conditioning
at enqueue to decide if packets to be enqueued or discarded.
Discussed more in later section.
1) One data-plane queue to buffer all DF traffic
This one queue maintains, possibly a circular, indexed buffer list.
Every time scheduled slot is an index in the buffer list. If enqueue
conditioning decides not to discard a packet, packet gets en-queued
at the relevant index in the buffer list in such a way that relevant
index pointer, and thus buffered relevant packets for that index, at
the head of the list is ready to be de-queued at next scheduled time.
Subsequent buffer index is scheduled for the subsequent scheduled
time slot and so forth. If a specific flow has not received any
packet for a scheduled time then buffer index for that flow remains
empty. A packet from other flows do not get buffered at that empty
index. That means during dequeue, at a schedule time service, an
empty index results in no packets to dequeue and thus nothing to be
transmitted from the DF queue at that point in time.
Shah & Thubert Expires July 12, 2014 [Page 6]
Internet-Draft Deterministic Forwarding PHB January 2014
.
|`.
EF (Low latency) ----------------||----> `.
High | `.
. | `.
rate queues |`. | `.
AF1 ----||---> `. | `.
| `. | `.
AF3 ----||---> '------------------> '------>
| .' Low | .'
BE ----||---> .' | .'
| .' | .'
.' | .'
Deterministic| .'
DF ----------------||----> .'
(scheduled time/interrupt driven de-queue)|.'
2) multiple sub-queues for each DF flows
If enqueue conditioning decides not to discard a packet, packet gets
enqueued in the relevant DF sub-queue designated for that flow. At a
scheduled time slot, scheduler dequeues a packet from the respective
sub-queue. Every scheduled time service interrupt is mapped to a
specific DF sub-queue to dequeue a packet from.
.
|`.
EF (Low latency) ----------------||----> `.
High | `.
. | `.
rate queues |`. | `.
AF1 ----||---> `. | `.
| `. | `.
AF3 ----||---> '------------------> '------>
| .' Low | .'
BE ----||---> .' | .'
| .' | .'
.' | .'
(DF queues) Deterministic| .'
DF (at interval 1, 6, 11 ..) ----||----> .'
DF (at interval 3, 8, 13 ..) ----||---->.'
Shah & Thubert Expires July 12, 2014 [Page 7]
Internet-Draft Deterministic Forwarding PHB January 2014
3.2. Conditioning DF traffic at Enqueue
DF traffic MUST be conditioned at the enqueue. As per PHB
definition, packets are required to be scheduled and delivered at a
precise absolute or relative time interval. Any packet that has
missed the window of its service time MUST be discarded. That would
also mean any packet coming from the previous hop MUST be conditioned
at the enqueue for validity of its scheduled service. For example if
a DF queue is provisioned to serve a packet with less than x ms of
jitter and for an arrived packet, if next scheduled time for a packet
results in more than x ms of jitter then such packet MUST be
discarded. The enqueued packet MUST also be checked against the size
of the data. If size of the data to be enqueued in a DF queue is
bigger than what scheduled time slot is provisioned for then such
packet MUST be discarded.
4. Diffserv behavior through non-DF DS domains
In cases where DF traffic is forwarded through multiple DS domains,
DS domains close to the source and receiver understand application's
deterministic service requirement well and so MUST be provisioned for
the precise time scheduled forwarding treatment. Intermediate DS
domains MAY support DF PHB. Intermediate domains that can not
support DF PHB, DF traffic from such domains SHOULD get EF treatment,
as defined in RFC5127 for Real Time Service aggregation. Sender and
Receiver DS domains, in such cases, MUST condition DF traffic at the
respective Edge. If EF service through intermediate DS domains can
have a predictable upper bound, receiver DS domain Edge can add a
correction to an incurred latency/jitter with its own defined time
interval for DF service.
5. Updates to RFC4594 and RFC5127
This specification updates RFC4594 with an addition of a new Diffserv
Class. It also updates RFC5127 to aggregate DF class of traffic to
Real Time Aggregation Class.
6. IANA Considerations
This document defines a new DSCP code-point DF. IANA maintains the
list of existing DSCPs. Proposal is to allocate a new one for the DF
code-point.
Shah & Thubert Expires July 12, 2014 [Page 8]
Internet-Draft Deterministic Forwarding PHB January 2014
7. Security Considerations
There is no security considerations required besides ones already
understood in the context of Differentiated services architecture
8. Acknowledgements
Fred Baker and Norm Finn.
9. References
9.1. Normative References
[RFC2474] Nichols, K., Blake, S., Baker, F., and D. Black,
"Definition of the Differentiated Services Field (DS
Field) in the IPv4 and IPv6 Headers", RFC 2474,
December 1998.
[RFC2475] Blake, S., Black, D., Carlson, M., Davies, E., Wang, Z.,
and W. Weiss, "An Architecture for Differentiated
Services", RFC 2475, December 1998.
[RFC2598] Jacobson, V., Nichols, K., and K. Poduri, "An Expedited
Forwarding PHB", RFC 2598, June 1999.
[RFC5127] Chan, K., Babiarz, J., and F. Baker, "Aggregation of
Diffserv Service Classes", RFC 5127, February 2008.
9.2. Informative References
[TiSCH] Thubert, P., Watteyne, T., and R. Assimiti, "An
Architecture for IPv6 over the TSCH mode of IEEE
802.15.4e, I-D.draft-ietf-6tisch-architecture", Nov 2013.
[LLN-DIFF]
Shah, S. and P. Thubert, "Differentiated Service Class
Recommendations for LLN Traffic,
I-D.draft-svshah-tsvwg-lln-diffserv-recommendations",
Aug 2013.
Shah & Thubert Expires July 12, 2014 [Page 9]
Internet-Draft Deterministic Forwarding PHB January 2014
Authors' Addresses
Shitanshu Shah
Cisco Systems
170 W. Tasman Drive
San Jose, CA 95134
US
Email: svshah@cisco.com
Pascal Thubert
Cisco Systems
Village d'Entreprises Green Side
400, Avenue de Roumanille
Batiment T3
Biot - Sophia Antipolis 06410
FRANCE
Email: pthubert@cisco.com
Shah & Thubert Expires July 12, 2014 [Page 10]