| DetNet | J. Farkas |
| Internet-Draft | B. Varga |
| Intended status: Standards Track | Ericsson |
| Expires: January 1, 2018 | R. Cummings |
| National Instruments | |
| J. Yuanlong | |
| Z. Yiyong | |
| Huawei | |
| June 30, 2017 |
DetNet Flow Information Model
draft-farkas-detnet-flow-information-model-01
This document describes flow information model for Deterministic Networking (DetNet). The DetNet service is provided either for a Layer 3 or a Layer 2 flow. This document provides DetNet flow information model both for Layer 3 and Layer 2 flows in an integrated fashion.
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 January 1, 2018.
Copyright (c) 2017 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 described in the Simplified BSD License.
A Deterministic Networking (DetNet) service provides a capability to carry a unicast or a multicast data flow for an application with constrained requirements on network performance, e.g., low packet loss rate and/or latency. The DetNet service is provided either for a Layer 3 (L3) flow or a Layer 2 (L2) flow by an IP/MPLS network, see, e.g., [I-D.ietf-detnet-dp-alt]. Similarly, Time-Sensitive Networking (TSN) [IEEE8021TSN]) can be used for L2 flows in a bridged network. DetNet and TSN have common architecture as expressed in [IETFDetNet] and [I-D.ietf-detnet-architecture]. DetNet service can be leveraged both by L3 and L2 flows, i.e., by DetNet L3 flows and DetNet L2 flows. Therefore, the DetNet flow information model provided by this document covers both DetNet L3 flows and DetNet L2 flows in an integrated fashion. Thus, the DetNet flow information model is based on [I-D.ietf-detnet-architecture] and on the data model specified by [IEEE8021Qcc]. Furthermore, the DetNet flow information model relies on the flow identification possibilities described in [IEEE8021CB], which is used by [IEEE8021Qcc] as well. In addition to TSN data model, [IEEE8021Qcc] also specifies configuration of TSN features (e.g., traffic scheduling specified by [IEEE8021Qbv]). Due to the common architecture and flow model, configuration features can be leveraged in certain deployment scenarios, e.g., when the network that provides the DetNet service includes both L3 and L2 network segments.
Based on the DetNet architecture [I-D.ietf-detnet-architecture] (see Section 4), this document (this revision) only considers the Centralized Network / Distributed User Model out of the models specified by [IEEE8021Qcc]. That is, there is a User-Network Interface (UNI) between an end system and a network. Furthermore, there is a central entity for the control of the network. For instance, the central entity implements a Path Computation Element (PCE) for the calculation and establishment of paths needed for packet replication and elimination, if any.
[[NOTE (to be removed from a future revision): The Goals and Non goals subsections are only for initial revisions, they are to be removed from future revisions of this draft.]]
As it is expressed in the Charter [IETFDetNet], the DetNet WG collaborates with IEEE 802.1 TSN in order to define a common architecture for both Layer 2 and Layer 3, which is beneficial for various reasons, e.g., in order to simplify implementations. The flow information model should be also common along those lines. As the TSN flow information/data model specified by [IEEE8021Qcc] is mature, the DetNet flow information model described in this document is based on [IEEE8021Qcc], which is an amendment to [IEEE8021Q].
The Centralized Network / Distributed User Model of [IEEE8021Qcc] is used in this revision as a start of the work. Further models can be also useful for DetNet, e.g., the Fully Centralized Model for the Industrial M2M use case [I-D.ietf-detnet-use-cases].
This document intends to specify flow information model only.
This document (this revision) does not intend to specify either flow data model or DetNet configuration. From these aspects, the goals of this document differ from the goals of [IEEE8021Qcc], which also specifies data model and configuration of certain TSN features.
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].
The lowercase forms with an initial capital "Must", "Must Not", "Shall", "Shall Not", "Should", "Should Not", "May", and "Optional" in this document are to be interpreted in the sense defined in [RFC2119], but are used where the normative behavior is defined in documents published by SDOs other than the IETF.
This document uses the terminology established in Section 2 of the DetNet architecture document [I-D.ietf-detnet-architecture]. The DetNet <=> TSN dictionary of [I-D.ietf-detnet-architecture] is used to perform translation from [IEEE8021Qcc] to this document. Additional terms used in this document:
The following naming conventions were used for naming information model components in this document. It is recommended that extensions of the model use the same conventions.
Deterministic service is required by time/loss sensitive application(s) running on an end system during communication with its peer(s). Such a data exchange has various requirements on delay and/or loss parameters.
The DetNet architecture [I-D.ietf-detnet-architecture] distinguishes two kinds of end systems: Source and Destination. The same distinction is applied for the DetNet flow information model. In addition to the end systems interested in a flow, the status information of the flow is also important. Therefore, the DetNet flow information model relies on four high level groups:
There are two operations for each flow with respect to a Source or a Destination:
[[NOTE (to be removed from a future revision): Adding Modify operation can be considered to address cases when a flow is slightly changed, e.g., only MaxPayloadSize (Section 6.2) has been changed. The advantage of having a Modify is that it allows to initiate a change of flow spec while leaving the current flow is operating until the change is accepted. If there is no linkage between the Join and the Leave, then in figuring out whether the new flow spec can be supported, the central entity has to assume that the resources committed to the current flow are in use. If it is a Modify the central entity knows that the resources supporting the current flow can be available for supporting the altered flow. Modify is considered to be an optional operation due to possible control-plane limitations. ]]
As the DetNet UNI can provide service for both L3 and L2 flows, end systems may not need to implement the L3 <=> L2 Transfer Function specified by [IEEE8021CB] (see, e.g., subclause 6.3; see also subclause 46.1 in [IEEE8021Qcc]). An edge node may implement a function similar to the Transfer Function, see, e.g., the Svc Proxy in Figure 1 in [I-D.ietf-detnet-dp-alt].
The flows leveraging DetNet service can be unicast or multicast data flows for an application with constrained requirements on network performance, e.g., low packet loss rate and/or latency. Therefore, they can require different connectivity types: point-to-point (p2p) or point-to-multipoint (p2mp). The p2mp connectivity is created by a transport layer function (e.g., p2mp LSP) [I-D.ietf-detnet-dp-alt]. (Note that mp2mp connectivity is a superposition of p2mp connections.)
Many flows using DetNet service are periodic with fix packet size (i.e., Constant Bit Rate (CBR) flows), or periodic with variable packet size.
Delay and loss parameters are correlated because the effect of late delivery can result data loss for an application. However, not all applications require hard limits on both parameters (delay and loss). For example, some real-time applications allow graceful degradation if loss happens (e.g., sample-based processing, media distribution). Some others may require high-bandwidth connections that make the usage of techniques like packet replication economically challenging or even impossible. Some applications may not tolerate loss, but are not delay sensitive (e.g., bufferless sensors). Time/loss sensitive applications may have somewhat special requirements especially for loss (e.g., no loss in two consecutive communication cycles; very low outage time, etc.).
Flows have the following attributes:
Flow attributes are described in the following sections.
Identification options for DetNet flows at the UNI and within the DetNet domain are specified as follows; see Section 6.1.1 for DetNet L3 flows (at UNI), Section 6.1.2 for DetNet L2 flows (at UNI) and Section 6.1.3 for DetNetwork flows (within the network).
DetNet L3 flows can be identified and specified by the following attributes:
DetNet L2 flows can be identified and specified by the following attributes:
[[NOTE (to be removed from a future revision): The Multiple Stream Registration Protocol (MSRP) [IEEE8021Q] uses StreamID to match Talker registrations with their corresponding Listener registrations, i.e., to identify Streams (L2 TSN flows). The StreamID includes the following subcomponents:
]]
DetNetwork flows can be identified and specified by the following attributes:
[[NOTE (to be removed from a future revision): Attributes are based on latest data plane solution.]]
TrafficSpecification specifies how the Source transmits packets for the flow. This is effectively the promise/request of the Source to the network. The network uses this traffic specification to allocate resources and adjust queue parameters in network nodes.
TrafficSpecification has the following attributes:
[[NOTE (to be removed from a future revision): These attributes can be used to describe any type of traffic (e.g., CBR, VBR, etc.) and can be used during resource allocation to represent worst case scenarios. Further optional attributes can be considered to achieve more efficient resource allocation. Such optional attributes might be worth for flows with soft requirements (i.e., the flow is only loss sensitive or only delay sensitive, but not both delay-and-loss sensitive). Possible options how to extend TrafficSpecification attributes is for further discussion. Identified options are described in the following notes.]]
[[NOTE1: Based on the already defined attributes the most similar additional attributes for VBR type flows can be defined as follows:
]]
[[NOTE2: another alternative to deal better with various traffic types can rely on [RFC6003], which describes the support of Metro Ethernet Forum (MEF) Ethernet traffic parameters for using for resource reservation purposes. Such a Bandwidth Profile can be also adapted to describe the set of traffic parameters for a Detnet flow. Committed Rate indicates the rate at which traffic commits to be sent by the source (described in terms of the CIR (Committed Information Rate) and CBS (Committed Burst Size) attributes.) Excess Rate indicates the extent by which the traffic sent by the source exceeds the committed rate. The Excess Rate is described in terms of the EIR (Excess Information Rate) and EBS (Excess Burst Size) attributes. ]]
[[NOTE3: a third alternative is to define application based traffic models such as [GPP22885] defines periodic and event-driven traffic model, and 5G PPP work defines traffic model for MTC (Machine Type Communication) use cases. Periodic traffic type is usually for status update between devices or devices transmit status report to a central unit in regular basis. TrafficPeriod, defines the period of the status update message. DataSize, defines the data size of the massage which is constant. 3GPP also defines approximately-periodic transmission with variations on period and uncertainty in the time arrival of the packets. Event-triggered traffic type corresponds traffic being triggered by an MTC device event. MinIntervalBetweenEvent, defines the minimum interval between two events. Event-triggered transmission will not happen all the time, whenever an alert is sent, it waits until the issue being solved to be able to send another alert. MaxPacketPerEvent, defines the max number of packets within one message. ]]
FlowRank provides the rank of this flow relative to others flows in the network. This rank is used to determine success/failure of flow establishment. Rank (boolean) is used by the network to decide which flows can and cannot exist when network resources reach their limit. Rank is used to help to determine which flows can be dropped (i.e., removed from node configuration) if a port of a node becomes oversubscribed (e.g., due to network reconfiguration). The true value is more important than the false value (i.e., flows with false are dropped first).
[[NOTE (to be removed from a future revision): FlowRank specified by [IEEE8021Qcc] is according to L2 logic, where lower values are preferred.]]
The Source object specifies:
The Source object includes the following attributes:
For the join operation, the DataFlowSpecification, FlowRank, EndSystemInterfaces, and TrafficSpecification SHALL be included within the Source. For the join operation, the UserToNetworkRequirements and InterfaceCapabilities groups MAY be included within the Source.
For the leave operation, the DataFlowSpecification and EndSystemInterfaces SHALL be included within the Source.
The Destination object includes the following attributes:
For the join operation, the DataFlowSpecification and EndSystemInterfaces SHALL be included within the Destination. For the join operation, the UserToNetworkRequirements and InterfaceCapabilities groups MAY be included within the Destination.
For the leave operation, the DataFlowSpecification and EndSystemInterfaces SHALL be included within the Destination.
[[NOTE (to be removed from a future revision): Should we add DestinationRank? It could distinguish the importance of Destinations if the flow cannot be provided for all Destinations.]]
Source and Destination end systems have the following common attributes in addition to DataFlowSpecification (Section 6.1).
EndSystemInterfaces is a list of identifiers, one for each physical interface (port) in the end system acting as a Source or Destination. An interface is identified by an IP or a MAC address.
EndSystemInterfaces can refer also to logical sub-Interfaces if supported by the end system, e.g., based on IfIndex parameter.
InterfaceCapabilities specifies the network capabilities of all interfaces (ports) contained in the EndSystemInterfaces object (Section 9.1). These capabilities may be configured via the InterfaceConfiguration object (Section 11.2) of the Status object (Section 11).
Note that an end system may have multiple interfaces with different network capabilities. In this case, each interface should be specified in a distinct top-level Source or Destination object (i.e., one entry in EndSystemInterfaces (Section 9.1)). Use of multiple entries in EndSystemInterfaces is intended for network capabilities that span multiple interfaces (e.g., packet replication and elimination).";.
InterfaceCapabilities attributes:
[[NOTE (to be removed from a future revision): InterfaceCapabilities attributes are to be defined. For information, [IEEE8021Qcc] specifies the following attributes:
]]
UserToNetworkRequirements specifies user requirements for the flow, such as latency and reliability.
The UserToNetworkRequirements object includes the following attributes:
NumReplicationTrees specifies the number of maximally disjoint trees that the network should configure to provide packet replication and elimination for the flow. NumReplicationTrees is provided by the Source only. Destinations SHALL set this element to one. Value zero and one indicate no packet replication and elimination for the flow. When NumReplicationTrees is greater than one, packet replication and elimination is to be used for the flow. If the Source sets this element to greater than one, and packet replication and elimination is not possible in the network (e.g., no disjoint paths, or the nodes do not support packet replication and elimination), then the FailureCode of the Status object is non-zero (Section 11.1).
MaxLatency is the maximum latency from Source to Destination(s) for a single packet of the flow. MaxLatency is specified as an integer number of nanoseconds. When this requirement is specified by the Source, it must be satisfied for all Destinations. When this requirement is specified by a Destination, it must be satisfied for that particular Destination only. If the UserToNetworkRequirements group is not provided within the Source or Destination object, then value zero SHALL be used for this element. Value zero represents a special use for the maximum latency requirement. Value zero locks-down the initial latency that the network provides in the AccumulatedLatency parameter of the Status object (Section 11) after the successful configuration of the flow, such that any subsequent increase in the latency beyond that initial value causes the flow to fail.
[[NOTE-1 (to be removed from a future revision): Should we add a parameter to specify the maximum packet loss rate that can be tolerated for the flow?]]
[[NOTE-2 (to be removed from a future revision): TrafficSpecification (Section 6.2) specifies the Peak Information Rate (PIR) of the flow, which is a kind of user requirement to the network. Should we add Committed Information Rate (CIR), i.e., the minimum rate the user requests to be guaranteed for the flow by the network?]]
The DetNet Domain may change the encapsulation of a DetNet L2 or L3 flow at the UNI. That impacts not only how a flow can be recognised inside the DetNet domain but also the resource reservation calculations.
The DetNet Domain object specifies:
The DetNet domain object includes the following attributes:
DetnetDomainCapabilities specifies the network capabilities, which can be used to provide DetNet service. DetNet Edge nodes may change the encapsulation of a flow according to the data plane used inside the DetNet domain.
DetnetDomainCapabilities object includes the following attributes:
The Status object is provided by the network each Source and Destination of the flow. The Status object provides the status of the flow with respect to the establishment of the flow by the network. The Status object is delivered via the corresponding UNI to each Source and Destination end system of the flow. The Status is distinct for each Source or Destination because the AccumulatedLatency and InterfaceConfiguration objects are distinct, see below.
The Status object SHALL include the attributes a), b), c); and MAY include attributes d), e):
DataFlowSpecification identifies the flow for which status is provided. DataFlowSpecification is described in (Section 6.1) If the Status object is provided without a Source or Destination object in a protocol message via a UNI, then the DataFlowSpecification object SHALL be included within the Status object for both join and leave operations. If the Status object immediately follows a Source or Destination object in the protocol message, then the DataFlowSpecification object is obtained from the Source/Destination object, and therefore DataFlowSpecification is not required within the Status object.
AccumulatedLatency provides the worst-case latency that a single packet of the flow can encounter along its current path(s) in the network. When provided to a Source, AccumulatedLatency is the worst-case latency for all Destinations (worst path). AccumulatedLatency is specified as an integer number of nanoseconds. Latency is measured using the time at which the data frame´s message timestamp point passes the reference plane marking the boundary between the network media and PHY. The message timestamp point is specified by IEEE Std 802.1AS [IEEE8021AS] for various media. For a successful Status, the network returns a value less than or equal to the MaxLatency of the UserToNetworkRequirements (Section 9.3). If the NumReplicationTrees of the UserToNetworkRequirements (Section 9.3) is one, then the AccumlatedLatency SHALL provide the worst latency for the current path from the Source to each Destination. If the path is changed (e.g., due to rerouting), then the AccumulatedLatency changes accordingly. If the NumReplicationTrees of the UserToNetworkRequirements (Section 9.3) is greater than one, AccumlatedLatency SHALL provide the worst latency for all paths in use from the Source to each Destination.
StatusInfo provides information regarding the status of a flow´s configuration in the network.
The StatusInfo object MAY include the following attributes:
[[NOTE (to be removed from a future revision): FailureCodes to be defined for DetNet. Table 46-1 of [IEEE8021Qcc] describes TSN failure codes.]]
InterfaceConfiguration provides configuration of interfaces in the Source/Destination. This configuration assists the network in meeting the requirements of the flow. The InterfaceConfiguration object is according to the capabilities of the interface. InterfaceConfiguration can be distinct for each Source or Destination of each flow. If the InterfaceConfiguration object is not provided within the Status object, then the network SHALL assume zero elements as the default (no interface configuration).
The InterfaceConfiguration object MAY include one or more the following attributes:
FailedInterfaces provides a list of one or more physical interfaces (ports) in the failed node when a failure occurs in network configuration.
The InterfaceConfiguration object includes the following attributes:
InterfaceName is the name of the interface (port) within the node. This interface name SHALL be persistent, and unique within the node.
This document describes DetNet flow information model both for DetNet L3 flows and DetNet L2 flows based on the TSN data model specified by [IEEE8021Qcc]. This revision is extended with DetNet specific flow information model elements.
N/A.
N/A.
| [I-D.ietf-detnet-architecture] | Finn, N. and P. Thubert, "Deterministic Networking Architecture", Internet-Draft draft-ietf-detnet-architecture-00, September 2016. |
| [I-D.ietf-detnet-dp-alt] | Korhonen, J., Farkas, J., Mirsky, G., Thubert, P., Zhuangyan, Z. and L. Berger, "DetNet Data Plane Protocol and Solution Alternatives", Internet-Draft draft-ietf-detnet-dp-alt-00, October 2016. |
| [I-D.ietf-detnet-use-cases] | Grossman, E., Gunther, C., Thubert, P., Wetterwald, P., Raymond, J., Korhonen, J., Kaneko, Y., Das, S., Zha, Y., Varga, B., Farkas, J., Goetz, F. and J. Schmitt, "Deterministic Networking Use Cases", Internet-Draft draft-ietf-detnet-use-cases-01, February 2016. |
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997. |
| [RFC6003] | Papadimitriou, D., "Ethernet Traffic Parameters", RFC 6003, DOI 10.17487/RFC6003, October 2010. |