Operations, Administration and Maintenance (OAM) features for RAW

Internet-Draft	OAM features for RAW	September 2022
Theoleyre, et al.	Expires 6 March 2023	[Page]

Abstract

Some critical applications may use a wireless infrastructure. However, wireless networks exhibit a bandwidth of several orders of magnitude lower than wired networks. Besides, wireless transmissions are lossy by nature; the probability that a packet cannot be decoded correctly by the receiver may be quite high. In these conditions, providing high reliability and a low delay is challenging. This document lists the requirements of the Operation, Administration, and Maintenance (OAM) features are recommended to construct a predictable communication infrastructure on top of a collection of wireless segments. This document describes the benefits, problems, and trade-offs for using OAM in wireless networks to achieve Service Level Objectives (SLO).¶

1. Introduction

Reliable and Available Wireless (RAW) is an effort that extends DetNet to approach end-to-end deterministic performances over a network that includes scheduled wireless segments. In wired networks, many approaches try to enable Quality of Service (QoS) by implementing traffic differentiation so that routers handle each type of packets differently.¶

Deterministic Networking (DetNet) [RFC8655] has proposed to provide a bounded end-to-end latency on top of the network infrastructure, comprising both Layer 2 bridged and Layer 3 routed segments. Their work encompasses the data plane, OAM, time synchronization, management, control, and security aspects.¶

However, wireless networks create specific challenges. First of all, radio bandwidth is significantly lower than in wired networks. In these conditions, the volume of signaling messages has to be very limited. Even worse, wireless links are lossy: a Layer 2 transmission may or may not be decoded correctly by the receiver, depending on a broad set of parameters. Thus, providing high reliability through wireless segments is particularly challenging.¶

Wired networks rely on the concept of links. All the devices attached to a link receive any transmission. The concept of a link in wireless networks is somewhat different from what many are used to in wireline networks. A receiver may or may not receive a transmission, depending on the presence of a colliding transmission, the radio channel's quality, and the external interference. Besides, a wireless transmission is broadcast by nature: any neighboring device may be able to decode it. This document includes detailed information on the implications for the OAM features.¶

Last but not least, radio links present volatile characteristics. If the wireless networks use an unlicensed band, packet losses are not anymore temporally and spatially independent. Typically, links may exhibit a very bursty characteristic, where several consecutive packets may be dropped because of, e.g., temporary external interference. Thus, providing availability and reliability on top of the wireless infrastructure requires specific Layer 3 mechanisms to counteract these bursty losses.¶

Operations, Administration, and Maintenance (OAM) Tools are of primary importance for IP networks [RFC7276]. They define a toolset for fault detection, isolation, and performance measurement.¶

The primary purpose of this document is to detail the specific requirements of the OAM features recommended to construct a predictable communication infrastructure on top of a collection of wireless segments. This document describes the benefits, problems, and trade-offs for using OAM in wireless networks to provide availability and predictability.¶

1.1. Terminology

In this document, the term OAM will be used according to its definition specified in [RFC6291]. We expect to implement an OAM framework in RAW networks to maintain a real-time view of the network infrastructure, and its ability to respect the Service Level Objectives (SLO), such as delay and reliability, assigned to each data flow.¶

We re-use here the same terminology as [I-D.ietf-detnet-oam-framework]:¶

OAM entity: a data flow to be monitored for defects and/or its performance metrics measured.;¶
Test End Point (TEP): OAM devices crossed when entering/exiting the network. In RAW, it corresponds mostly to the source or destination of a data flow. OAM message can be exchanged between two TEPs;¶
Monitoring endPoint (MonEP): an OAM system along the flow; a MonEP MAY respond to an OAM message generated by the TEP;¶
control/management/data plane: the control and management planes are used to configure and control the network (long-term). On a per-node basis, the data plane applies rules and policies for each packet. For example, selecting the time-frequency block or the next hop on a packet-by-packet basis. Relative to a data flow, the control and/or management plane can be out-of-band;¶
Active measurement methods (as defined in [RFC7799]) modify a normal data flow by inserting novel fields, injecting specially constructed test packets [RFC2544]). It is critical for the quality of information obtained using an active method that generated test packets are in-band with the monitored data flow. In other words, a test packet is required to cross the same network nodes and links and receive the same Quality of Service (QoS) treatment as a data packet. Active methods may implement one of these two strategies:¶
- In-band: control information follows the same path as the data packets. In other words, a failure in the data plane may prevent the control information from reaching the destination (e.g., end-device or controller).¶
- out-of-band: control information is sent separately from the data packets. Thus, the behavior of control vs. data packets may differ;¶
Passive measurement methods [RFC7799] infer information by observing unmodified existing flows.¶

We also adopt the following terminology, which is particularly relevant for RAW segments.¶

piggybacking vs. dedicated control packets: control information may be encapsulated in specific (dedicated) control packets. Alternatively, it may be piggybacked in existing data packets, when the MTU is larger than the actual packet length. Piggybacking makes specifically sense in wireless networks, as the cost (bandwidth and energy) is sublinear with the packet size. Indeed, the cost to access the medium (e.g., early wake-up to deal with clock drifts) cannot be neglected, and is counted once, whatever the packet size.¶
router-over vs. mesh under: a control packet is either forwarded directly without being processed (mesh under) or handled hop-by-hop by each router. While the latter option consumes more resources, it allows collecting additional intermediary information, particularly relevant in wireless networks. For instance, each router may insert its own ID in the packet's hader, so that the destination can a posteriori know the list of IDs that actually forwarded a packet.¶
Defect: a temporary change in the network (e.g., a radio link which is broken due to a mobile obstacle);¶
Fault: a irrevocable change which may affect the network performance, e.g., a node runs out of energy.¶
End-to-end delay: the time between the packet generation and its reception by the destination.¶

1.2. Acronyms

OAM Operations, Administration, and Maintenance¶

DetNet Deterministic Networking¶

PSE Path Selection Engine [I-D.pthubert-raw-architecture]¶

QoS Quality of Service¶

RAW Reliable and Available Wireless¶

SLO Service Level Objective¶

SNMP Simple Network Management Protocol¶

SDN Software-Defined Network¶

1.3. Requirements Language

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶

2. Role of OAM in RAW

RAW networks expect to make the communications reliable and predictable over a wireless network infrastructure. Most critical applications will define an SLO required for the data flows it generates. RAW expcts to exploit OAM to improve the RAW operation at the service and the forwarding sub-layers.¶

To respect strict guarantees, RAW relies on the Path Selection Engine (PSE) (as defined in [I-D.pthubert-raw-architecture] to monitor and maintain the L3 network. An L2 scheduler may be used to allocate transmission opportunities, based on the radio link characteristics, the SLO of the flows, or the number of packets to forward. The PSE exploits the L2 resources reserved by the scheduler and organizes the L3 paths to introduce redundancy, fault tolerance and create backup paths. OAM represents the core of the pre-provisioning process by supervising the network. It maintains a global view of the network resources to detect defects, faults, over-provisioning, anomalies.¶

Fault tolerance also assumes that multiple paths must be provisioned so that an end-to-end circuit remains operational regardless of the conditions. The Packet Replication and Elimination Function ([I-D.thubert-bier-replication-elimination]) on a node is typically controlled by the PSE. OAM mechanisms can be used to monitor that PREOF is working correctly on a node and within the domain.¶

To be energy-efficient, out-of-band OAM SHOULD only be used to report aggregated statistics (e.g., counters, histograms) from the nodes using, e.g., SNMP or Netconf/Restconf using YANG-based data models. The out-of-band OAM flow MAY use a dedicated control and management channel, dedicated to this purpose.¶

RAW supports both proactive and on-demand troubleshooting. Proactively, it is necessary to detect anomalies, report defects, or reduce over-provisioning if it is not required. However, on-demand may also be required to identify the cause of a specific defect. Indeed, some specific faults may only be detected with a global, detailed view of the network, which is too expensive to acquire in the normal operating mode.¶

The specific characteristics of RAW are discussed below.¶

2.1. Link concept and quality

In wireless networks, a link does not exist physically. A device has a set of neighbors that correspond to all the devices that have a non-null probability of receiving its packets correctly. We make a distinction between:¶

point-to-point (p2p) link with one transmitter and one receiver. These links are used to transmit unicast packets.¶
point-to-multipoint (p2mp) link associates one transmitter and a collection of receivers. For instance, broadcast packets assume the existence of p2mp links to avoid duplicating a broadcast packet to reach each possible radio neighbor.¶

In scheduled radio networks, p2mp and p2p links are commonly not scheduled simultaneously to save energy and/or to reduce the number of collisions. More precisely, only a fraction of the neighbors may wake up at a given instant.¶

Anycast is used in p2mp links to improve the reliability. A collection of receivers are scheduled to wake up simultaneously, so that the transmission fails only if none of the receivers can decode the packet.¶

Each wireless link is associated with a link quality, often measured as the Packet Delivery Ratio (PDR), i.e., the probability that the receiver can decode the packet correctly. It is worth noting that this link quality depends on many criteria, such as the level of external interference, the presence of concurrent transmissions, or the radio channel state. This link quality is even time-variant. For p2mp links, consequently, we have a collection of PDR (one value per receiver). Other more sophisticated, aggregated metrics exist for these p2mp links, such as [anycast-property]¶

2.2. Broadcast Transmissions

In modern switched networks, unicast transmissions are delivered exclusively to the destination . Wireless networks are much closer to the traditional shared access wired networks. Practically, unicast and broadcast frames are handled similarly at the physical layer. The link layer is just in charge of filtering the frames to discard irrelevant receptions (e.g., different unicast MAC addresses).¶

However, contrary to wired networks, we cannot ensure that a packet is received by all the devices attached to the Layer 2 segment. It depends on the radio channel state between the transmitter(s) and the receiver(s). In particular, concurrent transmissions may be possible or not, depending on the radio conditions (e.g., do the different transmitters use a different radio channel or are they sufficiently spatially separated?)¶

2.3. Complex Layer 2 Forwarding

Multiple neighbors may receive a transmission. Thus, anycast Layer 2 forwarding helps to maximize reliability by assigning multiple receivers to a single transmission. That way, the packet is lost only if none of the receivers decode it. Practically, it has been proven that different neighbors may exhibit very different radio conditions, and that reception independence may hold for some of them [anycast-property].¶

2.4. End-to-end delay

In a wireless network, additional transmissions opportunities are provisioned to accommodate for packet losses. Thus, the end-to-end delay consists of:¶

Transmission delay, which is fixed and depends mainly on the data rate, and the presence or absence of an acknowledgement.¶
Residence time, corresponds to the buffering delay and depends on the schedule. To account for retransmissions, the residence time is equal to the difference between the time of last reception from the previous hop (among all the retransmissions) and the time of emission of the last retransmission.¶

3. Operation

OAM features will enable RAW with robust operation both for forwarding and routing purposes.¶

3.1. Information Collection

The model for exchanging information should be the same as for a DetNet network to ensure inter-operability. YANG may typically fulfill this objective.¶

However, RAW networks imply specific constraints (e.g., low bandwidth, packet losses, cost of medium access) that may require to minimize the volume of information to collect. Thus, we discuss in Section 4.2 different ways to collect information, i.e., transfer the OAM information physically from the emitter to the receiver. This corresponds to passive OAM as defined in [RFC7799].¶

3.2. Continuity Check

Similarly to DetNet, we need to verify that the source and the destination are connected (at least one valid path exists).¶

3.3. Connectivity Verification

As in DetNet, we have to verify the absence of misconnection. We focus here on the RAW specificities.¶

Because of radio transmissions' broadcast nature, several receivers may be active at the same time to enable anycast Layer 2 forwarding. Thus, the connectivity verification must test any combination. We also consider priority-based mechanisms for anycast forwarding, i.e., all the receivers have different probabilities of forwarding a packet. To verify a delay SLO for a given flow, we must also consider all the possible combinations, leading to a probability distribution function for end-to-end transmissions. If this verification is implemented naively, the number of combinations to test may be exponential and too costly for wireless networks with low bandwidth.¶

3.4. Route Tracing

Wireless networks are broadcast by nature: a radio transmission can be decoded by any radio neighbor. In multihop wireless networks, several paths exist between two endpoints. In hub networks, a device may be covered by several Access Points. We should choose the most efficient path or AP, concerning specifically the reliability, and the delay.¶

Thus, multipath routing / multi-attachment can be viewed as making the network more fault-tolerant. Even better, we can exploit the broadcast nature of wireless networks: we may have multiple Monitoring Endpoints (MonEP) for each of these kinds of hop. While it may be reasonable in the multi-attachment case, the complexity quickly increases with the path length. Indeed, each Maintenance Intermediate Endpoint has several possible next hops in the forwarding plane. Thus, all the possible paths between two maintenance endpoints should be retrieved, which may quickly become intractable if we apply a naive approach.¶

3.5. Fault Verification/detection

Wired networks tend to present stable performances. On the contrary, wireless networks are time-variant. We must consequently make a distinction between normal evolutions and malfunction.¶

3.6. Fault Isolation/identification

The network has isolated and identified the cause of the fault. While DetNet already expects to identify malfunctions, some problems are specific to wireless networks. We must consequently collect metrics and implement algorithms tailored for wireless networking.¶

For instance, the decrease in the link quality may be caused by several factors: external interference, obstacles, multipath fading, mobility. It is fundamental to be able to discriminate the different causes to make the right decision.¶

4. Administration

The RAW network has to expose a collection of metrics to support an operator making proper decisions, including:¶

Packet losses: the time-window average and maximum values of the number of packet losses have to be measured. Many critical applications stop working if a few consecutive packets are dropped;¶
Received Signal Strength Indicator (RSSI) is a very common metric in wireless to denote the link quality. The radio chipset is in charge of translating a received signal strength into a normalized quality indicator;¶
Delay: the time elapsed between a packet generation / enqueuing and its reception by the next hop;¶
Buffer occupancy: the number of packets present in the buffer, for each of the existing flows.¶
Battery lifetime: the expected remaining battery lifetime of the device. Since many RAW devices might be battery-powered, this is an important metric for an operator to make proper decisions.¶
Mobility: if a device is known to be mobile, this might be considered by an operator to take proper decisions.¶

These metrics should be collected per device, virtual circuit, and path, as DetNet already does. However, in RAW, we have to deal with them at a finer granularity:¶

per radio channel to measure, e.g., the level of external interference, and to be able to apply counter-measures (e.g., blacklisting).¶
per physical radio technology / interface, if a device has multiple NICs.¶
per link to detect misbehaving link (asymmetrical link, fluctuating quality).¶
per resource block: a collision in the schedule is particularly challenging to identify in radio networks with spectrum reuse. In particular, a collision may not be systematic (depending on the radio characteristics and the traffic profile).¶

4.1. Worst-case metrics

RAW inherits the same requirements as DetNet: we need to know the distribution of a collection of metrics. However, wireless networks are known to be highly variable. Changes may be frequent, and may exhibit a periodical pattern. Collecting and analyzing this amount of measurements is challenging.¶

Wireless networks are known to be lossy, and RAW has to implement strategies to improve reliability on top of unreliable links. Reliability is typically achieved through Automatic Repeat Request (ARQ), and Forward Error Correction (FEC). Since the different flows don't have the same SLO, RAW must adjust the ARQ and FEC based on the link and path characteristics.¶

4.2. Efficient measurement retrieval (Passive OAM)

We have to minimize the number of statistics / measurements to exchange:¶

energy efficiency: low-power devices have to limit the volume of monitoring information since every bit consumes energy.¶
bandwidth: wireless networks exhibit a bandwidth significantly lower than wired, best-effort networks.¶
per-packet cost: it is often more expensive to send several packets instead of combining them in a single link-layer frame.¶

In conclusion, we have to take care of power and bandwidth consumption. The following techniques aim to reduce the cost of such maintenance:¶

on-path collection: some control information is inserted in the data packets if they do not fragment the packet (i.e., the MTU is not exceeded). Information Elements represent a standardized way to handle such information. IP hop by hop extension headers may help to collect metrics all along the path;¶
flags/fields: we have to set-up flags in the packets to monitor to be able to monitor the forwarding process accurately. A sequence number field may help to detect packet losses. Similarly, path inference tools such as [ipath] insert additional information in the headers to identify the path followed by a packet a posteriori.¶
hierarchical monitoring: localized and centralized mechanisms have to be combined together. Typically, a local mechanism should continuously monitor a set of metrics and trigger remote OAM exchanges only when a fault is detected (but possibly not identified). For instance, local temporary defects must not trigger expensive OAM transmissions. Besides, the wireless segments often represent the weakest parts of a path: the volume of control information they produce has to be fixed accordingly.¶

Several passive techniques can be combined. For instance, the DetNet forwarding sublayer MAY combine In-band Network Telemetry (INT) with P4, iOAM and iPath to compute and report different statistics in the track (e.g., number of link-layer retransmissions, link reliability).¶

4.3. Reporting OAM packets to the source (Active OAM)

The Test EndPoint will collect measurements from the OAM probes received in the monitored track. However, the aggregated statistics must then be reported to the other Test Endpoint that injected the probes. Unfortunately, the monitored track MAY be unidirectional. In this case, the statistics have to be reported out-of-band (through, e.g., a dedicated control or management channel).¶

It is worth noting that Active OAM and Passive OAM techniques are not mutually exclusive. In particular, Active OAM is useful when a statistic cannot be accurately acquired passively.¶

Besides, Active OAM may also use piggybacking techniques: the OAM packet may be piggybacked in a frame if the MTU is sufficient. Indeed, increasing the number of transmissions in radio networks may very negatively impact the performance of radio networks, particularly for scheduled access, with fixed timeslot durations. Thus, OAM packets may be buffered until another frame has sufficient space, and has to be transmitted to the same neighbor. In conclusion, active OAM packets may be out-of-band or in-band.¶

5. Maintenance

Maintenance needs to facilitate the maintenance (repairs and upgrades). In wireless networks, repairs are expected to occur much more frequently, since the link quality may be highly time-variant. Thus, maintenance represents a key feature for RAW.¶

5.1. Soft transition after reconfiguration

Because of the wireless medium, the link quality may fluctuate, and the network needs to reconfigure itself continuously. During this transient state, flows may begin to be gradually re-forwarded, consuming resources in different parts of the network. OAM has to make a distinction between a metric that changed because of an usual network change (e.g., flow redirection) and an unexpected event (e.g., a fault).¶

5.2. Predictive maintenance

RAW needs to implement self-optimization features. While the network is configured to be fault-tolerant, a reconfiguration may be required to keep on respecting long-term objectives. Obviously, the network keeps on respecting the SLO after a node crash, but a reconfiguration is required to handle future faults. In other words, the reconfiguration delay MUST be strictly smaller than the inter-fault time.¶

The network must continuously retrieve the state of the network, to judge about the relevance of a reconfiguration, quantifying:¶

the cost of the sub-optimality: resources may not be used optimally (e.g., a better path exists);¶
the reconfiguration cost: the controller needs to trigger some reconfigurations. For this transient period, resources may be twice reserved, and control packets have to be transmitted.¶

Thus, reconfiguration may only be triggered if the gain is significant.¶

10. Informative References

[anycast-property]: Teles Hermeto, R., Gallais, A., and F. Theoleyre, "Is Link-Layer Anycast Scheduling Relevant for IEEE 802.15.4-TSCH Networks?", 2019, <https://doi.org/10.1109/LCNSymposium47956.2019.9000679>.
[I-D.ietf-detnet-oam-framework]: Mirsky, G., Theoleyre, F., Papadopoulos, G. Z., Bernardos, C. J., Varga, B., and J. Farkas, "Framework of Operations, Administration and Maintenance (OAM) for Deterministic Networking (DetNet)", Work in Progress, Internet-Draft, draft-ietf-detnet-oam-framework-06, 13 June 2022, <https://datatracker.ietf.org/doc/html/draft-ietf-detnet-oam-framework-06>.
[I-D.pthubert-raw-architecture]: Thubert, P., Papadopoulos, G. Z., and L. Berger, "Reliable and Available Wireless Architecture/Framework", Work in Progress, Internet-Draft, draft-pthubert-raw-architecture-09, 7 July 2021, <https://datatracker.ietf.org/doc/html/draft-pthubert-raw-architecture-09>.
[I-D.thubert-bier-replication-elimination]: Thubert, P., Eckert, T., Brodard, Z., and H. Jiang, "BIER-TE extensions for Packet Replication and Elimination Function (PREF) and OAM", Work in Progress, Internet-Draft, draft-thubert-bier-replication-elimination-03, 3 March 2018, <https://datatracker.ietf.org/doc/html/draft-thubert-bier-replication-elimination-03>.
[ipath]: Gao, Y., Dong, W., Chen, C., Bu, J., Wu, W., and X. Liu, "iPath: path inference in wireless sensor networks.", 2016, <https://doi.org/10.1109/TNET.2014.2371459>.
[RFC2119]: Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997, <https://www.rfc-editor.org/info/rfc2119>.
[RFC2544]: Bradner, S. and J. McQuaid, "Benchmarking Methodology for Network Interconnect Devices", RFC 2544, DOI 10.17487/RFC2544, March 1999, <https://www.rfc-editor.org/info/rfc2544>.
[RFC6291]: Andersson, L., van Helvoort, H., Bonica, R., Romascanu, D., and S. Mansfield, "Guidelines for the Use of the "OAM" Acronym in the IETF", BCP 161, RFC 6291, DOI 10.17487/RFC6291, June 2011, <https://www.rfc-editor.org/info/rfc6291>.
[RFC7276]: Mizrahi, T., Sprecher, N., Bellagamba, E., and Y. Weingarten, "An Overview of Operations, Administration, and Maintenance (OAM) Tools", RFC 7276, DOI 10.17487/RFC7276, June 2014, <https://www.rfc-editor.org/info/rfc7276>.
[RFC7799]: Morton, A., "Active and Passive Metrics and Methods (with Hybrid Types In-Between)", RFC 7799, DOI 10.17487/RFC7799, May 2016, <https://www.rfc-editor.org/info/rfc7799>.
[RFC8174]: Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, May 2017, <https://www.rfc-editor.org/info/rfc8174>.
[RFC8655]: Finn, N., Thubert, P., Varga, B., and J. Farkas, "Deterministic Networking Architecture", RFC 8655, DOI 10.17487/RFC8655, October 2019, <https://www.rfc-editor.org/info/rfc8655>.