OMNI and the 6M’s of modern internetworking

By on 18 May 2022

Category: Tech matters

Tags: ,

Blog home

This post is the third in a series on a new layer in the Internet architecture known as the adaptation layer. The first two posts examined the Overlay Multilink Network Interface (OMNI) Adaptation Layer (OAL) mechanics and provisions for reliability, integrity, efficiency and security. In this post, we’ll explore the advantages of the adaptation layer that support the ‘6 M’s of modern internetworking’.

The OAL establishes a new layer in the Internet architecture below the IP layer (Layer 3) and above the data link layer (Layer 2). The IP layer engages the OAL via an OMNI interface configured over multiple Layer 2 (physical or virtual) underlay interfaces. The OAL encapsulates IP packets produced at Layer 3 in the adaptation layer IPv6 headers then performs fragmentation followed by Layer 2 encapsulation to produce ‘carrier packets’ for transmission over underlay interfaces.

OMNI Client end systems (and near-end systems), as well as a nominal collection of supporting infrastructure nodes known as Proxy/Servers, Relays and Gateways, configure an OMNI interface as their point of attachment to a Non-Broadcast, Multiple Access (NBMA) virtual OMNI link. The OMNI link spans the adaptation layer over the entire Internet, while the vast majority of Internet routers can continue to forward carrier packets the same as ordinary IP packets. The adaptation layer can therefore be incrementally deployed on increasing numbers of OMNI nodes without disturbing the vast majority of the deployed Internet base.

In addition to OMNI and the OAL, the Automatic Extended Route Optimization (AERO) control messaging service provides the neighbor coordination functions necessary for incremental deployment. But, before embarking on solutions we must first consider the primary motivating reasons for establishing a new adaptation layer in the architecture, that is, the ‘why’ as opposed to the ‘how’. This post examines the ‘6 M’s of modern internetworking’, namely Multilink, Multinet, Mobility, Multicast, Multihop and MTU assurance.

Multilink — Harnessing redundant links for reliable communications

More and more, ‘Client’ devices such as laptop computers, cell phones, air, land, sea and space vehicles include multiple data link technologies that could provide cost, performance and/or reliability operational advantages if grouped as a bundle of alternate and/or redundant communications links. For example, airplanes commonly include multiple link types such as VHF and L-Band terrestrial radios plus several forms of satellite services.

With the advancement in 5G cellular services, next generation cellphones will also include multiple link types including cellular, Wi-Fi, Bluetooth and even satellite and/or omnidirectional peer-to-peer radios. For such devices, the adaptation layer presents a new capability that can effectively employ multiple links simultaneously instead of singularly per the current operational model.

The OMNI interface provides the IP layer of such multilink Clients with a single network interface abstraction. The adaptation layer below provides dynamic mapping of incoming and outgoing carrier packets via the best underlay link or links based on factors such as link quality, cost, and performance.

Peer end systems can further engage the AERO control messaging service through IPv6 Neighbor Discovery (ND) message exchanges over the OMNI link to negotiate traffic profiles for using incoming and outgoing data links for a given traffic flow based on traffic selectors. Consider, for example, a cellphone that can engage all of its radio interface types simultaneously instead of one at a time while also communicating with a peer cellphone doing the same thing that could be located either nearby or in a far distant corner of the globe.

It is important to understand that Client applications can use a single, stable IP address at Layer 3 when engaging peers even though the adaptation layer may spread the carrier packets over multiple underlay interfaces. Each peer, therefore, can use the IP layer address as a constant and unchanging identifier for any communications sessions invoked by upper layers and need not have any visibility into the adaptation layer underlay link selections.

Multinet — Spanning the globe over diverse administrative domains

The public Internet can be regarded as a giant collection of links and network layer devices (routers, bridges, switches and so forth) that seamlessly span the globe allowing worldwide services such as web browsing, online banking, and social media. But, there are also many private networks worldwide with restricted connections to the public Internet (for example, through firewalls, proxies, and Network Address Translators (NATs)) or with no connection to the public Internet at all. For example, large corporations often have their own private ‘Intranets’ that span the globe and yet have only very restrictive (or even non-existent) public Internet physical points of connection.

The adaptation layer introduces the ability to ‘bridge’ such private Intranets as concatenated partitions of an overlay virtual link to allow for end-to-end overlay traversal even though carrier packets in the underlay may traverse the private networks of multiple diverse organizations with no security trust relationships. The AERO/OMNI architecture accomplishes this by establishing node types known as ‘Gateways’ between adjacent private Intranets. The Gateway model is not new but had its origins in the ‘Catenet Model for Internetworking’ published in 1978 (and still earlier publications).

AERO/OMNI Gateways establish a Border Gateway Protocol (BGP) overlay network to form an adaptation layer OMNI link over the (concatenated) private networks. Each Gateway establishes a secure underlay link with adjacent private Intranets (or their Gateways) while engaging the security parameters of each. The deployment of additional Gateways to interconnect other private Intranets extends the OMNI link and even allows for secured wide-area peer-to-peer communications over diverse private networks that would not ordinarily trust one another. Moreover, the adaptation layer supports multilink forwarding services over multinet concatenations so that even peer nodes located across a wide area concatenation of diverse administrative domains can continue to engage in multilink for reliability, performance and cost-effectiveness.

Mobility — Accommodating disruption at lower layers for upper layer stability

The adaptation layer provides Mobile Clients with a fixed Layer 3 IP Mobile Network Prefix (MNP) maintained by the mobility service. The MNP remains allocated to the Client even if its Layer 2 data link connections change dynamically due to mobility. The OMNI interface adaptation layer abstraction, therefore, conceals lower-layer mobility changes from upper layers, which see only stable IP addresses within the MNP.

The adaptation layer also provides each Client with an abstraction that sees other Clients as single-hop neighbors reachable over the OMNI virtual link while other nodes can be reached via nearby OMNI relays that forward Layer 3 packets to off-link correspondents. AERO/OMNI Clients (whether mobile or fixed) coordinate each of their underlay interfaces with a first-hop Proxy/Server, which in turn coordinates with a Hub Proxy/Server in a hub-and-spoke arrangement. The Hub Proxy/Server then injects the Client’s MNP into the BGP routing system so that all aspects of the multilink, multinet and mobility services are coordinated.

As a Client moves, its data link layer IP addresses may change dynamically, for example, as a result of changing to a new wireless base station. The Client updates its OMNI link neighbors by sending unicast IPv6 ND Neighbor Advertisement (NA) messages to announce new data link layer address binding. This allows upper layer protocol sessions to continue uninterrupted without readdressing even if lower layer addresses are changing dynamically. A Client that moves away from an old Proxy/Server on a specific underlay interface can also send unicast IPv6 ND Router Solicitation (RS) messages to associate with a new (first-hop) Proxy/Server and possibly also a new Hub Proxy/Server. These new Proxy/Servers can then coordinate the Clients’ movements with the OMNI link-wide mobility service.

Multicast — Accommodating network efficiency through reduced transmissions

The subject of IP multicast has been of keen interest to the Internet community for many decades. The IP multicast service promises to improve network efficiency through reduced transmissions by sending a single packet that is received by multiple multicast group members instead of just a single unicast destination. However, extending multicast services to the Internet has proven challenging due to the difficulty of ensuring multicast routing services across a wide area, and not just within local area network (LAN) links.

Since the adaptation layer establishes an OMNI virtual link that spans the (concatenated) wide area underlay network, IP multicasting can consider the overlay the same as for any NBMA link with LAN properties. The OMNI link can therefore establish multicast forwarding state in Proxy/Servers so that a small amount of packet duplication can be engaged if necessary to ensure that packets reach all multicast group members. This is supported by extending OMNI link Neighbor Cache Entries (NCEs) to also maintain multicast group membership state according to Protocol Independent Multicast (PIM) messaging.

Multihop — Extending the OMNI link over the mobile edge

The OMNI link extends over well connected underlay networks up to a point where fixed network infrastructure ends and gives way to ad hoc structures where communications must be relayed between the mobile nodes themselves, for example, in a Mobile Ad-hoc Network (MANET) or Vehicular Ad-hoc Network (VANET). This means that peer-to-peer or Vehicle-to-Vehicle (V2V) multihop relaying may be necessary to convey the packets from a Mobile Client located many hops away from a node within the range of infrastructure (for example, Vehicle-to-Infrastructure (V2I)).

The adaptation layer of each AERO/OMNI node configures IPv6 addresses taken from a Unique-Local Address (ULA) range provisioned for the OMNI link. Mobile Clients that do not yet have a connection to infrastructure can also assign temporary ULAs for the operation of initial exchanges until they receive an actual ULA assignment from the OMNI link. When a Client joins a MANET/VANET, it injects its adaptation layer ULA into the (multihop) routing service that establishes a return path for packets destined for the Client. The Client can then send IPv6 ND Router Solicitation (RS) messages to receive Router Advertisements (RA) from an OMNI link Proxy/Server that may be many hops away. The MANET/VANET routing protocol supports multihop forwarding to extend the adaptation layer virtual link over the ad hoc edge.

MTU assurance — Large packet services for upper layers

The Internet today operates with a fundamentally flawed service for discovering the largest packet size that can traverse a given path without loss due to a size restriction. The service (known as Path MTU discovery) relies on a feedback channel from the network that is prone to loss or corruption of important control messages. For this reason, Path MTU discovery in the Internet has impeded growth to larger packet sizes and rarely allows for transmission of IP packets that exceed 1,500 octets. This limited size can be restrictive for upper-layer protocols that prefer to send larger packets for greater efficiency and performance.

The adaptation layer addresses this issue by providing upper layers with an assured MTU of at least 65,535 octets, that is, even though fragmentation may be needed to present lower layers with carrier packets small enough to traverse the path without loss due to a size restriction. The adaptation layer can also offer larger MTUs to upper layers if Jumbogram support is desired, with the understanding that these larger packets may encounter a ‘best-effort’ (not assured) forwarding service.

The adaptation layer further supports a new class of packets known as IP Parcels, defined as a single IP packet that contains multiple upper-layer protocol segments, that is, a ‘packet-of-packets’. The adaptation layer breaks parcels larger than 65,535 octets into smaller sub-parcels and then applies fragmentation if necessary to provide assured delivery services. The adaptation layer egress node can then reassemble the sub-parcels into a smaller number of larger parcels, or can instead deliver the sub-parcels immediately to upper layers since they include an integral number of upper-layer protocol segments. Studies have shown that upper layer protocols can receive improved performance using such packaging.

The adaptation layer discussed above opens new opportunities and advantages not possible in traditional approaches. The next article in this series will explore how AERO, OMNI and Delay-Tolerant Networking (DTN) services together provide a multi-layered approach suitable for a wide variety of mobile internetworking applications. The article will further explain the multi-layered addressing scheme and inter-workings of the various layers.

Fred Templin is an Internet networking research engineer working in the industry since 1986, where he was deeply involved in the evolution of the Internet Protocol over Ethernet, FDDI, ATM and other data link and network layer technologies.

Rate this article

The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.

Leave a Reply

Your email address will not be published.