Graft and Prune: A Lighter Fix for Streaming Video When Mobile Users Switch Networks

Every time your phone hops from one cell tower to another while you watch a live video, the network has to figure out, in milliseconds, where to send the next packets of your stream. In today’s internet that job falls to IP addresses and routing tables that were never designed with mobility in mind. A research team at King Saud University in Riyadh argues that a future architecture called Named Data Networking, or NDN, could handle this far more gracefully, and they have built a new mechanism to prove it. Their scheme, called Graft and Prune, tackles a stubborn problem that arises when mobile consumers of data move around while receiving real-time streams in NDN’s push communication mode.

Named Data Networking is one of the most prominent clean-slate redesigns of the internet’s core. Instead of delivering packets to numeric host addresses, NDN routes traffic by name: a consumer expresses interest in a piece of named data, and the network fetches it from wherever it lives. Routers keep track of pending requests in a Pending Interest Table, the PIT, which records where each Interest came from so the matching Data packet can be sent back along the reverse path. This design makes content caching natural and decouples data from location, which sounds ideal for mobile users. Yet mobility still creates friction, because the forwarding paths that carry data toward a consumer were set up when the consumer was attached to a different access point.

The friction becomes acute in push mode. In classical pull-based NDN, each Interest retrieves exactly one Data packet, so a moving consumer simply re-issues Interests from its new location and the network rebuilds paths packet by packet. Push-based communication changes that bargain. Using a construct known as a Persistent Interest, a consumer issues a single request and the producer then pushes a continuous stream of Data packets back without waiting for further Interests. This is far more efficient for real-time streaming, because it removes the round-trip delay of asking for every packet individually. The catch is that the forwarding state installed for that Persistent Interest is long-lived, and when the consumer hands over to a new access point, the old path lingers in the network as a ghost.

Those ghost paths, the researchers explain, are not merely cosmetic clutter. Until the PIT entries on the stale route expire, the network keeps wasting bandwidth by trying to deliver packets toward a location the consumer has already left, and the stream itself is disrupted because the new path has not yet been established. The persistence duration of a stale path, measured from the moment of handover to the moment the network has fully cleaned it up, is therefore a critical performance metric, and the team provides what they describe as a reproducible benchmark of this path-cleanup efficiency for each scheme they tested.

Graft and Prune is best understood as an engineering refinement of an earlier idea. A prior mechanism called Prune-and-Graft, or P&G, first removes the stale path and then establishes a new one toward the producer. That ordering works, but it opens a data-loss window in one-to-one communication: between the prune and the graft, packets have nowhere to go. The new scheme simply reverses the sequence. It grafts a fresh forwarding path toward the producer first, and only then prunes the stale one. By keeping the old path alive until the new one is ready, G&P preserves data continuity throughout the handover, eliminating the loss window that P&G incurs.

Three properties distinguish the design. It is lightweight, adding minimal signaling overhead to an already busy handover moment. It is anchor-less, meaning it does not depend on any fixed rendezvous point, home agent, or mobility anchor in the network, which keeps the architecture true to NDN’s location-independent philosophy. And it is fully NDN-compliant, requiring no changes to the packet formats or forwarding semantics that the architecture already defines. In a field where many mobility proposals quietly break the protocol’s rules, that compliance matters, because it means the mechanism could in principle be deployed on standard NDN routers without architectural surgery.

To find out whether the idea survives contact with reality, the team evaluated G&P against three anchor-less baselines: Prune-and-Graft, Path Stretch, and Bit Embedding. The experiments used ndnSIM, an open-source simulator for NDN built on the ns-3 network simulator, and ran two separate simulation campaigns, one with a stationary producer and one with a mobile producer, covering both one-to-one consumer scenarios and one-to-many scenarios in which several consumers follow the same stream. Consumer mobility speeds were varied across the trials to see how each scheme behaves for a pedestrian, a cyclist, or a fast-moving vehicle.

The results, reported in the journal Multimedia Tools and Applications, show that G&P maintains competitive performance on all state-of-the-art metrics compared with the baseline schemes across every tested condition, and that it scales gracefully as the number of consumers increases. That last point is significant for the one-to-many case, which mirrors how live streaming actually works in the wild: many viewers attached to the same producer, each potentially moving at a different moment. A mechanism that degrades as the audience grows would be a poor fit for that workload, whereas one that holds its performance under load is a genuine candidate for deployment.

The work arrives at a moment when the traffic patterns of the mobile internet are tilting heavily toward exactly the kind of continuous, delay-sensitive streams that push mode is designed to serve. Industry forecasts cited in the paper project that 5G will carry more than 80 percent of total mobile traffic by 2030, rising to 83 percent by 2031, with global mobile data traffic growing steadily through the decade. Video streaming, video calling, and interactive applications dominate that growth, and all of them punish the packet loss and service disruption that stale forwarding paths cause during handover. A protocol that shortens the window of disruption, or removes it entirely, translates directly into fewer frozen frames and fewer dropped calls.

The researchers, Achraf Gazdar, Aljawharah Almuaythir, and Muna Al-Razgan of the Department of Software Engineering at King Saud University, position G&P as a robust and practical protocol for real-time push-based streaming in mobile NDN environments rather than a theoretical curiosity. Its contribution is deliberately modest in scope and large in consequence: a reordering of two well-understood operations, graft before prune, that closes a data-loss window, plus a systematic characterization of how long stale paths haunt the network under each competing scheme. As Named Data Networking research matures from architecture papers toward deployable protocols, refinements of this kind, which respect the architecture’s constraints while fixing its mobility blind spots, are the pieces that will determine whether content-centric networking can deliver on its promise for the mobile, streaming-dominated internet now taking shape.

Subject of Research: Consumer mobility management in Named Data Networking under push communication mode for real-time streaming

Article Title: Graft and Prune: enhanced consumer mobility in NDN under push communication mode in real-time streaming applications

Article References: Gazdar, A., Almuaythir, A., & Al-Razgan, M. (2026). Graft and Prune: enhanced consumer mobility in NDN under push communication mode in real-time streaming applications. Multimedia Tools and Applications, 85(10), Article 763. https://doi.org/10.1007/s11042-026-21917-z

Image Credits: AI Generated

DOI: 10.1007/s11042-026-21917-z

Keywords: Named Data Networking, consumer mobility, push communication, real-time streaming, Persistent Interest, stale paths, ndnSIM, handover, content-centric networking, forwarding paths, protocol design, network simulation