Next-Generation Apps Enabled by SPAN
Technologist Jaime Llorca explores the next-generation consumer and industry applications that will require the scalability and low-latency inherent in SPAN.
The first wake up call unveiling the inefficiencies of current connection-oriented TCP/IP based network architectures came from content distribution. The 2000’s saw a rapid change in Internet usage, from point-to-point communication services or “conversations” between pairs of hosts, to shared content access.
Content distribution networks (CDNs) appeared as initial attempts to accommodate content delivery within the current Internet architecture. By deploying cache servers at geographically distributed network locations and storing content closer to end users, CDNs alleviate the congestion caused by multiple redundant point-to-point connections delivering the same content from origin servers to multiple end users.
While CDNs have “sufficed” to provide acceptable quality of experience (QoE) for the distribution of relatively static web content, the increasing dominance of professional and user-generated video on demand (VoD) services – such as Netflix, Hulu, Disney+, Youtube, Vimeo – and more recently live video streaming – such as live sports/events, Facebook/Instagram live, Tiktok, Twitch – have exposed the inherent limitations of the current Internet and its overlay solutions. In essence, CDNs have just moved the origin a few layers closer to the user, but what is known as the CDN edge is, at the very best, sitting at Internet exchange points (IXPs) at the telco core, and redundant content-unaware point-to-point connections are still required to stream content from CDN edge caches, resulting in unbearable metro and aggregation network congestion and bandwidth costs (see Rhett Sampson’s Live Sports Delivery article for more details on such challenges).
This is only getting worse with the advent of next-generation system automation and augmented experience services, which not only keep up the demand for high-quality video, but also add a new dimension of real-time interactivity and associated low latency requirements.
The future is about real-time interaction with content, both for industrial and consumer applications.
Think about the future of manufacturing, farming, agriculture, transportation, etc. The concept of smart factories, cities, farms, transport, logistics… is about real-time aggregation of multimedia streams from sensor devices attached to physical systems onto edge cloud servers that can process that information to understand systems’ health, predict future performance and distribute resulting actions back to actuators on those same physical systems, enabling real-time analysis and automation.
Think about the future of personal interactions and entertainment. The concept of augmented/virtual reality, telepresence, immersive video, multi-player gaming… is about real-time aggregation of source video streams, in-network media production and post-processing, and distribution of highly personalized streams to individual users.
It is apparent that this rapidly emerging class of resource- and interaction-intensive applications will disrupt the way to live, work, and interact with the physical world. But their high bandwidth, and especially low latency requirements, make delivering them over the current internet and CDN overlays not even cost-prohibitive, but literally insurmountable.
The efficient delivery of industrial automation and augmented experience services, which combine live and static data sources, intensive storage and computation requirements, and latencies on the sub-millisecond to few milliseconds range, require a fundamental new approach to content delivery. An approach that must include:
- Optimized use of highly distributed compute, storage, and network resources, spanning end devices, access points, telco edge, and telco core cloud layers.
- Programmable and elastic network slices that can be dynamically reconfigured and customized for specific application requirements.
- Content-aware connectionless networking with native multicast support to enable seamless content dissemination over the distributed cloud continuum.
Our SPAN-AI technology was designed to combine these key features as a result of the pressing need to efficiently deliver next-generation services. In SPAN-AI, the network becomes the cloud. Any network device can contribute storage, compute, and networking resources. Local AI agents guided by global control algorithms orchestrate the combined use of network resources to meet applications requirements and adaptive content-aware routing protocols enable cooperative content distribution.