AmLight supporting Network Research Exhibition (NRE) demonstrations at the Supercomputing Conference(SC25)

This year the AmLight team is supporting several Network Research Exhibition (NREs):

SC25-NRE-018: A Cryo-EM Without Borders: Empowering Distributed Research via AmLight and Collaborative (Human) Networks
Caltech Booth #2824

Abstract: Over the past decade, cryo-electron microscopy (Cryo-EM) has emerged as an important technique in structural biology, dramatically expanding our capacity to visualize biological macromolecules at near-atomic resolution. This advancement has been particularly impactful in biomedical research, enabling scientists to determine the detailed structures of critical targets such as the HIV vaccine research and SARS-CoV-2 spike protein. Insights obtained from Cryo-EM have directly accelerated vaccine and therapeutic development, demonstrating its potential to transform our approach to tackling infectious diseases, many of which disproportionately burden the Global South. Despite these advances, researchers in resource-constrained settings, particularly in Sub-Saharan Africa and South America, face significant barriers in accessing Cryo-EM infrastructure due to the high cost, sophisticated instrumentation, and specialized technical expertise required.

To address these challenges, centralization of Cryo-EM capabilities within specialized imaging core facilities has proven highly effective in many regions, including Europe and North America. Such facilities concentrate resources and expertise, providing a collaborative environment that maximizes utilization and ensures that advanced instrumentation is not underutilized or misused. However, for researchers geographically distant from these facilities, especially those across Africa and South America, access has remained challenging. The recent developments in network infrastructure, particularly through initiatives like AmLight, offer an effective solution to this geographic disparity. Enhanced network capacity could enable real-time or near-real-time remote data transfer and processing, allowing scientists to remotely collect and analyze Cryo-EM datasets. By leveraging improved network access, imaging core facilities can extend their reach, democratize access to cutting-edge microscopy, and empowering researchers in the Global South to address pressing health challenges through advanced structural biology.

Here, we provide a compelling use-case scenario where researchers at the SAMRC Antibody Immunity Research Unit (AIRU) Centre for HIV & STI’s at the National Institute for Communicable Diseases (NICD) in South Africa prepare a sample consisting of an HIV vaccine candidate bound to an antibody fragment. The samples are screened at the University of Cape Town (UCT) before being shipped to the University of California, Santa Cruz (UCSC) for data acquisition and preprocessing. The resulting preprocessed dataset is transferred back to UCT for further data processing and interpretation.

SC25-NRE-121: Real-Time In-Network Machine Learning and P4 Testbed Deployment on FPGA SmartNICs, DPUs, and Switches
TBD

Abstract: The San Diego Supercomputer Center and National Research Platform are deploying a large-scale programmable testbed integrating P4-programmable SmartNICs, DPUs, and switches for real-time in-network computing research. This infrastructure includes seven 1U DC-powered measurement and monitoring servers, known as Interactive Global Research Observatory Knowledge base nodes, supplied through the PacWave NSF Award 2029306. Each IGROK node is equipped with a Bluefield 2 DPUs featuring 2×100 Ethernet interfaces and is integrated into the AutoGOLE/SENSE topology, enabling distributed monitoring and control. This SC25 Network Research Exhibition highlights two closely related efforts: 1) real-time machine learning (ML) inference embedded directly in the network data plane using fixed-point arithmetic and Taylor series approximations on AMD/Xilinx Alveo U55C FPGA SmartNICs, and 2) the deployment of a flexible, scalable P4 experimentation environment spanning geographically distributed sites.

SC25-NRE-125: Caltech/ESnet Booth 2824 A Next Generation Multi-Terabit/sec Campus and Global Network System for Data Intensive Sciences
Caltech Booth #2824

Abstract: The Global Network Advancement Group (GNA-G), through its Data Intensive Sciences (DIS) and SENSE/AutoGOLE working groups, is leading a worldwide collaboration to develop next-generation intelligent networking systems that support data-intensive scientific research across continents. At SC25, the Network Research Exhibition showcases advances involving programmable open-source network operating systems, multi-domain virtual circuit overlays, novel routing and load-balancing techniques, and stateful traffic engineering to accelerate scientific workflows. These efforts address critical challenges in global data distribution, resource coordination, and the efficient use of high-performance research and education networking infrastructures that enable large-scale scientific programs such as the LHC, the Vera Rubin Observatory, and the National and Global Research Platforms. By deploying dynamic, adaptive, software-driven architectures—potentially enhanced with machine learning—the GNA-G community is charting the course for future network operations that seamlessly integrate networks as first-class resources alongside computing and storage, improving performance for current and emerging data-intensive science worldwide.

SC25-NRE-126: GP4L – Global Platform For Labs
Caltech Booth #2824

Abstract: GP4L (Global Platform for Lab) is a global overlay network integrating 100G/400G programmable switches and SmartNICs from the GNA-G AutoGOLE/SENSE Persistent Multi-Resource Testbed and GEANT’s RARE initiative, enabling production-oriented experimentation without impacting operational networks. It leverages open network operating systems like SONiC/FRR and RARE/FreeRtr to support traditional routing while allowing seamless opt-in deployment of next-generation features. At SC25, the demonstration highlights current deployed sites, enhanced orchestration and automation capabilities—including integration with the GNA-G GREN Map initiative—and advancements in 100G software dataplane performance using VPP, SmartNICs, and multiple control plane options such as Holo Routing. These efforts also enable large-scale testing of innovative networking protocols like PolKA, a stateless source-routing architecture that uses polynomial-encoded labels for more efficient traffic engineering and telemetry across geographically distributed resources.

SC25-NRE-128: High Performance Networking with the Sao Paulo Backbone SP Linking 8 Universities and the Bella Link
Caltech Booth #2824

Abstract: Rednesp, the research and education network of São Paulo, Brazil, connects dozens of major institutions across the state while providing international connectivity to the United States, Europe, Chile, and Argentina. Its new 100 Gbps “Backbone SP” infrastructure links eight leading universities and research centers, and includes a high-speed connection to the Brazilian Synchrotron Light Laboratory (LNLS/CNPEM). With intercontinental capacity recently upgraded to multiple 100 Gbps and 800 Gbps links via AmLight, RNP, and the EllaLink submarine cable, overall bandwidth to the U.S. is expected to exceed 1 Tbps by the end of 2025. Building on the 550 Gbps throughput achievement at SC24, Rednesp aims to reach a new record of 700 Gbps at SC25 and is deploying 400 Gbps equipment to enable future demonstrations across São Paulo. Looking ahead, Rednesp will adopt advanced services from SENSE, NRP, and GP4L while collaborating with global research groups to drive next-generation data-intensive science and strengthen scientific innovation throughout Latin America.

SC25-NRE-129: University of Sao Paulo (USP) High Performance Dark Fiber Infrastructure
Caltech Booth #2824

Abstract: The University of São Paulo (USP), Brazil’s top-ranked public university and one of the leading research institutions in Latin America, operates across 10 campuses with its main site located in the state’s capital city. To support its growing data and research needs, USP interconnected four major campuses—São Paulo, São Carlos, Ribeirão Preto, and Piracicaba—through a dedicated dark fiber network consisting of high-capacity optical cables buried along state highways. With current inter-campus bandwidth at 100 Gbps and plans to scale to 600 Gbps by the end of 2025, USP is well positioned for advanced high-performance networking and experimentation. Its primary campus already connects to Rednesp at 100 Gbps, with an upgrade to 400 Gbps underway, aligning with Rednesp’s expanding international capacity of up to 1 Tbps to the United States. USP aims to adopt advanced services from SENSE, NRP, and GP4L and collaborate with global working groups such as GNA-G and AutoGOLE/SENSE, strengthening its role in next-generation data-intensive science and driving impactful research and education throughout Latin America.

SC25-NRE-130: A Case Study for Data-Intensive Traffic from Vera Rubin Observatory Supported by Path Aware PolKA Network
Caltech Booth #2824

Abstract: This demonstration showcases a collaborative effort to enhance the Vera Rubin Observatory Long-Haul Network (LHN) using PolKA (Polynomial Key-based Architecture) path-aware networking, designed to reliably transport 13GB astronomical images every 27 seconds across a 40 Gbps international backbone from Chile to the U.S. By enabling intelligent, fine-grained route selection based on real-time performance metrics such as latency, bandwidth, and policy constraints, PolKA’s programmable source-routing offers faster transfers, fault resilience, and improved traffic steering over high-latency links exceeding 180ms RTT. Leveraging the AmLight-ExP SDN-enabled 100G infrastructure, this demonstration highlights multi-path routing, flow isolation for Rubin science traffic, and rapid failover visualization — advancing the development of scalable, path-aware networks engineered for the next generation of data-intensive science.

SC25-NRI-132: AmLight 2.0: Flexible control, deep visibility, and programmability @ Tbps!
SCinet25 Support

Abstract: The AmLight network uses optical spectrum and leased capacity to build a reliable, leading-edge network infrastructure for research and education. Its purpose is to enable research and collaboration between the U.S., Latin America, the Caribbean, and South Africa. AmLight supports high-performance network connectivity required by international science and engineering research and education collaborations involving the National Science Foundation (NSF) research community.

In 2023, AmLight deployed its current custom-made Software-Defined Networking (SDN) programmable fabric to enable network experimentation and flexible forwarding pipelines. AmLight is fully controlled by AmLight’s SDN controller named Kytos-ng [https://kytos-ng.io]. 

In 2024/2025, AmLight expanded its footprint to Argentina, with a 400G optical channel, and to Brazil, with the amazing 1.1Tbps capacity over its optical spectrum on the Monet submarine cable. With more than 1Tbps of capacity between the South and North hemispheres and more than 24 long-haul 100G links connecting 12 sites in the U.S, Latin America, and South Africa, AmLight offers new opportunities to the academic community, such as dynamic provisioning, complex pathfinding, per-packet network programmability and telemetry, integration with academic orchestrators, such as AutoGOLE/SENSE, and testbeds, such as FABRIC and RARE.