Imagine your datacenter's network as a highway system — right now, every car has to stop at toll booths (electronic switches) even when lanes are empty. NEPHELE built an optical bypass that lets data travel on light beams instead, skipping the bottlenecks entirely. The system automatically rearranges these light paths based on what applications actually need at any given moment. The result is a datacenter network that grows smoothly as you add more servers, uses dramatically less electricity, and responds to workload changes in real time.
Datacenter traffic is growing at double-digit rates every year, pushing conventional electronic network architectures to their limits in power consumption, cost, and latency. As datacenters move toward disaggregated compute and storage, the bandwidth and latency demands become even more severe. Operators need a network that scales without burning through electricity budgets or requiring constant expensive upgrades.
The project built and demonstrated fully functional prototypes: a 400Gb/s hybrid Top-of-Rack switch, a 50Gb/s electronic-optical smart Network Interface Card, and a fast optical pod switch, all orchestrated by a custom SDN controller with application-aware interfaces. These were validated end-to-end in both lab and operational environments for intra- and inter-datacenter connectivity.
If you are a datacenter operator dealing with exploding traffic growth and unsustainable power bills — this project developed a hybrid electronic-optical network architecture that saves up to 94% power and 30% cost while scaling linearly with the number of hosts. The 400Gb/s Top-of-Rack switch and 50Gb/s smart Network Interface Card were prototyped and demonstrated in operational environments.
If you are a telecom provider managing traffic between multiple datacenters with strict Service Level Agreements — this project built an SDN-controlled inter-datacenter connectivity system that dynamically allocates bandwidth based on migrating virtual machines and existing SLAs. It was demonstrated in an operational environment covering both intra- and inter-datacenter scenarios across telecom and datacenter operator domains.
If you are running high-frequency trading platforms or real-time analytics on disaggregated compute and storage infrastructure — this project developed a network with low latency and quasi-deterministic Quality of Service that consolidates compute and storage networks over a single Ethernet optical TDMA network. The hardware-level dynamic reconfigurability was validated in both lab and operational environments.
The project reports up to 30% cost savings compared to conventional datacenter network architectures. The EU invested EUR 3,053,636 across 7 partners over 3 years to develop the prototypes. Actual deployment costs would depend on datacenter scale, but the architecture is designed to scale linearly with hosts, avoiding the exponential cost curves of traditional electronic switching.
The architecture was specifically designed to scale linearly with the number of datacenter hosts, addressing the 2-digit traffic growth challenge. Prototypes included a 400Gb/s hybrid Top-of-Rack switch and 50Gb/s smart Network Interface Card. The system was demonstrated in an operational environment for both intra- and inter-datacenter scenarios.
The project was funded under Horizon 2020 RIA (Research and Innovation Action), meaning IP typically stays with the consortium partners who created it. With 4 industry partners including 2 SMEs across 5 countries, licensing discussions would need to go through the consortium. Based on available project data, specific patent filings are not detailed.
NEPHELE was designed with integration in mind — it uses standard Ethernet granularity and consolidates compute and storage over a single network. The SDN controller provides a northbound API for applications and a southbound interface that abstracts the physical layer, allowing it to work alongside existing infrastructure while enabling hardware-level reconfigurability.
Two key demonstrations were completed: a physical layer validation of slotted operation and end-to-end connectivity in a lab environment, and a full inter- and intra-datacenter demonstration in an operational environment. Fully functional network subsystems were prototyped including the 400Gb/s Top-of-Rack switch, 50Gb/s smart NIC, and fast optical pod switch.
The project developed its own SDN controller with standard northbound and southbound interfaces. The northbound interface provides an Application Programming Interface that links application requirements to network control, enabling what the project calls Application Defined Networking. Integration with existing SDN ecosystems would require interface adaptation.
The NEPHELE consortium brings together 7 partners from 5 countries (Germany, Greece, Israel, Italy, UK) with a strong industry tilt — 4 out of 7 partners are from industry (57% industry ratio), including 2 SMEs. This is a healthy mix for technology transfer: the academic and research partners developed the science while industry partners ensured the prototypes meet real-world datacenter requirements. The coordinator is a Greek research institute (EREVNITIKO PANEPISTIMIAKO INSTITOUTO), and the geographic spread across major European tech markets plus Israel suggests access to diverse datacenter ecosystems and telecom operators for validation.
The coordinator is EREVNITIKO PANEPISTIMIAKO INSTITOUTO SYSTIMATON EPIKOINONION KAI YPOLOGISTON based in Greece. SciTransfer can facilitate an introduction to the research team.
Want to explore licensing this optical datacenter technology or connecting with the NEPHELE team? Contact SciTransfer for a tailored introduction and technical briefing.
