SRS’ Post

Can two tenants - companies, operators, or governmental agencies - safely share the same satellite? 🛰️ That's a challenging question, as low Earth orbit (LEO) satellites reshape the entire network topology every few minutes, and scarce onboard computing resources force different tenants (i.e., slices) to share the same virtual network functions. This opens the door to cross-slice side-channel risks that evolve with every time window. Our new preprint tackles this head-on: ✅ Treats cross-slice co-location risk as a measurable objective (grounded in ISO/NIST) ✅ Separates avoidable migrations from the ones orbital motion forces ✅ Uses an ADMM decomposition that stays solvable where the exact solver times out. 📄 https://lnkd.in/g6nzcxhv Result: near-zero co-location risk, full delay compliance, and feasibility where the monolithic method breaks. How would you model the security-vs-continuity trade-off when the infrastructure itself is moving? Curious to hear from the NFV, slicing, and non-terrestrial-network folks. Grateful to my co-authors, Dr. Wael Jaafar, Prof. Sami Muhaidat, and Prof. Halim Yanikomeroglu, for the collaboration. #6G #LEOSatellites #NetworkSlicing #NFV #NonTerrestrialNetworks #NetworkSecurity

📡 Week 5: point an ordinary 5G phone at a satellite, and two things break instantly. The first four weeks of this series were about the satellite side — link budgets, real passes, scheduling, some ML. This week I came at it from the side I spent years working in: 5G itself. If you pointed a normal 5G handset at a satellite instead of a cell tower, what would actually break? Two things — and both come from assumptions baked into the 5G NR standard: ⏱️ 1. Timing advance 5G tells distant handsets to transmit slightly early so signals arrive aligned. But the largest standard 5G NR timing-advance command tops out at ~667 microseconds — about 100 km. A LEO satellite is 500–2000 km away. → Running it on a real Starlink pass over Aberdeen: the timing advance needed hit ~9,700 microseconds — more than 14 times the standard 5G limit — with the satellite still 1,386 km away at closest approach. Across the whole pass, essentially NONE of it fits within standard 5G timing. 📶 2. Doppler shift A cell tower stands still, so the frequency you receive is the frequency it sent. A satellite moving at 7.5 km/s shifts the carrier — up as it approaches, down as it recedes, through zero overhead. At S-band (~2 GHz, the direct-to-handset band) that's tens of kHz; at Ku-band, a few hundred. It has to be tracked continuously, not corrected once. 🔍 The point isn't "these numbers are big." It's that they're big relative to a specific line the 5G standard actually draws — the 667 µs / ~100 km limits are real 3GPP values, and a satellite blows straight through them. That's exactly why 3GPP's NTN work (Release 17+) exists: GNSS-assisted timing, extended timing advance, Doppler pre-compensation. The red line on my timing plot isn't decoration — it's where standard 5G stops working, and the entire real pass sits above it. Part of a 12-week public build connecting AI, 5G and satellite networks (NTN) — every week shipped, tested, and explained in plain terms. #5G #NTN #SatelliteCommunications #3GPP #RFEngineering #Python #6G

We should have our partners summarize our offerings more often! 😄 Keysight Technologies wrote up a great article on how to go from code to constellation by connecting #Keysight hardware to #COSMOS. They said: “COSMOS supplies the control layer over that hardware. It sends commands, pulls telemetry, and runs scripted test sequences over standard interfaces. Because it speaks SCPI to bench instruments and CCSDS packets to flight hardware, the same harness that drives a benchtop power supply drives the satellite. Its containerized, Kubernetes-scalable architecture means the setup built for one prototype carries through flatsat, integrated test, and operations up to a full constellation, without a ground-software rebuild at each phase.” We couldn’t have said it better ourselves! We’re proud to partner with Keysight and work directly with them to build COSMOS plugins that control the different hardware they create, no matter the phase of the #satellite build. Read the blog to learn more about our continued collaboration: https://lnkd.in/gRfRmAFR

𝗔𝗱𝘃𝗮𝗻𝗰𝗶𝗻𝗴 𝗥𝗲𝘀𝗶𝗹𝗶𝗲𝗻𝘁 𝗮𝗻𝗱 𝗦𝗲𝗰𝘂𝗿𝗲 𝗣𝗡𝗧 QASCOM has developed and successfully tested a 𝗻𝗲𝘄 𝗮𝘀𝘀𝘂𝗿𝗲𝗱 𝗣𝗡𝗧 𝗿𝗲𝗰𝗲𝗶𝘃𝗲𝗿 𝗽𝗿𝗼𝘁𝗼𝘁𝘆𝗽𝗲, combining Secure Element-based storage and cryptographic processing to support Galileo E6 range authentication and Galileo OSNMA. With the inclusion of 5G data, we support a secure User Terminal architecture, ensuring the integrity and robustness of the PNT solution, targeting critical applications. The prototype demonstrated strong resilience in complex simulated environments, including scenarios involving advanced spoofing attacks. We would like to thank Inside GNSS for featuring our work. Read the full article here: https://lnkd.in/e8yvbx-b

Volume 2 of The Substrate Report is live. Episode 1 — The Design Center Shift. AI is moving from application to architecture in the network. The AI-RAN Alliance (NVIDIA, SoftBank, Samsung, ~40 members) is standardizing how AI workloads and cellular technology share the same compute substrate. SoftBank's AITRAS is the operational proof point — the cell tower becomes an edge data center earning revenue from both mobile traffic and external AI inference, hardware utilization approaching 100%. The architectural challenge isn't bolting GPUs onto edge sites. It's co-designing hardware and software — programmable smart NICs, GPU Direct RDMA, in-line front-haul offload — so the data path actually flows. The cognitive RAN is the substrate-aware network applied at the radio access layer. It's the same architectural pattern showing up simultaneously in automotive (software-defined vehicles), energy (distributed energy resources), and edge infrastructure broadly. This is the design center shift. Volume 2 walks through it across three episodes this week. https://lnkd.in/g8tw8mWg Subscribe: https://lnkd.in/gdGRAQz5 #AIRAN #CognitiveRAN #SubstrateAware #AIbyDesign #5G #6G #NVIDIA #SoftBank #Mavenir

Almost three years ago at MWC, we were blown away by SRS's proof of concept featuring a CU/DU deployment on a Raspberry Pi—and they were equally impressed by Eridan's innovation. We partnered with SRS shortly after and successfully integrated their 5G Enterprise solution with our RU. Thanks to its absolute reliability, it quickly became the most-used gNB setup in our lab. Now, thanks to U.S. government efforts, resulting in OCUDU Ecosystem Foundation project, this technology is accessible worldwide. Paired with an Eridan RU, it promises to deliver the exact disruption that Open RAN was designed to enable! #eridan #ocudu #openran #oran

Duos Edge AI has executed a Master Services Agreement with Nistar to deploy up to 2 MW of critical IT-load capacity, along with the utility infrastructure, at Duos’ Columbus, Georgia campus. The deployment is designed to support high-density AI and high-performance computing workloads with Tier III-equivalent infrastructure, N+1 redundant power and cooling, and a Ready for Service framework built for fast installation and reliable operations. The agreement adds to the current contract for a total of 10 MW at the site, with additional capacity planned later this year. Learn more about the announcement here: https://lnkd.in/eymXJS89 #DuosTechnologies #DuosEdgeAI #DUOT #Nistar #EdgeDataCenter #AIInfrastructure #DigitalInfrastructure #HPC #GPU #TechGrowth #iMillerPR $DUOT

Duos Edge AI announces today that it has executed a Master Services Agreement with Nistar to deploy up to 2 MW of critical IT-load capacity, together with the utility infrastructure. Under the terms of the agreement, Duos will provide a fully “Ready for Service” deployment designed to support high-density AI and high-performance compute (“HPC”) workloads, enabling seamless server installation and operation for Nistar. The Nistar deployment will be located at Duos’ Columbus, Georgia campus, which has expansion capacity of up to 20 MW. Read the full announcement: https://lnkd.in/eeQ5BtqV #DuosEdgeAI #EdgeComputing #DigitalInfrastructure #AIInfrastructure #DuosTech #DUOT Duos Technologies, Inc.

New TechPost: Next Generation Port Extender (NGPE): The Right Interface Speed on the Right Box, for the AI Era As AI traffic pushes networks toward 800G and beyond, what happens to all your 1G/10G edge ports? Pankaj Kumar's latest tech post introduces Next Generation Port Extender (NGPE) — pairing an MX aggregation device with low-cost satellites (EX/QFX) via Junos Node Unifier(JNU), running on a standards-based EVPN-VXLAN fabric with zero manual fabric engineering. Right interface speed, right box, one logical router. A great read for network engineers building AI-era edge networks. https://lnkd.in/gGMDM-vz #HPENetworking #NGPE #MX

More missions require more than high-performance onboard computing. They require confidence that the hardware can be delivered at scale and on time. According to Novaspace's 2026 forecast, 𝗻𝗲𝗮𝗿𝗹𝘆 𝟭𝟳,𝟬𝟬𝟬 𝘀𝗺𝗮𝗹𝗹 𝘀𝗮𝘁𝗲𝗹𝗹𝗶𝘁𝗲𝘀 𝗮𝗿𝗲 𝗲𝘅𝗽𝗲𝗰𝘁𝗲𝗱 𝘁𝗼 𝗯𝗲 𝗹𝗮𝘂𝗻𝗰𝗵𝗲𝗱 𝗯𝗲𝘁𝘄𝗲𝗲𝗻 𝟮𝟬𝟮𝟲 𝗮𝗻𝗱 𝟮𝟬𝟯𝟱, driven by growing demand for constellations in critical applications, from Earth Observation and secure connectivity to national defense and in-orbit services. That means satellite manufacturers won't just need better technology, they'll need suppliers who can keep pace with demand at volume. At KP Labs, we've been scaling up accordingly. Following an increase in commercial orders, 𝘄𝗲'𝘃𝗲 𝘁𝗿𝗶𝗽𝗹𝗲𝗱 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗼𝗳 𝗟𝗲𝗼𝗽𝗮𝗿𝗱 𝗗𝗣𝗨 𝗼𝘃𝗲𝗿 𝘁𝗵𝗲 𝗽𝗮𝘀𝘁 𝘀𝗶𝘅 𝗺𝗼𝗻𝘁𝗵𝘀. Last week, 𝘄𝗲 𝗱𝗲𝗹𝗶𝘃𝗲𝗿𝗲𝗱 𝗟𝗲𝗼𝗽𝗮𝗿𝗱 𝗗𝗣𝗨 𝘂𝗻𝗶𝘁𝘀 𝘁𝗼 𝗮𝗻𝗼𝘁𝗵𝗲𝗿 𝘁𝘄𝗼 𝗰𝗼𝗺𝗺𝗲𝗿𝗰𝗶𝗮𝗹 𝗽𝗮𝗿𝘁𝗻𝗲𝗿𝘀, 𝗼𝗻 𝘀𝗰𝗵𝗲𝗱𝘂𝗹𝗲, proof that this scale-up isn't just theoretical. Leopard DPU is our 𝗳𝗹𝗶𝗴𝗵𝘁-𝗽𝗿𝗼𝘃𝗲𝗻 𝗗𝗮𝘁𝗮 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 𝗨𝗻𝗶𝘁, built for payload acquisition, onboard processing, AI inference, and mass data storage on nano and microsatellites. It's designed as an 𝗶𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻-𝗿𝗲𝗮𝗱𝘆 𝘀𝘂𝗯𝘀𝘆𝘀𝘁𝗲𝗺, so mission teams can move straight from delivery into integration, without added development effort or delay. Increased production capacity means predictable 𝗹𝗲𝗮𝗱 𝘁𝗶𝗺𝗲𝘀 𝗮𝗻𝗱 𝗼𝗻-𝘁𝗶𝗺𝗲 𝗱𝗲𝗹𝗶𝘃𝗲𝗿𝗶𝗲𝘀, so mission teams can plan with confidence. For satellite manufacturers, reliable delivery is becoming just as important as computing performance. We're building both. #KPLabs #missioncompete #LeopardDPU