Zach S.

I'm pleased to share some exciting updates before I begin my extended leave. 1- Graviton R8gn instances are now generally available 🚀 Featuring the latest 6th generation AWS Nitro Cards, and offering up to 600 Gbps network bandwidth, the highest network bandwidth among network optimized EC2 instances. Full announcement here: https://lnkd.in/dVvZpuzc 2- Graviton R8gb instances are now generally available 🥳 With up to 150 Gbps of EBS bandwidth, these instances offer higher EBS performance compared to same-sized equivalent Graviton4-based instances. Full announcement here: https://lnkd.in/d4Ah3J83 3- We have announced the general availability of EC2 instance attestation, a new capability that lets customers verify their EC2 instances are running exactly the software they created and approved. 🔐 EC2 instance attestation is powered by Nitro Trusted Platform Module (NitroTPM) and Attestable Amazon Machine Images (AMIs). Customers can build an AMI that includes a cryptographic measurement representing all the contents of that AMI. Using NitroTPM, customers can then verify whether a target EC2 instance has the same measurement as the reference measurement generated by the AMI. EC2 instance attestation integrates with AWS Key Management Service (KMS), allowing customers to restrict key operations to instances that pass specific attestation conditions. EC2 instance attestation is available in all AWS Commercial Regions, including the AWS GovCloud (US) Regions. Full announcement here: https://lnkd.in/dww6i8fs Stay tuned for more exciting updates.

3

Infrastructure is capital-intensive and long-lived. Once a network decision is made, it locks in cost structures, timelines, and operational constraints for years. That’s why deployment speed and flexibility matter just as much as raw performance. Before co-founding Dayton Photonics, I kept seeing the same pattern: demand was there, funding was often there, but projects slowed down in execution. Permitting, trenching, spectrum coordination, and construction timelines pushed real connectivity months or years out. In the meantime, operators carried the pressure. We focus on high-capacity free-space optical communication because it addresses a specific operational gap. There are scenarios where fiber is ideal, and there are scenarios where waiting on fiber creates unnecessary delay. Having an additional tool changes planning conversations. It creates redundancy options. It improves speed to revenue. It reduces certain regulatory friction points.

26

For those pushing the boundaries of performance, a critical bottleneck often lies not within CPU cycles, but in memory bandwidth utilization. While cache hits are vital, many advanced applications remain memory-bound due to inefficient data access patterns. Random memory access severely hampers modern CPU architectures, which are designed for speculative execution and prefetching based on predictable, sequential data streams. When algorithms necessitate constant, unpredictable jumps in memory, the memory controller and CPU prefetchers are rendered ineffective, leading to significant stalls and underutilized bandwidth. Strategically re-designing data structures and algorithms to promote *contiguous memory access* can fundamentally transform performance. This architectural shift enables optimal memory streaming, allowing prefetchers to operate at peak efficiency and often yielding 2x-5x throughput improvements in computationally intensive, data-bound scenarios. This is a non-trivial but highly impactful optimization for true high-performance computing. #PerformanceEngineering #HPC #SystemDesign #Optimization #MemoryManagement #AdvancedCoding

1

I've watched our industry churn out an impressive number of AI-powered Network Operations Proofs of Concept over the past few years. Don't get me wrong, many of these projects are technically brilliant ... But … are we actually creating value? Our goal should be getting closed loops to work in real environments, not adding another PoC to the collection. In my view, we risk wasting time and resources, getting lost in the details of a PoC, sometimes “distracted” by the wow effect of a successful AI model, rather than focusing on end-to-end operational loops. Autonomous networks don't emerge because we've proven an AI model works, a data pipeline runs, or an automation script executes. They come to life when closed loops operate reliably in real environments - with clear governance, defined accountability, and genuine operational ownership. At some point, we need to make the leap: stop proving the technology works and start deploying actual systems. In my experience, one small real-world closed loop teaches you more than dozens of isolated POCs. I'm curious to hear other perspectives on this: are POCs still the right approach to accelerate autonomy, or are they sometimes holding us back? Giovanni Catalano, Andrea Sannino, Luca Mastrota, Luca Maragnani, Alberto Licciardello, Francesco Maria Magnini, Max Gasparroni, Alessandro Magnino, Giovanni Germani, Marco Rossetti, PhD #AutonomousNetworks, #TelcoTransformation, #AIinNetworks, #NetworkAutomation

137

10 Comments

Welp… AWS me-south-1 could’ve been a bust but I decided to go multi-region from day one. 🤯 Our workloads normally run in me-south-1, but we can instantly flip to eu-central-1 with zero manual work and zero customer impact. A real conflict in the region recently tested that setup, and MAKSAT still had 0 downtime. Honestly, sometimes the best feeling as a CTO is when no one notices anything… because everything just keeps working. 

12

This morning’s deep dive into Quality of Service (QoS) was a clear reminder of how essential it is to maintain performance and reliability in networked systems. what i deducted · QoS isn’t just about bandwidth—it’s about intelligently prioritizing traffic to ensure critical applications run smoothly, especially under load. · Mechanisms like traffic classification, queuing, and policing help manage delay, jitter, and packet loss. · Without QoS, voice, video, and real-time data can degrade, impacting user experience and business operations. In today’s hybrid and cloud-heavy environments, implementing thoughtful QoS policies is no longer optional—it’s foundational for consistent service delivery. What strategies are you using to manage network performance? #QualityOfService #QoS #Networking #ITInfrastructure #NetworkManagement #TechLearning

1

AT&T rebuilt its AI orchestration to handle 8 billion+ tokens a day, cutting costs by 90% and tripling throughput by shifting to a multi‑agent architecture powered by small language models and LangChain. A major accelerator was its deep collaboration with Microsoft Azure, which underpins the new Ask AT&T Workflows platform — a drag‑and‑drop agent builder now used by 100,000+ employees. The partnership enables secure, scalable automation grounded in AT&T’s own data, while Azure’s search and model ecosystem keeps the stack flexible and future‑ready. A strong example of how enterprise–cloud partnerships can unlock real, measurable AI impact. https://lnkd.in/enja9HZg

21

1 Comment