Adam Kalisz’s Post

eBPF: The Kernel’s Superpower Legacy tools like iptables can’t keep up with Kubernetes-scale environments. Regulatory pressure demands zero-trust. Hyperscalers have already made the shift. eBPF changes the game: programmable networking, real-time observability, and security at kernel speed. In my latest post, I break down how it works and why enterprises need it NOW, not later. https://lnkd.in/ehGk6GYx #eBPF #Kubernetes #OpenShift #CloudNative #PlatformEngineering

Big Data meets Big Compute! For throughput computing, Through the Pipelines Thank you Elza Wong for organizing this webinar based on much effort by CXL Consortium member companies! During the webinar this morning, Anil Godbole, Geof Findley, Jianping(JP) Jiang, and I presented practical examples of how CXL is mitigating system bottlenecks for traditional database and search engines which feed the emerging AI/ML LLMs. We discussed attributes of CXL specification as an efficient interconnect protocol that industry partners are putting to work to deliver higher memory capacity and bandwidth. As we shared system performance results, we also demonstrated how memory pooling/sharing of disaggregated memory help increase compute and memory utilization through various pipeline stages of emerging AI/ML workloads. The essential elements for this efficient interconnect include ·      Shared memory semantics & direct Load/Store ·      High-throughput, low-latency Fabrics with efficient protocol ·      Predictable response time (low jitter) ·      Fine-granularity flow-control ·      Memory ordering guarantees (ensures correctness in distributed long-context generation & training) ·      Reliable, loss-less transfer (sustains throughput under load avoids costly retransmission cycles) ·      Confidential Compute & Security (protects data at-rest and in-motion for enterprise-grade privacy) CXL is finding its path not only within the traditional general-purpose computing and scale-up architecture but also as a major contributor to collecting, organizing, and feeding data for large AI/ML workloads. CXL Consortium Open Compute Project Foundation

Following the CXL Consortium webinar, we presented several practical use cases at OCP Global Summit for CXL technology as it is gaining momentum with commercially available silicon and software! Demonstrated how CXL enables traditional search and database techniques feed Big Data to Big Compute for the emerging AI/ML’s throughput computing via RAG pipelines End to end memory fabric orchestration How CXL plays a role in Super Clusters via bridges between GPU scale-up fabrics and the standard CXL Fabric Stretching the memory hierarchy pyramid to place more memory at the top layers of the bandwidth tiers with CXL and optical interconnects Held an interconnect panel of experts and described how the currently available CXL solutions address the emerging needs for Bulk memory while the emerging scale-up fabrics will commingle to serve future GPU-to-GPU interconnect needs The role of short-reach optical interconnects to increase system level flexibility to stretch CXL domain of influence and service Open Compute Project Foundation CXL Consortium

Now deploy even bigger workloads on containers - up to 4 vCPUs, 12 GiB memory, and 20 GB disk per instance. With the entire Developer Platform ecosystem, you can build and run full-stack applications globally, closer to your users on Region Earth - "Network is the computer". #cloudflare #containers https://lnkd.in/gNQJUYdR

Most AI models are smart thinkers but poor connectors. MCP servers change that by becoming the "USB port" of intelligent applications. Here’s what they are; and why they matter. #MCP #ModelContextProtocol #AIIntegration #TechExplained #NeoQuant #FutureOfAI

Verifiable Blind Observable Estimation Achieves Composably Secure Quantum Advantage for Near-Term Tasks Researchers have developed a new method for verifying computations performed on remote servers, guaranteeing the accuracy of estimated results even when the server is untrusted, and doing so with minimal additional computational cost. #quantum #quantumcomputing #technology https://lnkd.in/eZve8yQC

📢Providing high quality compute systems is our DNA at Fsas Technologies. The latest introduction of the #PRIMERGY Server M8 Series is combining high quality, latest technology and our sustainability expectations: The 🆕 standard for trusted data centers! We're proud to launch our new PRIMERGY Server M8 Series, featuring the versatile RX2540 M8 and RX2530 M8. Built on a legacy of quality, these servers are engineered for the reliability and adaptability demanded by modern enterprise IT. Powered by the latest Intel® Xeon® 6 Processors with P-cores, the PRIMERGY Server M8 series offers substantial enhancements in compute power and memory bandwidth, ensuring stable performance for your critical applications and AI workloads, day in and day out. 👉 Discover the foundation of a dependable data center: https://okt.to/FqAlaj #PRIMERGYM8 #ServerLaunch #ReliableIT #DataCenterStandard #IntelXeon

Building High-Performance TCP Servers in Go: Custom Binary Protocols for Ultra-Low Latency Applications As a best-selling author, I invite you to explore my books on Amazon. Don't forget to follow me on Medium and show your support. Thank you! Your support means the world! In my work with network systems, I've often encountered situations where standard HTTP protocols introduce unnecessary overhead. For latency-sensitive applications like financial trading platforms or real-time gaming servers, every microsecond counts. That's why I turned to building custom TCP servers in Go, leveraging its powerful concurrency model and efficient runtime to achieve performance levels that HTTP-based solutions struggle to match. Go's goroutines and channels provide an excellent foundation for handling thousands of concurrent connections. The language's built-in networking packages offer robust primitives while leaving room for optimization. When I started designing this server, I focused on three key areas: efficient connection management, a lean binary protocol, and zero-copy data processing to minimi https://lnkd.in/gURTFQ8G

🔍 What happens when you ping google.com? (The journey of a simple command) Seems simple, right? Actually, it's a masterclass in networking. Here's the 7-step journey: 1️⃣ DNS Resolution • Check local cache first • Query DNS server: google.com → 172.217.164.110 • Time: ~20ms 2️⃣ ARP (Address Resolution Protocol) • Find router's MAC address • IP → Physical address mapping 3️⃣ ICMP Packet Creation • Internet Control Message Protocol • Ping uses ICMP Echo Request 4️⃣ Routing Through Internet • Packet hops through 10-15 routers • Each router decides next best path • BGP protocol guides the journey 5️⃣ Google's Load Balancer • Receives packet at edge server • Decides which data center to use 6️⃣ ICMP Echo Reply • Google sends response back • Same route, reverse direction 7️⃣ Round Trip Complete • Your terminal shows: "64 bytes from google.com: time=23ms" 💡 Why this matters: • Network debugging skills • Understanding latency sources • Designing distributed systems 23ms seems instant, but it traveled thousands of miles! What networking concept confused you the most? Let's discuss! #Networking #TCP #DNS #SystemDesign #SoftwareEngineering #Internet

Blind deployments are risky. Want to know exactly what your server’s doing in real time? Our new guide shows you how to monitor your Webdock server with Prometheus, Grafana, and Node Exporter - plus set up alerts with Alertmanager - so you catch issues before they worsen. Key steps covered: •Install Node Exporter and expose metrics (port 9100)  •Configure Prometheus to scrape your server metrics (e.g. CPU, memory, disk)  •Set up Grafana dashboards to visualize performance over time  •Use Alertmanager with custom alert rules (e.g. high CPU, low memory, disk full) to trigger notifications  You’ll get visibility, not just logs - and peace of mind when load or anomalies appear. 🔗 Full setup walkthrough: https://lnkd.in/e_q_MvBh #Webdock #Monitoring #Prometheus #DevOps #Sysadmins #CloudHosting