Tom Dean’s Post

Today, we're happy to announce the release of Linkerd 2.17, a new version of Linkerd that introduces several major new features to the project: 🔹 egress traffic visibility and control, 🔹 rate limiting, 🔹 federated services, 🔹 OpenTelemetry support for distributed tracing. Federated services is a powerful new multicluster primitive that combines services running in multiple clusters into a single logical service. Read the full blog post to learn more: https://hubs.ly/Q02-xrBs0 #Linkerd #ServiceMesh #OpenSource #CNCF

Today, we're happy to announce the release of Linkerd 2.17, a new version of Linkerd that introduces several major new features to the project: 🔹 egress traffic visibility and control, 🔹 rate limiting, 🔹 federated services, 🔹 OpenTelemetry support for distributed tracing. Federated services is a powerful new multicluster primitive that combines services running in multiple clusters into a single logical service. Read the full blog post to learn more: https://hubs.ly/Q02-xrQ90 #Linkerd #ServiceMesh #OpenSource #CNCF

Linkerd 2.17 is out! Check out the awesome new features! Buoyant Linkerd #servicemesh #kubernetes #k8s #cloud #cloudnative Cloud Native Computing Foundation (CNCF) #opensource #linux #tomsentyou

🌟 Day 350/365: #365DaysOfDSA Challenge 🌟 Today, I tackled the Pairs With Given XOR problem, adding another exciting milestone to my DSA journey! 🚀 This problem sharpened my understanding of XOR operations and efficient pair-finding techniques. It reminded me how the elegance of bitwise operations can simplify complex challenges. With only 15 days left in this year-long journey, the finish line is in sight, but the learning and growth continue stronger than ever. 💡 ✨ Key takeaway: Simplify the approach by leveraging the properties of XOR for faster computations. If you're also on a problem-solving streak or exploring XOR magic, let’s connect and discuss! 🔗 Discord Server: https://lnkd.in/gstPhXQX Scaler #DataStructure #Algorithms #CodingChallenge #ProgrammingJourney #CodeNewbieLeetCode LeetCode #AchievementUnlocked #scalerdiscord #codewithscaler #365daysofcodescaler #365daysofcode #CodingJourney #DSA #ProblemSolving #MilestoneAchievement #KeepLearning

🎫 Join Dr. Jochen Papenbrock of NVIDIA for Building and Running a Scaling, Enterprise FinLLM Using OSS and Accelerated Computing at our Open Source in Finance Forum on 26th June in #London 🔗 https://bit.ly/3wlV2a1 Talk Abstract In this talk, Jochen presents various OSS tools from NeMo LLM Framework that can be leveraged on an accelerated computing platform for building scaling, enterprise-ready (Financial) LLMs with outstanding performance, accuracy, stability and guardrailed trustworthiness. Initial step is to select one of the best community-build LLMs like Llama, Falcon, Mixtral or Nemotron. Next steps is to use tools for model customization, retrieval augmented generation, and guardrailing. Then, the model will be deployed with an inference platform that is optimized for throughput and scale. Input data can be tokenized, curated, and deduplicated at scale. 📆 OSFF Schedule: https://bit.ly/3QvvEp4 Event presented by FINOS & The Linux Foundation #OSFF2024 #OSinFinance #opensource #fintech #financialservices #techevent #conference #opensourcecommunity

I went to SW2con this week - a conference about the bleeding edge in software development. Two things I learned: 1) East-west traffic in data centers is growing significantly due to AI training workloads. At the most basic level, GPUs are sending and receiving large subsets of parameters back and forth during the training processes. This causes large amounts of packet loss, which results in longer training times. This is, of course, expensive. Hedgehog is a neat company solving this: https://githedgehog.com/ 2) A presentation by Github highlighted the importance of UX, context, and prompt engineering in building an experience that developers love in CoPilot. On UX, research by Github showed that code completion needed to run in under 150 ms to maximize user engagement. On context, Github found the unique combination of given the LLM context but which avoided the crawling of what otherwise might be sensitive information (hint: look at open tabs in the editor). On prompt engineering, it took Github over 6 months just to get all the prompting right to provide helpful results.

Want to know what Pi Squared is all about? Watch this great presentation by our CTO Xiaohong Chen where he shared how Proof of Proof enables any developers for any programming languages to build in Web3!

🔔 Adaptive parallelism through #OpenMP within the #LLVM framework Adrián Munera Sánchez, representing our project partner Barcelona Supercomputing Center, will showcasing the innovative research on "Adaptive parallelism in OpenMP through Dynamic Variants" at ISC High Performance. Our approach includes two key contributions: 1️⃣ Automatic instrumentation of variants, facilitating the collection of system metrics during a warm-up phase. 2️⃣ Implementation of runtime heuristics designed to effectively choose the optimal variant, taking into account user preferences and system conditions. These contributions are seamlessly integrated into the LLVM 17.0.0 compilation framework. Save the dates: 📆 16th May: Join us for a presentation at the #C3PO workshop during #ISC High Performance. 📆 17th June: Stay tuned for the release of the camera-ready version and public accessibility of our gitlab repository. Shaping together the future of parallelism!

Day-24 of #100DaysOfRTL Challenge: PIPO Shift Register Today, I’m diving into the Parallel-In Parallel-Out (PIPO) shift register, a fundamental component in digital systems for holding and transferring parallel data. It is also called as storage register or buffer register. The applications of the PIPO shift register include the following: >>The PIPO shift register is mainly used to add time delay to digital circuits. >>Shift registers are used for converting data and also to shift the data from left to right and right to left. >>It is used for storing the data. This is a temporary storage device where both data loading & retrieval processes take place within the parallel mode. Checkout my GitHub Repository: https://lnkd.in/g8aVHwn6 #DigitalDesign #Verilog #RTL #ShiftRegister #PIPO #100DaysOfRTL #LearningJourney

This is Chat GPT’s interpretation of #Flape. Flash to tape. But Flape is so much more. Having a super fast NVMe frontend based on #objectstorage with ActiveScale, based on file with StorNext or backup centric with our DXi, off-loading or migrating to super economic and green tape storage with our ActiveScale Cold Storage solution or ‘just’ StorNext vaulting, all within one bucket or namespace! The best of both worlds seamlessly made possible with Quantum.