Igal Figlin

Midmarket customers need simplicity, not hyperscale complexity. 🧩 While AWS, Azure, and Google Cloud were built for global scale, they often fall short for businesses that value clarity, control, and real support. This white paper breaks down the key challenges—like opaque pricing, overengineering, and vendor lock-in—and shows why more partners are choosing CaaB. ✅ Transparent pricing ✅ Turnkey disaster recovery ✅ No lock-in Download the white paper to explore how a purpose-built platform like CaaB - Cloud as a Business helps reduce risk, not shift it.

1

1 Comment

With AI adoption and deeper integration into enterprise products, the world moves toward being collectively intelligent. First movers will have a distinct advantage. Fully backward-integrated AI companies are best positioned to win on economics.

1

The more autonomous AI becomes, the more binaries it will create. The more binaries created, the more security is required at the gate, and the more governance is needed before deployment to production. 🐸 https://bit.ly/4seEMPq #JFrog #LiquidSw

5

Have you heard of the llm-d project? It improves LLM performance and drives down costs of inference serving on Kubernetes. It's based around a few key projects including vLLM and kgateway. If you want a deep dive into how it works with a detailed step-by-step "packet flow" through llmd, take a look at this blog I wrote recently. 👉 https://lnkd.in/g2mbEh3m

44

3 Comments

People often ask me about the differences in architecture between Amazon Dynamo (the 2007 SOSP paper), DynamoDB (the AWS serverless NoSQL database), and Aurora DSQL (the AWS serverless SQL databases). I captured the key points on durability, high availability, programming model, and trade-offs in my blog: https://lnkd.in/gkMG_tVD TL;DR: Despite the shared name, Dynamo and DynamoDB have little in common architecturally, and make very different trade-off decisions.

631

8 Comments

Today in Satya's keynote at #MSBuild, we announced Copilot Tuning. Super excited to share the first version of the product we have been working on since late last summer. Existing enterprise applications use large models that have been trained on all the data on the Internet, but not enterprise data. Imagine if enterprises (from manufacturers, to financial companies, to energy companies, to agricultural firms) could customize Copilot with models fine-tuned with their data, such that the Copilot talks like them, thinks like them, and works like them. And all of this in a secure, tenant-isolated way, so that the fine tuned models are only available to people who had access to the underlying data, and turnkey and self-serve, so that you don’t need a data science degree to create fine-tuned, customized models. It can be built by subject matter experts without requiring an army of data scientists. Huge congratulations to the entire team that has been working passionately to build this powerful, industry leading capability, in such a short time. And to our early partners: EY, McCarthy Tétrault, and Land O'Lakes, Inc.. Learn more in this blog about the details of the program and the innovations across Engineering, Product, Business, and Science (read the ICML'25 paper for one of these)! https://lnkd.in/giuvHJAb #CopilotTuning

801

58 Comments

Memory is one of the most expensive resources in the data center. With cloud and edge workloads pushing the limits, optimizing memory is key to lowering TCO and boosting performance. 🚀 Our new paper shows how Intel® In-Memory Analytics Accelerator (IAA) + zswap unlock efficient memory-tiering for next-gen environments. 🔍 Read more: https://lnkd.in/gnrZuASW Memory-tiering in action: With growing demands in cloud and edge computing, memory optimization is critical. 🔹 Memory-tiering—with zswap as a key technique—uses compression to expand effective memory capacity and reduce Total Cost of Ownership (TCO). The Intel® In-Memory Analytics Accelerator (IAA), integrated into Intel® Xeon® 6 processors (P-cores), accelerates compression with parallel processing. Paired with zswap, a Linux kernel feature that compresses memory pages in RAM to avoid slow disk swapping, it delivers lower latency and higher efficiency. 📑 In the paper, we dive into: ✅ Faster performance with IAA-accelerated compression ✅ Lower TCO through smarter memory utilization ✅ Scalable solutions for memory-constrained environments A big thanks to all the collaborators and mentors (Kanchana Sridhar,Rakib Al-Fahad, PhD,Vinodh Gopal,Wajdi Feghali,Ryan Saffores,Mark Adams) for their support and guidance 🙏 💬 How are you tackling memory optimization in your memory-constrained environments? Looking forward to hearing thoughts and sparking more discussions on making memory go further in modern data centers! #IntelIAA #MemoryTiering #zswap #DataCenter #CloudComputing #EdgeComputing #Intel #IntelBusiness

94

3 Comments