Luis Recio

Just released another episode in the "Modern Database" Playlist which has 45+ database episodes. This time we are talking to Philipp Fent from CedarDB about #CedarDB and learning about the innovations and the amazing engineering behind this modern ultra fast database. Key Takeaways: - CedarDB is built from the ground up to utilize modern hardware effectively. - The system compiles SQL directly to machine code for performance. - Parallel processing is a key feature, allowing efficient use of multiple cores. - CedarDB aims to be Postgres compatible while innovating on performance. - Transactional workloads are handled efficiently without sacrificing analytical capabilities. - Data ingestion is optimized for both row-oriented and columnar formats. - The system uses optimistic concurrency control to manage write conflicts. - Query optimization leverages statistics to improve join performance. - Future developments include schema evolution and disaggregated storage. - CedarDB is designed to be flexible and adaptable for various workloads. Watch the entire episode and enjoy! Link below in the comments. Like the post and share it with your friends for the love of databases.

86

7 comentarios

Accidentally discovered a new way to work with GitHub Copilot today: write new tests based off test coverage reports. Copilot (or I) run this command to generate a report of all the uncovered lines in the current diff: pytest --cov --cov-report=xml && \ diff-cover coverage.xml --html-report coverage_report.html && \ open coverage_report.html Then Copilot reviews the report, summarizes the uncovered situations, and proposes new test cases to add, or if the code is truly unreachable, suggests deleting it. Just another reason why you should add diff-cover (or similar tool) to your projects!

115

4 comentarios

Sonnet 4.5 is special in many ways, but one I worked directly, is very exciting. Sonnet 4.5 is the first model that is aware of how much usable context it has left (test it - you can just ask it on https://claude.ai). Why is this a big deal? When models run in an agentic loop they need to decide when to stop and provide an answer. If they run out of tokens without providing an answer thats very bad. So most models have been trained to avoid that at any cost. But without knowing how many tokens are left models will cut their efforts short just to be on the safe side - we call that "agentic laziness". With Context Awareness feature models like Sonnet 4.5 always know how many tokens they have left and can use it to the fullest. You can also use this feature in a creative way: ask the model to spend no more than X thousand tokens or ask it to compact it's context (if such tool is available) before they run out of context. Read more at: https://lnkd.in/eMT3pQAh

276

3 comentarios

A couple of weeks ago, someone asked me: "Can we use LlamaIndex workflows with microservices?". Short answer: yes. Long answer: I built a small-scale e-commerce system leveraging Docker, Inc and Kafka to demo it 👉 youtu.be/gPR0pMtY--c ⭐ Star the repo: https://lnkd.in/ddhtW6-B 📚 Learn about workflows: https://lnkd.in/duRFTNxN

56

4 comentarios

𝟰 𝘄𝗮𝘆𝘀 𝘁𝗼 𝗺𝗮𝗸𝗲 𝘆𝗼𝘂𝗿 𝗖𝗜/𝗖𝗗 𝗽𝗶𝗽𝗲𝗹𝗶𝗻𝗲 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝘄𝗼𝗿𝗸 CI/CD is not just about automation. It’s about shipping with confidence, catching issues early and moving fast without chaos. Here’s how teams can get it right: 🔵 1. 𝗧𝗲𝘀𝘁 𝗲𝘃𝗲𝗿𝘆𝘁𝗵𝗶𝗻𝗴 𝗯𝘆 𝗱𝗲𝗳𝗮𝘂𝗹𝘁 Unit, integration and security tests should run automatically. If something fails, the pipeline blocks it. 🟣 2. 𝗞𝗲𝗲𝗽 𝗽𝗶𝗽𝗲𝗹𝗶𝗻𝗲𝘀 𝗳𝗮𝘀𝘁 Slow pipelines get ignored. Speed matters if you want quick feedback and real adoption. 🟠 3. 𝗦𝗵𝗶𝗳𝘁 𝘀𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗹𝗲𝗳𝘁 Scan early, manage secrets properly and bake in permission checks from the start. 🟢 4. 𝗠𝗮𝗸𝗲 𝗱𝗲𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁𝘀 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗮𝗯𝗹𝗲 The goal is safe, repeatable releases that don’t need heroics or manual fixes. Gif Credit: Chandresh Desai #cicd #devops #automation

61

1 comentario

Recently I've been getting a lot of requests for how to build an agent with MCP servers. In this article I walk through how to quickly build a ReAct agent using the Neo4j Cypher MCP server mixed with custom tools. #neo4j #langchain #mcp

53

1 comentario

When we began breaking our monolith into microservices, it felt like setting up separate departments in a company. Each service got its own domain: orders, users, payments, notifications. Small, focused, autonomous. Sounds clean, right? But then came the questions: How will services talk to each other? What happens when one crashes? How do we manage configs across environments? Where’s the log when something breaks? So we built out the microservices architecture the real one: 🔧 Service Mesh for Communication Introduced a sidecar proxy (Istio) to manage service-to-service traffic. Had retries, circuit breakers, and observability without polluting business logic. 🔗 API Gateway for the Front Door All traffic hit the gateway first. It handled auth, routing, rate-limiting so our internal services could stay lightweight. 🛢️ Dedicated Databases per Service No more shared DB nightmares. Each service owned its data, and used event-based patterns for eventual consistency. 📬 Asynchronous Messaging with Queues Some services talked via RabbitMQ or Kafka. Less coupling, more resilience and the luxury of delayed processing. 🔍 Centralized Monitoring and Logging With Prometheus + Grafana for metrics and Loki + Fluentd for logs, finally had answers to “what just happened?” 🚀 CI/CD and Containerization Services lived in Docker containers, deployed via pipelines to Kubernetes clusters. Each with independent lifecycles, auto-scaled based on demand. The result? A modular, resilient system. But also a reminder that microservices aren’t just small services, they’re a shift in thinking: ownership, observability, and operational maturity. If you’re considering microservices, don’t just ask “How do we split this up?” Ask: “Are we ready to run, monitor, and scale 20+ independent systems like a pro?” Because the moment you break the monolith, you’re signing up for distributed systems engineering. #Microservices #Architecture #DevOps #CloudNative #Kubernetes #PlatformEngineering #SoftwareEngineering #SRE #API #TechLeadership #hellodeolu #technicalmarketing

65

6 comentarios

Next week at SREcon in Dublin, several of my colleagues will present talks and lead discussions on critical aspects of reliability engineering. Pierre Theo Klein, Garrett Holthaus, and Rubén Barroso will present “STPA for Software Systems–Illuminate the Unknown Unknowns”: SREs know about some of the flaws and vulnerabilities in their systems. They might also have intuition on where to look for additional issues–"known unknowns." But what about the "unknown unknowns"–outages waiting to happen that nobody is even looking for? What if we could understand our systems enough to be able to methodically explore these complex interactions and build a comprehensive list of possible outage scenarios? In STPA, we model systems based on control-feedback loops, creating a hierarchical control structure, or HCS. In this session, we'll use a real Google system to show how an HCS can help you gain a new perspective and understanding of your system. We'll note similarities to common patterns in software design so you can start thinking about similar vulnerabilities in your own systems. Tomasz Koczorowski will present “the Bitter and the Sweet of Running a Planet-Scale Build & CI Stack at Google”, about managing Google's planet-scale Build & CI stack from an SRE perspective. He’ll explore the complexities of supporting over 100,000 monthly users and massive workloads, maintaining a 98% cache hit rate across diverse computing environments. Discover resource management strategies for doubling year-over-year growth, the application of the Pareto principle, and critical caching layers. You will gain insights into optimizing the build stack and maintaining a 24/7 service with a small SRE team through strategic planning, monitoring, and collaboration. He will present availability risks, standardization, and the pros and cons of a monolithic Build & CI stack Jennifer Petoff will lead a discussion on “Oops to Ops: Establishing SRE Practices”: Establishing SRE practices, gaining buy-in, and building momentum around reliability can be a significant hurdle for many organizations. This interactive discussion session, designed for both SRE newcomers and seasoned practitioners, will explore common challenges and effective strategies for implementing SRE principles based on our collective experience. We'll delve into practical approaches for demonstrating the value of SRE, fostering a culture of shared ownership, and overcoming resistance to change. Participants are encouraged to share their experiences, ask questions, and contribute to a collaborative exchange of ideas on how to successfully integrate SRE into their organizations and build a strong foundation for operational excellence. (continued in comments)

48

3 comentarios

🚀🚀🚀 The AWS Lambda Runtimes team is announcing 3 new runtime previews! We are thrilled to introduce three new preview OCI images: Python 3.14 Preview / Node.js 24 Preview / NET 10 Preview You can pull them today from our public ECR with the following tags: python:3.14-preview, nodejs:24-preview, and dotnet:10-preview We've created 3 dedicated GitHub issues in the aws/aws-lambda-base-images repo. We look forward to your feedback! #AWS #Serverless #Lambda #Python #NodeJS #dotNET

123

7 comentarios

Choosing a framework to measure software development can be a difficult and confusing step, but it doesn’t have to be. We've just published an article that aims to guide you in selecting the right framework for your team for right now. Looking forward to your feedback and hearing how your team and organization are approaching this topic! https://lnkd.in/eUNyBvdr #DORA #GBGB #metrics

38

𝗧𝗵𝗿𝗶𝗹𝗹𝗲𝗱 𝘁𝗼 𝗕𝗲𝗴𝗶𝗻 𝗠𝘆 𝗝𝗼𝘂𝗿𝗻𝗲𝘆 𝗮𝘁 𝗛𝗮𝗿𝗽𝗲𝗿! After just one week at Harper, I'm amazed by how much I've learned—and even more excited about what's ahead. So many backend stacks are fragmented—databases, caches, queues—each adding latency and complexity. Harper takes a bold approach: unifying these into one high-performance runtime. Application logic, caching, messaging, and storage all run in-process, delivering speed and simplicity that change the game. For anyone building web experiences where performance and reliability matter, Harper's value is clear: 🚀 𝗕𝗹𝗮𝘇𝗶𝗻𝗴 𝗳𝗮𝘀𝘁: backend ops in the sub-millisecond range. ⚡ 𝗨𝗻𝗶𝗳𝗶𝗲𝗱 𝗿𝘂𝗻𝘁𝗶𝗺𝗲: no patchwork of systems to manage. 🌍 𝗘𝗳𝗳𝗼𝗿𝘁𝗹𝗲𝘀𝘀 𝘀𝗰𝗮𝗹𝗲: distribute globally with ease. I'm excited to help customers architect with Harper. If you're curious, let's connect—or come find us at JSConf North America (https://lnkd.in/dq3kKnr9) this October. I'd love to share more and hear about the challenges you're solving. • Harper Website: https://lnkd.in/d77CXb55 • Harper GitHub: https://lnkd.in/d4ujxCBG #webperformance #ecommerce #webperf #seo #gratitude

54

6 comentarios

With a better understanding of Cilium, I can now break down one of the slides from "Liberating Kubernetes from kube-proxy and iptables". Here’s a summary of what pod-to-pod communication really looks like without kube-proxy and iptables, with the client on Node A and the Nginx server on Node B. ↳ The Client Pod (1.1.3.1) on Node A wants to talk to the Nginx Server Pod (1.1.1.1) on Node B. ↳ The Nginx server is exposed through a Kubernetes Service with ClusterIP (3.3.3.3:80) instead of the pod’s direct IP. ↳ When the client makes a request to the server: ↳ The packet first leaves through the pod’s own eth0 interface. ↳ It then enters the host network stack through the veth pair endpoint (lxc0) that connects the pod to the host. ↳ At this point, Cilium intercepts the traffic in its eBPF datapath, looks up the Service eBPF map, and decides how to forward it. ↳ At this stage: ↳ Source Address: 1.1.3.1 ↳ Destination Address: 3.3.3.3:80 (ClusterIP) ↳ Cilium looks through the Service eBPF Map (SVC map lookup), checking which backend pods are associated with the ClusterIP. ↳ It selects one of the endpoints. ↳ Performs DNAT (rewrites the destination on the packet from ClusterIP → endpoint IP, i.e, 3.3.3.3 → 1.1.1.1). ↳ Cilium then creates entries in the eBPF conntrack map: ↳ The first entry is the SVC CT (Service Connection Tracking entry): source is the client pod, destination is the ClusterIP. ↳ Second entry is the Egress CT (Egress Connection Tracking entry): same source, but with the real endpoint IP of the Nginx pod as the destination. ↳ When the reply comes back from the Nginx pod, Cilium intercepts again and uses the Egress CT entry to perform reverse DNAT, so the client still sees the response as coming from the cluster IP (3.3.3.3) instead of the pod’s direct IP. Link to Slides: https://lnkd.in/dKbsNDiw

474

19 comentarios