Shwan Ciyako

Well written blog post from my colleague about a specialised Terraform HTTP Backend which allows you to store your terraform.state file in HashiCorp Vault.

18

It's been years since Java 8 was released, but let's revisit this change in Java. it’s crucial to understand the memory management within the JVM to optimize our applications and prevent memory-related issues. Two key concepts in this area are PermGen Space and Metaspace. PermGen Space: Permanent Generation (PermGen) was used in versions up to Java 7 to store metadata about the classes, methods, and other reflective data. It had a fixed size, which often led to OutOfMemoryError: PermGen space if the application loaded too many classes or methods, especially in large enterprise applications or those using frameworks that dynamically generate classes (like Hibernate or Spring). Transition to Metaspace: With the release of Java 8, PermGen was replaced by Metaspace. Metaspace uses native memory (outside of the heap) for class metadata, offering more flexibility and alleviating the fixed-size constraint of PermGen. The size of Metaspace can grow dynamically, reducing the likelihood of running out of memory, though it still can be controlled via the -XX:MaxMetaspaceSize flag. Advantages of Metaspace: Dynamic Sizing: Metaspace can expand as needed, which helps avoid the OutOfMemoryError that was common with PermGen. Improved Performance: Class metadata management in Metaspace is more efficient, leading to better performance of Java applications. Best Practices: Monitor Metaspace Usage: Use tools like jconsole, jvisualvm, or JMX to monitor and profile Metaspace usage. Set Limits: Use -XX:MetaspaceSize and -XX:MaxMetaspaceSize to control the initial and maximum size of Metaspace, ensuring your application doesn’t consume all available native memory. Optimize Class Loading: Review your application’s class loading and unloading patterns. Avoid excessive class loading, especially in web applications with frequent redeployments. Web : https://lnkd.in/g8GnqxGF #Java #Programming #JVM #MemoryManagement #JavaDeveloper #TechTalk

9

🚀 gpt-4-turbo (2024-04-09), the latest stable release from OpenAI is now available on #AzureOpenAIService in East US 2 and Sweden Central Azure regions! We will be rolling out to more regions over the coming weeks, including a schedule for when we will begin replacing 1106-preview and 0125-preview with the stable version. We're working hard to add support for JSON mode and function calling for vision input, but it is currently only available with text input. ✨We've also recently expanded our support for fine-tuning to East US 2 and Switzerland West bring the latest support for gpt-35-turbo model versions 1106 and 0125. This includes supporting multi-turn chat training, like ultrachat_200k. Stay up on the latest: https://lnkd.in/gkX_sukm Announcement: https://lnkd.in/ggG4_qjg 🎉 #AzureAI #GenerativeAI #GPT4Turbo #GPT35Turbo #FineTuning #GenerallyAvailable #GA #Announcement

198

1 kommentar

Learn more about Moderating and detect harm with Azure AI Content Safety. #Azure #AI #Governance

6

Have not been on YouTube for some time, but coming back hopefully more often :) Join me in this tutorial on running Gemma2 on a Kubernetes cluster to achieve "Local Gemini". Learn step-by-step how to set up and manage Gemma2 efficiently within your Kubernetes environment. Whether you're a beginner or an experienced user, this video will guide you through the process seamlessly. Watch now and enhance your skills in managing Gemma2 on Kubernetes! Disclaimer: This is 101 simple deployment so needs some proofing before going into prod! https://lnkd.in/dzznp8fK #Gemma2 #LLM #K8s #Google #ChampionInnovators #ChatGPT #Gemini

14

4 kommentarer

Very nice article about displaying Azure infrastructure costs in Grafana via the Infinity Plugin. You do not always have to use the newest SaaS tools in order to provide basic information ;) https://lnkd.in/er3dKFgS

11

3 kommentarer

Join Björn Sundling as he explains the intricacies of Azure DevOps management with #PowerShell. Learn about the Azure DevOps API, OAuth authentication, and the innovative AZDM framework. #PSConfEU

6

In an ever-evolving cyberthreat landscape, you need an effective security posture strategy. Microsoft 365 Defender makes it simpler to identify and prioritize the right controls—and even makes recommendations based on threat intelligence. Learn more.

2

It was time to pick up the pen again and write how you can streamline Azure infrastructure deployments using the new extendable parameter files using Azure Bicep. In this post, I dive into how you can enable the new experimental feature "extendable parameter files". This brings in another level of modularity and efficiency when you are deploying Azure resources. Check out the post at: https://lnkd.in/esaBtQAU #azure #blog #bicep #azurebicep #iac #devops #cloudcomputing #modularization

16

What does life with #openbsd look like in the longer run? Some insights to be found in "You Have Installed OpenBSD. Now For The Daily Tasks." https://lnkd.in/dExJatNX (prettified, tracked: https://lnkd.in/dWfdFmWp) #OpenBSD #install #syspatch #supported #releases #packages #pkg_add #sysupgrade #release #snapshot #current #sysadmin #keepitrunning #keepitsecure #security #development

7

How to easily integrate AI in Java business applications? LangChain4j is a promising choice to do so. It is the Java variant of the popular python framework LangChain for Generative AI. LangChain4j supports developers in using different LLMs, Vector-databases and other AI features. Compared with Spring AI, it has more features, but seems to be less mature. In my article you see how integration and usage in with Spring Boot: https://lnkd.in/gfSWhv3P #AI #Java #LangChain4j #SpringBoot #MachineLearning #ArtificialIntelligence #TechInnovation #DeveloperTools

12

1 kommentar

When architecting platforms the question always comes up - is it better to buy or build a component of the solution? This entails talking with vendors who aspire to grow their pipeline and ultimately land a new customer, or sell an existing customer on a new product offering. Many times it can be difficult to determine "fit" since so much is at stake from both sides of the equation. Which is why I salute Snowflake. Had a meeting with a sales rep today who could have easily pushed some new technology upon us. Instead, the person truly cared about our needs - to the point of recommending that we not use Snowflake in the way we were considering. Trust. It takes time to build. However, my trust in this sales rep took a large step forward today. Snowflake is an absolutely incredible platform in many ways that can be very beneficial to us - just not in this one path. In the era of tight markets, thinning margins, and customer churn - Snowflake did right in bringing this individual on board to represent them and their products. Just want to say to Snowflake - thanks...

12

1 kommentar

Let's talk about Teams In this session we talk about one of the most requested features for Teams Rooms on Android: Custom Background. Now available in Teams Admin Center Michael Fogh Broholt Carsten Lund Meilbak #teamsrooms #custombackground #mvpbuzz

14

Excellent resource for incorporating Large Language Models (LLM) into your Raspberry Pi setup. This comprehensive guide takes you from a bag of parts to developing your own program with LLM capabilities, all while running it on your low-power device. Check it out here: https://lnkd.in/d4Dwa7-i #ollama #raspberrypi #llm #ai

8

I have cleared the certification exam for #GoogleCloudCertified Professional Cloud Architect for the third time. You can check it out here: https://lnkd.in/dfVgXQZx I also promised Carole Kurowski that I would write some tips for people interested in taking the exam and muse on how the exams these past four years reflect on cloud computing trends. Please leave a comment if this interests you and if there's something specific you would like to know about! #GoogleCloud #CloudArchitect

21

The SSH vulnerability announced this week (https://lnkd.in/dRAu6mQh openssh-vulnerability-could-lead-to.html) hopefully spurred you into action if you run servers with SSH publically accessible. Please don't put your servers behind a VPN. It makes no sense to route all your traffic through a single remote VPN IP if you can have a direct peer-to-peer encrypted connection instead. Wireguard is super-fast, incredibly easy to configure and deploy, and built-in to the Linux kernel. If you are traveling and don't have a dedicated IP, use a dynamic DNS service (like the awesome free DuckDNS) to update your IP and add a systemd timer to your server to refresh your wireguard interface. See https://lnkd.in/dZnZ2yMX. The wireguard package has a convenient reresolve-dns.sh script to do this for you. You know you will sleep MUCH better if your server is only accessible from a single IP address. With a little bit more creativity, one can even combine Wireguard access with SSO access and create a simple zero-trust network.