Yann Leger’s Post

Why Choose Red Hat for Kubernetes? What's different about Red Hat's approach to Kubernetes versus proprietary solutions? Discover how the OpenShift solution gives you the agility to build new applications and restructure traditional apps to become cloud-native. Ready to get started? Contact a CoreHive Computing Red Hat expert for a consultation.

Anyone actually run an inference with this yet? Any tips on local hardware, EC2 instance sizes or types? Initial attempts have (I guess) run out of GPU memory. https://lnkd.in/e-pxpMza

Really excited to be working with my friend and fellow committer Jordan West on a performance patch for Apache Cassandra that's already showing a MASSIVE 6x reduction in IOPS usage during compaction which makes a huge impact on EBS. In our initial testing we saw a 3x improvement to compaction throughput, just on one thread, and multiple threads will deliver an even bigger improvement. The same patch will reduce system overhead on *any* deployment, but will have the biggest impact when using disks that use quotas on IOPs. The impact is you'll be able to run WAY denser nodes on EBS for the same cost, more predictable random read performance, and we're not even finished. This same patch should deliver improvements to range scans (hello spark jobs!) and repairs. You can follow along here: https://lnkd.in/d5_hcyJQ

Check out this new video from Ravi Theja Desetty walking you through the big new features in LlamaIndex 0.9: https://lnkd.in/gXWUpBZw Get a step-by-step look at ⭐️ Our new ingestion pipeline feature ⭐️ Transformation caching in memory and in Redis ⭐️ Custom transformations ⭐️ Our new, simplified interface for parsing and splitting text ⭐️ Our new-and-improved token counting

One of the key benefits of using #ebpf is achieving 100% visibility without any blind spots. With the release of Coroot 1.3.0, we have enabled tracking of connections to external services by default.  This feature provided immediate insights into our lab environment. For instance, we discovered that our MongoDB reports usage statistics to api.segment.io. It’s a documented behaviour, but did you know about that? If you want complete control over your system, you know what to do :) https://lnkd.in/d7sBru5h

Here is a modern, production-ready, take on easy serviceability - no need for a screwdriver to change failed hard drive. What generally takes two people and 15 minutes now can be done by 1 person and in under 1 minute. #OCP #circleB #modular #infrastructure

Thrilled to support ✨ NEW usage-based-routing-v2 strategy 👉 https://lnkd.in/gKz7hzA4 (+4 more updates 👇) 💪 v2 handles redis race conditions 🛠️ routing w/ redis works across multiple instances 🎉 Async cooldown logic ⚡️ Async get_available_deployment logic

This is what I like about Cloud Architecture and applications: every single detail is about efficiency. Because efficiency means less CPU, less ram and less costs. In this video the CTO explains how they move their infrastructure to ARM64 serverless architecture in order to reduce costs and to use less power (and less carbon). It was not so easy to implement but they have always two points in their minds: - observe everything to be sure that they can compare with old versions - ability to fallback to working version whenever they need in just 30 seconds There are also some hints about serverless we should always stick to: - workloads that need to be executed in few seconds can be moved to serverless, but if response time can be forwarded then EC2 fits best - for Lambda you should reduce time and load especially for results: if the JSON is heavy in size, save it on S3 and give back a link This is an amazing video Take a look https://lnkd.in/drXMjCnu

🏎️ Take #AzureCosmosDB for a free 30 day test drive. No credit card required. Watch this short video with Meredith Moore and learn more about how to get started. https://lnkd.in/eBG5-4sD