While browsing Chinese casual games recently, I unexpectedly discovered a unique display card – Tesla T10. This source is from the professional data center GPU, originally designed by NVIDIA for cloud gaming services, mainly used for GeForce NOW cloud gaming servers. These retired display cards have now entered the second-hand market, currently priced around 1,350 RMB (approx. $190) on Chinese casual game sites. Due to the affordable price, I bought two to research their performance.
Deploying Prometheus monitoring system on Arista switches
Overview This article explains how to run node_exporter and snmp_exporter using Docker containers on an Arista switch to monitor switch status via Prometheus.
- Cloud
- ...
Deploying LLM Chatbot with Nvidia GPU on Kubernetes
Introduction With the rapid development of artificial intelligence technology, more and more enterprises and developers hope to integrate language models (LLM Chatbots) into their systems. This article guides readers on how to use Nvidia GPUs in a Kubernetes environment to deploy an efficient LLM Chatbot, from necessary drivers and tools to specific deployment steps.
How to monitor container PSI information
Introduction In the previous article, we explored PSI (Pressure Stall Information) and how to monitor system PSI data. This article will delve into how to monitor PSI data for a single container.
Understanding and applying Linux PSI (Pressure Stall Information) indicators
Introduction When CPU, memory, or I/O devices experience contention, workloads will experience latency, throughput loss, and face the risk of OOM termination. Without accurate measurement of such contention, users are forced to conservatively use their hardware resources to avoid unexpected interruptions caused by over-commitment. Since Linux kernel 4.20, the Linux kernel has added PSI (Pressure Stall Information), allowing users to precisely understand the performance impact of resource insufficiency on the entire system. This article will briefly describe how to read PSI information.
Deploy Charmed Kubernetes with OpenStack Integrator
Introduction Charmed Kubernetes is a Kubernetes deployment method provided by Canonical that allows you to deploy Kubernetes to various environments via juju. This article describes how to deploy Charmed Kubernetes on OpenStack and use the OpenStack Integrator to provide Persistent Volumes and Load Balancers to Kubernetes.
- Cloud
- ...
Quick Deploy Kubernetes Cluster: Practical Guide to Using kops on OpenStack
Kubernetes 提供多種部署選擇，而在眾多工具中，kops 以其易用性及高整合性脫穎而出。本文將深入介紹 kops 工具，並透過實際操作引導讀者在 OpenStack 環境中迅速建立一個 Kubernetes 叢集。
How to Choose SSDs Suitable for Ceph
隨著SSD價格的不斷下降，許多技術愛好者和企業開始考慮使用 Ceph 建立基於 SSD 的儲存池，以追求更高的效能。但要確保 Ceph 達到出色的效能，選擇合適的 SSD 極為關鍵。在本篇中，我們將探討如何選擇適合 Ceph 的 SSD。
AMD GPU and Deep Learning: Practical Learning Guide
過去經常聽說 AMD GPU 用於執行深度學習相關軟體非常麻煩，建議有深度學習需求的使用者購買 Nvidia 顯卡。然而，最近 LLM（大型語言模型）很熱門，許多研究單位釋出了基於 LLaMA 的模型，讓我覺得有趣並想測試。我手邊有較多 VRAM 的顯示卡都是 AMD 的，因此決定嘗試使用這些顯示卡來執行。
- Cloud
- ...
Kubernetes Cluster-API Introduction
Kubernetes 在雲原生世界發展這麼多年，也發展出了很多管理其集群生命週期的相關專案，如 kops, Rancher 等。而 VMware 則發起了一個名為 Cluster-API 的專案來利用 Kubernetes 本身的功能管理其他 Kubernetes 集群。本篇文章將會來簡單介紹 Cluster-API 這個專案。