https://www.chenshaowen.com/en/tags/application/Recent content in Application on Shaowen Chen's WebsiteHugo -- gohugo.ioen©2016 - {year}, All Rights Reserved.Sat, 17 Jan 2026 00:00:00 +0000weeklyhttps://www.chenshaowen.com/en/blog/test-and-deploy-pd-disagg-app-with-erdma-on-aliyun.htmlSat, 17 Jan 2026 00:00:00 +0000Sat, 17 Jan 2026 00:00:00 +0000https://www.chenshaowen.com/en/blog/test-and-deploy-pd-disagg-app-with-erdma-on-aliyun.htmlIn a PD disaggregated deployment, heterogeneous GPU models are often used to deploy the model across machines, which multiplies the cross-machine communication pressure. RDMA devices are usually brought in to accelerate kvcache transfer between nodes so as to achieve a lower FTTL. This post describes how to test eRDMA devices微信公众号Alibaba CloudeRDMARDMAPDDisaggregationAIApplicationOperationshttps://www.chenshaowen.com/en/blog/using-vllm-to-deploy-pd-disagg-app.htmlSat, 20 Sep 2025 00:00:00 +0000Sat, 20 Sep 2025 00:00:00 +0000https://www.chenshaowen.com/en/blog/using-vllm-to-deploy-pd-disagg-app.html1. Why Deploy LLM Applications with PD Disaggregation In the process of LLM inference, there are two serial stages: Process the entire input context and generate the KV Cache (Prefill stage) Incrementally generate new tokens (Decode stage) These two stages have different resource requirements. The Prefill stage has to compute微信公众号vLLMDeploymentPDDisaggregationApplicationhttps://www.chenshaowen.com/en/blog/record-a-troubleshooting-process-for-application-slowness.htmlWed, 08 Nov 2023 00:00:00 +0000Wed, 08 Nov 2023 00:00:00 +0000https://www.chenshaowen.com/en/blog/record-a-troubleshooting-process-for-application-slowness.html1. Symptoms The business side reported that the API of application app-a was slow. Looking at the logs, one particular Pod was slow, and deleting that Pod so it moved to another node fixed it. From the monitoring metrics you can see that the Pod’s CPU usage did indeed spike sharply.微信公众号CPUNetworkApplicationTroubleshootingKubernetesOperationshttps://www.chenshaowen.com/en/blog/multi-cluster-applications-under-kubevela.htmlFri, 17 Sep 2021 00:00:00 +0000Fri, 17 Sep 2021 00:00:00 +0000https://www.chenshaowen.com/en/blog/multi-cluster-applications-under-kubevela.htmlKubevela is currently at version 1.1. We usually consider 1.x releases to be relatively stable and safe to try in production. Through continuous tracking and learning, I have also come to appreciate some of the things Kubevela does well. This is a summary document. 1. What Problem Kubevela Solves Aimed at platform developers Several roles need to be distinguished: development, operations, and operations development.微信公众号KubevelaKubernetesApplicationOperationsMulti-ClusterOAMLearningCloud Native