Embracing Modern Workloads without Losing "IT"
summary
Claims a production-ready stack that runs big data, streaming, cognitive and containerized web apps together on-prem using Mesos + Kubernetes + Spark and a two-level scheduling approach. Demonstrates a Spark-on-Mesos component that cuts worst-case job runtimes and makes pipelines predictable, plus MPI on Mesos for HPC. Shows a live GPU inference demo of style-transfer where GPUs scale up to ~99% use via dynamic GPU allocation, and cites IBM SmartCloud deployments and thousands of users in production.
AI-written summary · about
chapters (6)
related talks
47mBuilding High Performance Real-time Stream Processing Microservices
05mIgnite: Enabling multi-cloud and breaking vendor lock-in with Cloud Sidecar
14mManaging decisions with Amazon Kinesis for better outcomes
08mStepan Pushkarev Interview
46mComposable microservices for streaming analytics
53mHow do we build high‑load microservices
01h 00mDeveloping Distributed Systems with Apache Mesos
14mPerformance Optimization in High-Volume Transaction Systems