Google Cloud - Google Cloud Run From Code to Cloud, Instantly

Cloud Run is Google Cloud’s fully managed serverless container platform designed to remove all infrastructure complexity so teams can deploy apps fast without managing servers, clusters, or nodes. It runs any container built for Linux x86_64 and supports all major languages, frameworks, and libraries. This video gives a practical overview of how Cloud Run works, why it’s so cost-efficient, and how developers and teams can use it for real workloads while keeping budgets under control. Cloud Run supports two execution models. Cloud Run Services are used for real-time and request-driven apps like APIs, web services, and event handlers. Cloud Run Jobs are used for batch workloads that need to run to completion instead of listening on a port. Cloud Run has a strict container runtime contract. Services must listen on 0.0.0.0 using the PORT environment variable, usually port 8080, and should not implement TLS because Cloud Run handles encryption. Jobs must not listen on a port at all and simply exit with code 0 or non-zero depending on success or failure. Resources are flexible. Each instance defaults to 1 vCPU and 512 MiB memory, with AVX2 CPU support. Cloud Run also supports NVIDIA L4 GPUs for AI inference and heavy compute tasks, with drivers automatically mounted so developers don’t need manual setup. Autoscaling is one of Cloud Run’s superpowers. It automatically scales up with load and down to zero when idle. This means you only pay when your code runs. Cold starts happen when scaling up from zero and usually take 10 to 40 seconds depending on image size and startup time. If you need zero cold starts, you can configure minimum instances to keep containers warm. Idle instances cost far less than active ones, making this a smart way to balance performance and cost. Pricing is pay-per-use, billed in 100 millisecond increments. Active time is billed at a higher rate, while idle time for warm instances is much cheaper. Cloud Run includes a generous free tier in us-central1 with vCPU-seconds, GiB-seconds, and free requests every month. Committed use discounts are available for predictable workloads. Cloud Run integrates seamlessly across Google Cloud. It works with Cloud SQL, Spanner, Firestore, Cloud Storage, Memorystore, Pub/Sub, Eventarc, Cloud Scheduler, and Workflows. Security uses service accounts and Application Default Credentials, and enterprise controls like Binary Authorization and VPC Service Controls can enforce governance. Cloud Run and GKE work well together. Cloud Run is ideal for simple, stateless, scale-to-zero apps. GKE is better for full control, custom networking, or stateful systems. Both share compatible APIs and container workflows, making hybrid deployments straightforward. If you want help building your own serverless AI stack, check out www.imbila.ai http://www.imbila.ai/services