Serverless Cold Starts, Measured: Where the Latency Actually Goes
We measure, not speculate. This video instruments cold starts on real serverless platforms — Lambda, Cloud Functions, and a container-based runner — and breaks the latency down into what actually took time: sandbox setup, runtime init, dependency loading, and first request handling. The measurements show which levers actually work: provisioned concurrency, warm-up pings, keeping dependencies small, and pruning the runtime's import graph. We also look at how the platform hides the cost — keep-alive windows and execution environment reuse — and what that means for latency-sensitive workloads with tight p99 budgets.
Topics covered:
- Measured cold-start breakdown on three platforms
- What the platform reuses between invocations
- Provisioned concurrency and warm pools
- Which optimizations actually move the p50 and p99
Related articles
Cold Starts Are Not a Myth
What a serverless cold start actually measures — microVM boot, runtime init, handler wiring — and how Lambda SnapStart and Firecracker change the numbers without changing the physics.
Serverless Pricing Is a Latency Multiplier
How GB-seconds billing, per-invocation fees, and the cold-start tax make every millisecond of function duration a line item — with real Lambda pricing math.
Multi-Region Deployments and Latency
Speed-of-light RTT floors, active-passive versus active-active architectures, DNS-based routing, and replication lag — the real tradeoffs of running in multiple regions.
More in Cloud & Infrastructure
Edge Computing Explained: Where Compute Actually Sits
What edge computing actually is — the compute tiers from device to far edge to cloud, latency and bandwidth budgets, and which workloads genuinely benefit.
DetailsContainer Orchestration Basics: API Server, Controllers, and Scheduler
Container orchestration from first principles — what the API server, controller manager, and scheduler actually do, with Kubernetes as the working example.
DetailsThe Cost of Distributed Systems: Coordination, Consistency, and Failure
What distributed systems actually cost — coordination, consistency, and failure taxes — quantified with quorum math, tail latency, and retry-storm dynamics.
DetailsDNS and Traffic Routing in the Cloud: From Resolver to Anycast
How DNS actually routes traffic in the cloud — record resolution, CDN anycast, geo routing, load balancer handoffs, and how TTLs shape your failover story.
DetailsMulti-Region Architecture: Active-Active, Failover, and Replication
The real mechanics of multi-region deployments — where writes land, how replication propagates, what failover flips, and the latency math that constrains every design.
DetailsObject Storage Under the Hood: PUT, GET, and Erasure Coding
What happens inside an object store — the PUT and GET paths, metadata partitions, erasure coding, and why object storage is eventually consistent.
DetailsAutoscaling Explained: The Controller Loop Behind Horizontal Scaling
How autoscaling actually works — the metrics window, desired-replica calculation, stabilization, and why naive CPU-based scaling oscillates under real load.
DetailsKubernetes Scheduling Visualized: The Filter-Score Pipeline
What the Kubernetes scheduler actually does — how pending Pods become assigned to Nodes, the filter-then-score pipeline, and how taints and affinity shape placement.
DetailsContainer Images Explained: Layers, Manifests, and Digests
How container images are actually built and run — layers, manifests, content digests, and what the runtime does at pull and run time.
DetailsDepth, delivered weekly
One technical dispatch a week — articles and episode notes before they go public.
One technical dispatch per week. No noise.