Cloud Computing: Autoscaling Data Center

Data Center Fleet

Request rate 18 req/s

Target utilization 70%

Service tier PaaS

Active servers

–

Utilization

–

Requests served

–

Requests dropped

–

How it works

Each server instance can handle a fixed number of requests per second, set by the service tier — a raw IaaS virtual machine does less work per instance than a fully managed SaaS backend. Every autoscale tick, a control loop compares the incoming request rate against the fleet's current capacity and computes how many servers should be active to keep utilization near the target — the same feedback loop cloud providers run behind "auto scaling groups".

desiredServers = ceil( rate / (targetUtil × capacityPerServer) )
utilization   = rate / (activeServers × capacityPerServer)
requests dropped when utilization > 1

Request rate — how many client requests arrive per second; spikes force the autoscaler to react.
Target utilization — the headroom the autoscaler tries to keep; a lower target scales out earlier but costs more idle capacity.
Service tier — IaaS instances need more of them to absorb the same load; PaaS and SaaS pack more effective throughput per instance, so fewer are needed.
Reset — clears the fleet back to a single idle server and zeroes the counters.

Real-world relevance: this is the core promise of elastic cloud infrastructure — capacity that grows and shrinks with demand instead of being provisioned once for peak load, which is what makes pay-as-you-go pricing possible.

Idle / off Active — healthy Active — busy Overloaded

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.