The power of servers, in serverless form
Fluid compute combines the efficiency of servers and the flexibility of serverless, enabling real-time, dynamic workloads like APIs, streaming, and AI.
Pay for compute, not idle time
Fluid runs many requests on a single instance. Billing is based on CPU usage.
Traditional serverless wastes idle time. Provisioned instances can sit idle between requests.
Fluid runs invocations concurrently. Overall cloud resources provisioned are dramatically reduced while invocations are kept alive in-memory.
CPU that’s charged when active, not idle. No more paying for I/O wait times, network latency, and API calls.
“By leveraging in-function concurrency, we were able to share compute resources between invocations, cutting costs by over 50% with zero code changes.”
- AI workloads. Run tasks with reduced latency and higher concurrency, delivering faster, scalable results for all users, regardless of the workload size.
- Business-critical APIs. Fluid Compute runs multiple API invocations concurrently on a single instance to handle traffic spikes.
- Server-side and partial pre-rendering. Pre-render dynamic pages before requests arrive to reduce work at request time.
- Middleware. Perform authentication checks and apply personalization with Fluid Compute.
Vercel Functions
Bridging servers and serverless
Taking the best of servers and serverless to create a new model in computing, scaling business-critical workloads efficiently across global environments.
| Features | Servers | Serverless | Fluid |
|---|---|---|---|
| Cold start handling | Not applicable | Cold starts | Cold-start prevention |
| Scaling | Manual scaling | Auto-scaling | Efficient auto-scaling |
| Concurrency | Horizontal | Vertical | Horizontal & vertical |
| Operational overhead | High maintenance | Minimal, inefficient | Minimal, optimized |
| Pricing model | Upfront cost per server | Pay-as-you-go model | Pay-as-you-compute |
| CPU efficiency | High efficiency | I/O bound inefficiency | Optimized I/O efficiency |
Fluid compute for dynamic web applications
- In-function concurrencyRun multiple invocations on a single function instance, reducing idle compute time and lowering costs.
- Cold-start reductionFunctions are pre-warmed and use bytecode caching to reduce cold starts.
- StreamingSend data to users as it becomes available, improving performance for AI, media, and real-time apps.
- Cross-region failoverReroutes traffic to backup regions during outages.
- Dynamic scalingAutomatically adjusts concurrency and resource allocation based on real-time demand.
- Post-response tasksKeep functions running after sending a response to handle tasks like logging or database updates.