Skip to main content

Spike Instances

Spike Instances are KVM-based virtual machines engineered to deliver peak performance exactly when your workloads demand it. Each instance can tap into a large pool of CPU and memory resources, automatically scaling up during compute-intensive phases — whether you’re rendering images, running data analytics, or performing complex simulations — and scaling down when demand diminishes.

Load-Based Billing

Rather than charging you for all allocated resources at a flat rate, Spike Instance bills strictly on actual resource load. CPU consumption is metered in CPU-seconds and Memory in GB-seconds, independently tracked. You pay only for the resources your instance loads in real time, not for entire idle capacity, ensuring transparent and fair costs.

Automatic, Real-Time Scaling

Spike Instances dynamically adjust resource allocations in real time:

  • Scale Up: When your application hits a compute-heavy section—such as rendering high-resolution frames or crunching large datasets—the instance automatically borrows additional CPU cores and memory.
  • Scale Down: During lighter phases, unused resources are released, preventing overprovisioning.

This on-demand elasticity removes the need to guess peak requirements and eliminates wasted spending on idle capacity.

Benefits

  • Cost Efficiency: Only pay for resource consumption; eliminate charges for idle cores and memory.
  • Performance on Demand: Burst to high core counts or extra memory when you need it, without pre-allocating.
  • Predictable Spend: Clear, load-based billing makes cost forecasting straightforward.
  • Simplicity: No manual resizing of instances or complex autoscaling groups — Spike Instances handle it automatically.
tip

Spike Instances combine powerful, on-demand performance with a transparent, load-based billing model.

You get the compute you need, exactly when your workload needs it, without reconfiguration or restart of the instance, and pay only for what the workload actually consumed.

Yes, it's that simple.

Example: High-Resolution Image Rendering

Consider a graphics studio running a batch of 3D image renders. The initial phase—loading scene assets and textures—uses minimal CPU and memory. As the rendering engine begins processing each frame, CPU and RAM demands spike dramatically:

  1. Preparation (Low Load):

    • Single-digit CPU utilization
    • Small memory footprint

  2. Rendering (High Load):

    • Dozens of CPU cores fully utilized
    • Hundreds of gigabytes of memory employed

With a traditional cloud instance, you’d provision for peak load and be billed at that rate throughout the entire job. With Spike Instance, you incur charges at low rates during the preparation phase and scale seamlessly to full capacity only when rendering begins—resulting in a substantially lower overall cost without sacrificing speed.

Runtime Resources

There are several types of runtime resources available on Puzl. In the following articles we dive deeper into each resource allocation and billing specifics.