Hybrid Scheduling

Composer Resource Allocation is a hybrid quantum scheduling system that uses knowledge on the state of the network to have an awareness of the connected quantum hardware and simulation resources.

Architecture

  flowchart TD
    classDef smallText font-size:12px;

    A["Composer Gateway (API)"] --> B["Composer Resource Allocation"]
    B --> C["Composer Orchestrator"]
    C --> D["Maestro (Simulator)"]
    C --> E["Composer Interfaces (QPU Access)"]
    E --> F{{"IBM"}}
    E --> G{{"IQM"}}
    E --> H{{"AWS Braket"}}
    E --> I{{"LRZ"}}

How It Works

Step 1: Job Submission

When a user submits a quantum program through Divi or the Composer Gateway API, the job enters the scheduling pipeline. Each job carries metadata including circuit structure, qubit count, shot count, and any hardware preferences.

Step 2: Resource Assessment

Composer Resource Allocation maintains a live view of available resources:

• Simulators: Maestro instances and their current capacity
• QPU Backends: Available hardware via Composer Interfaces, including queue depths and calibration status
• Network State: Latency and bandwidth between computing nodes

Step 3: Intelligent Routing

Based on the job’s requirements and available resources, the scheduler routes to the optimal backend:

Factor	Decision
Small circuit, few qubits	Route to Maestro simulator
Large circuit, many shots	Route to QPU hardware
Batch of mixed circuits	Split between simulator and hardware
High queue times on hardware	Fall back to simulation

Step 4: Execution and Aggregation

The Composer Orchestrator manages the job through its lifecycle:

1. Circuits are dispatched to assigned backends
2. Results are collected asynchronously
3. Error mitigation is applied where configured
4. Aggregated results are returned to the user