Mobile Menu

Benefits

  • Enables realistic system validation by combining synthetic and trace-based traffic.
  • Helps uncover bottlenecks, tail-latency issues and QoS violations early in the design cycle.
  • Supports quick “what-if” studies on workload intensity, traffic mix and correlation.
  • Provides a unified traffic framework that can be reused across hardware, software and network models.
  • Reduces dependence on early prototypes by virtualizing realistic workloads in VisualSim.

The Traffic Modeling library in VisualSim Architect provides a rich set of traffic and workload generators that drive hardware, software and network models with realistic stimulus. Designers can combine distribution-based, sequence-based, trace-driven and waveform-based generators to emulate real systems ranging from embedded controllers to large-scale data centers and AI accelerators.

The library supports both synthetic traffic (defined by probability distributions and time patterns) and application workloads (such as AI, DNN and protocol stacks), enabling thorough validation of performance, power, reliability and QoS.

Overview

This library includes multiple traffic modeling mechanisms that can be used independently or in combination:

  • Distribution-Based Generators – Use probability distributions to create packet, transaction or task arrivals. Includes support for libraries such as Colt for scientific computing in Java, with a wide range of probability distribution functions and statistical utilities.
  • Sequence-Based Traffic – Drives the system with a predefined or scriptable sequence of requests, packets, transactions or software calls.
  • Probability and Time-Based Models – On/off patterns, Markov-style state transitions, burst/idle cycles and time-window-based activation.
  • Trace-Based Traffic – Uses recorded traces from network, software, bus or processor activity (e.g., Ethernet, PCIe, AMBA, application logs) to replay realistic workloads.
  • Waveform Generators – Clock, pulse, periodic, sine wave and other waveforms to model periodic triggering, sampling and control signals.
  • Hardware / Network / Software Traffic Generators – Domain-specific generators that mimic typical behavior of cores, devices, links, protocols and tasks.
  • AI and DNN Workload Generators – Pre-configured workloads representing AI inference and training pipelines, deep neural nets and other compute-intensive applications.
  • Task Graph Workloads – Workload models built from task graphs (DAGs) covering multiple stages, with dependencies, parallelism and pre-configured application templates.

Supported Features

  • Distribution-based traffic:
    • Uniform, Normal, Exponential, Poisson, Pareto, Gamma and more.
    • Parameterizable mean, variance, burst length and idle intervals.
  • Sequence and script-driven generators:
    • CSV- and table-based sequences of time, size, address and type.
    • Scriptable pattern definitions for complex test cases.
  • Trace-based traffic:
    • Replay traces from network captures, bus monitors or software logs.
    • Filtering, scaling and sub-sampling of traces.
  • Waveform-based triggering:
    • Clock, pulse, periodic, sine and custom waveforms.
    • Duty-cycle and jitter controls.
  • AI and application workload templates:
    • Pre-configured graphs for AI/DNN, multimedia, protocol stacks and control loops.
    • Workload scaling based on batch size, resolution, frame rate or user count.
  • Time-varying load profiles (ramp-up, ramp-down, diurnal patterns).
  • Correlated traffic generation across multiple sources and sinks.
  • Support for multi-domain traffic (hardware, software, network) in the same model.

Key Parameters

  • Distribution_Type – Choice of probability distribution for arrivals.
  • Inter_Arrival_Time – Average or configured time between generated events.
  • Packet_or_Transaction_Size – Fixed or distribution-based size.
  • Burst_Length / Idle_Length – Parameters for bursty traffic models.
  • Sequence_Table – External file or table for sequence-based generators_

Applications

  • Semiconductor and SoC Architecture
    • Driving AMBA, PCIe, UCIe and NoC interconnects with diverse traffic patterns.
    • Stress testing cache hierarchies, memory controllers and coherence protocols.
    • Evaluating throughput and latency of CPU, GPU, NPU and accelerator clusters.
  • Networking and Data Centers
    • Modeling bursty and long-tailed flows in switches, routers and fabrics.
    • Evaluating QoS, priority schemes and congestion-control algorithms.
    • Replaying real network traces from production or lab environments.
  • Automotive, Aerospace and Embedded Systems
    • Simulating sensor, actuator and control-loop traffic under different driving/mission scenarios.
    • Validating ECU and avionics bus utilization under fault and stress conditions.
    • Traffic for CAN, LIN, FlexRay, Ethernet TSN and ARINC-style networks.
  • AI and High-Performance Computing
    • Workload generation for AI inference and training pipelines.
    • Task-graph-based generation for DNN layers and microservices.
    • Multi-tenant workloads in shared GPU/accelerator infrastructure.
  • Software and Cloud Services
    • User request and microservice traffic for cloud-native applications.
    • API call patterns, session traffic and back-end database load.

Integrations

The Traffic Modeling library integrates with:

  • Processor, cache, memory, AI accelerator and power libraries.
  • AMBA, PCIe, UCIe, Corelink, Arteris NoC and other interconnect components.
  • Network protocol models for Ethernet, TSN, TCP/IP and custom flows.
  • RTOS and software-task models using task graphs and schedulers.
  • Visualization, analysis and design-space exploration tools in VisualSim.

Schedule a consultation with our experts

    Subscribe