Simulation

Run traffic at 100M RPS against your architecture design.

A real-time tick engine updates every node and edge on your canvas every 100 milliseconds while traffic runs. Stateful service models track Lambda concurrency and cold starts, Kinesis shards, DynamoDB capacity, and SQS queue depth — before a single resource exists.

Spike · 12,400 RPS
Traffic simulation Live
⚠ Lambda concurrency at 92% Est. $2,847/mo under load
100msLive tick interval
100MRPS maximum
4Traffic patterns
30+Stateful service models
Live simulation engine

Not a calculator. A running system.

Simulation isn't a spreadsheet estimate — it's a live loop that pushes traffic through your actual canvas topology and updates every node while it runs.

Real-time tick simulation

A 100ms loop updates every node and edge on the canvas while the simulation runs — load, latency, errors, and cost tick live.

10 → 100M RPS

Logarithmic RPS slider from quiet dev environments to global-scale production traffic.

Traffic input modes

Set direct RPS, or express load as total requests per day, week, or month — Pinpole converts it to steady RPS.

Pause / resume / stop

Control the simulation mid-run. Edit traffic configuration while paused, then resume against the new profile.

Stateful service models

Per-tick state for Lambda concurrency and cold starts, Kinesis shards, DynamoDB capacity, SQS queue depth, and more.

Traffic distributors

ELB, Route 53, CloudFront, and Global Accelerator model round-robin and failover routing across your graph.

Traffic patterns

Model the load your architecture will actually see.

Steady state, organic growth, launch spikes, and daily cycles — each pattern surfaces different failure modes.

Constant

Steady-state load

Fixed RPS validates baseline behavior at your expected operating throughput. Size quotas, validate limits, and model baseline cost.

Sizing · quota validation · baseline cost
Ramp

Organic growth

Load increases linearly from zero to target RPS. Models growth and validates auto-scaling trigger points before you commit.

Capacity planning · scaling validation
Spike

Launch & viral moments

Instantaneous jump to peak RPS. Stress-tests concurrency limits, queue overflow, and cold-start cascades — the pattern that finds production incidents.

Launch prep · concurrency testing
Wave

Daily & batch cycles

Oscillating load between baseline and peak. Models payroll windows, Black Friday cycles, and periodic batch workloads.

Auto-scaling · on-demand vs provisioned
Per-node metrics & alerts

See bottlenecks where they happen — on the canvas.

Every node shows load %, sparklines, RPS, latency, error rate, and health status while the simulation runs. Edges animate with live throughput labels. Alerts fire with actionable suggestions.

API Gateway / events-api12,400 req/s
Lambda / event-processor847 / 1,000 concurrency
DynamoDB / events-table2,340 / 3,000 WCU
SQS / async-queue14,230 depth
Lambda invocations $184/mo DynamoDB WCU $1,190/mo API Gateway $442/mo Total estimate $2,847/mo
Live node overlays

Load %, 6-second sparklines, RPS, latency, error rate, and status — healthy, scaling, throttled, or error.

Live alerts

Critical throttling alerts and high-load or elevated-error warnings, each with actionable suggestions.

Latency aggregation

P50 / P95 / P99 computed across the whole graph, not just individual services.

Cost estimation

Real cost under real load — not calculator guesses.

The cost model runs inside the simulation, so your estimate reflects your actual traffic pattern, not a vendor calculator's steady-state assumption.

Real-time cost dashboard

Live $/sec, $/hour, $/day, and $/month projections update in the Simulation panel as the run progresses.

Per-node cost breakdown

Expandable list sorted by load — each node shows cost per second and service-specific metrics.

Per-service pricing models

Dedicated cost handlers for Lambda, API Gateway, DynamoDB, SQS, SNS, S3, ElastiCache, CloudFront, ELB, Kinesis, and more.

Bedrock & AgentCore pricing

Token and unit pricing for foundation models and AgentCore services — AI workload costs modelled like any other service.

Monthly request estimate

Derived from your current RPS so the dollar figure always has request-volume context.

Pattern-aware totals

Spike and wave patterns produce different bills than constant load. The estimate reflects the pattern you chose.

Modelled services

Dedicated tick processors, service by service.

Each of these has its own behavioural model in the live engine — capacity, throttling, and pricing. Anything else runs through a generic passthrough model.

LambdaAPI GatewayAppSyncDynamoDBSQSSNSEventBridgeS3ElastiCacheKinesisCloudFrontELBRoute 53AmplifyBedrockKnowledge BaseGuardrailsAgentCoreGlueDeequOpenSearchADOTMWAAMediaConvertMediaPackageMediaLiveKinesis Analytics / FlinkQLDB
Optimization

Recommendations ranked by severity.

After each simulation, Pinpole analyzes your architecture and returns findings with specific services, problems, and fixes — apply with one click.

WARNING

Lambda / event-processor — Concurrency limit (1,000) will be exceeded at 1,240 RPS. Recommended: increase reserved concurrency to 8,000.

ADVISORY

Firestore / events-collection — Write rate at peak exceeds hot-key limits. Estimated additional cost: $620/mo. Recommended: shard document paths.

INFO

Azure Functions / async-handler — Cold-start latency may affect p99 by ~340ms at peak. Consider Premium plan for latency-sensitive paths.

Every recommendation can be applied with one click — Pinpole patches node config or adds new services (an SQS buffer, a CloudFront CDN) and wires the edges for you. See the full Optimization page for how the recommendation engine works.

History & sharing

Every run is evidence you can share.

Execution persistence

Runs are saved with traffic config, duration, peak RPS, cost breakdown, node metrics, and recommendations.

Local history

Your last 50 runs are mirrored to the browser, so recent evidence is always one click away.

Embed simulate

Share a tokenized interactive simulation canvas via URL — no login needed to view. Used by MCP and external tools.

MCP simulation

AI agents can run batch simulations through the API and return an inline simulation canvas directly in chat.

Export GIF

Capture the animated simulation as a GIF for demos, docs, and architecture review decks.

Architecture snapshots

Each run stores the exact canvas state, so you can compare versions and roll back regressions.

Cost forecasting

Live monthly cost estimates update as simulation runs — per service, under load, not from a static calculator.

Failure analysis

Find timeout chains, throttling limits, and queue saturation before your users do.

Execution history

Every simulation is logged with a full architecture snapshot. Compare runs and share evidence in architecture reviews.

Ready to deploy? See the deploy workflow →

Stop guessing.
Start simulating.

No infrastructure spend · No cloud account required to simulate