Read the dashboard

Progress checklist

• Dashboard reachable on :8080 (public or localhost via SSM)
• Can interpret Standard÷Express p50 ratio

Overview

The harnessharness
Docker container on the lab EC2 instance that continuously GETs the same keys from Express and Standard, exposes a dashboard on :8080, and reports p50/p90 latency ratios. serves a small dashboard on port 8080. Continuous GETs against the same hot keys produce latency percentiles for Express and Standard. The headline ratio is Standard p50 ÷ Express p50 (higher means Express is faster on that percentile). Prefer the ratio over absolute milliseconds from a quiet lab.

Field	Meaning
Express p50 / p90 / p99	GET latency for the directory bucket
Standard p50 / p90 / p99	Same keys against the general-purpose bucket
Standard ÷ Express p50	How many times slower Standard is vs Express at p50
samples / errors	Rolling window (default 300s) plus cumulative errors

export AWS_PROFILE=sandbox
export AWS_REGION=ap-southeast-2

Steps

1. Open the dashboard — pick public IP or SSM port-forward to view on localhost.

   Public URL

   ./scripts/demo.sh dash

   Looks like (verified shape; IP is yours):

   http://203.0.113.10:8080/
   Local view (SSM): ./scripts/demo.sh port-forward  →  http://127.0.0.1:8080/

   Open that URL in a browser. Security group ingress allows TCP 8080 (lab default 0.0.0.0/0).

   SSM port-forward (localhost)

   Requires the Session Manager plugin on your laptop and an instance role with AmazonSSMManagedInstanceCore (see IAM and EC2). Leave the session running while you browse.

   # optional: LOCAL_PORT=18080 ./scripts/demo.sh port-forward
   ./scripts/demo.sh port-forward

   Looks like (verified shape):

   Forwarding remote :8080 → http://127.0.0.1:8080/  (Ctrl+C to stop)
   …
   Port 8080 opened for sessionId …
   Waiting for connections...

   Then open http://127.0.0.1:8080/ (or your LOCAL_PORT). Equivalent raw CLI:

   aws ssm start-session \
     --target "$(jq -r .instanceId .lab-state.json)" \
     --document-name AWS-StartPortForwardingSession \
     --parameters '{"portNumber":["8080"],"localPortNumber":["8080"]}'

2. Wait until both series have enough samples (hundreds of GETs), then read the headline (Express is N× faster), the ms saved at p50, and the equal-scale race bars (Standard = full width). p50 / p90 / p99 stay in the two panels.

   

   Live harness UI (this lab ≈4× at p50). Bucket suffix and account id are redacted.

   Tip

   Co-location matters. If the instance is not in the AZ that maps to apse2-az1, Express ratios will look worse than the design intent.

3. Cross-check with CLI status and /api/stats.

   ./scripts/demo.sh status
   curl -sS "http://$(jq -r .publicIp .lab-state.json):8080/api/stats" | jq '{
     express: .express|{count,errors,p50_ms,p90_ms},
     standard: .standard|{count,errors,p50_ms,p90_ms},
     standard_over_express_p50
   }'

   Looks like (verified in this lab; your numbers will differ — cite the ratio, not a fixed ms claim):

   {
     "express": { "count": 3671, "errors": 0, "p50_ms": 6.07, "p90_ms": 6.38 },
     "standard": { "count": 3671, "errors": 0, "p50_ms": 25.19, "p90_ms": 28.05 },
     "standard_over_express_p50": 4.15
   }

   Via port-forward, use http://127.0.0.1:8080/api/stats instead of the public IP.

Verify

Check	Expect
Dashboard HTML	Loads on :8080 (public or 127.0.0.1)
Sample counts	Climbing / non-zero in the rolling window
Errors	0 on both sides after warm-up
Ratio	standard_over_express_p50 ≫ 1 when co-located (this lab ~4× at p50)
SSM forward	session-manager-plugin installed; get-connection-status → connected

Next

When finished, Tear down so On-Demand and storage charges stop. Then read Where Express fits before treating Express as production-only storage.

References

• S3 Express One Zone
• Optimizing for performance (co-locate compute)
• Differences for directory buckets
• Start a Session Manager session (port forwarding)
• Install the Session Manager plugin