Performance validation suite

This suite validates that Edera delivers container-like performance while providing VM-level isolation. The tests establish baselines and compare performance between Edera zones and standard containers.

Benchmarking a multi-threaded CPU workload? See CPU performance benchmarking for a self-contained guide covering annotation requirements, multi-core configuration, and real-world thread counts.

Before you begin: Complete the prerequisites including cloning the learn repo and running make setup.

All manifests are available in the learn repository.

Test 1: Network performance with iperf

iperf3 is the industry-standard tool for measuring network bandwidth and throughput. It creates TCP or UDP streams between a client and server to measure maximum achievable throughput.

Network test overview

This test runs iperf3 client and server containers in the same pod (same Edera zone) to measure intra-zone network performance. Both containers share the zone’s network stack, communicating via the pod IP address.

The test validates that Edera’s network virtualization doesn’t significantly impact communication within a zone—the most common pattern for tightly coupled services like sidecars, service meshes, and multi-container pods.

Run the comparison

The easiest way to run this test is with the comparison target, which runs both Edera and baseline benchmarks and outputs a summary table:

make iperf-compare

This will:

1. Deploy iperf pods (Edera and baseline)
2. Run 30-second benchmarks on each
3. Output a comparison table showing throughput and percentage of baseline

Example output:

┌─────────────┬────────────────┬──────────────┐
│ Test        │ Throughput     │ % of Baseline│
├─────────────┼────────────────┼──────────────┤
│ Edera       │   27.4 Gbps    │  92%         │
│ Baseline    │   29.9 Gbps    │ 100%         │
└─────────────┴────────────────┴──────────────┘

Run tests individually

Edera:

make iperf
kubectl exec -it iperf-edera -c iperf-client -- iperf3 -c $(kubectl get pod iperf-edera -o jsonpath='{.status.podIP}') -t 30

Baseline:

make iperf-baseline
kubectl exec -it iperf-baseline -c iperf-client -- iperf3 -c localhost -t 30

Success criteria:
Compare your Edera throughput against baseline. Results vary based on instance type and network conditions in your environment.

Cleanup:

kubectl delete pod iperf-edera iperf-baseline

Test 2: CPU performance with sysbench

sysbench is a scriptable multi-threaded benchmark tool commonly used for database and system performance testing.

CPU test overview

The sysbench CPU benchmark calculates prime numbers up to a specified limit (20,000 in our test) using a trial division algorithm. This is a pure compute workload that:

• Runs entirely in CPU and memory (no disk I/O)
• Uses consistent, reproducible calculations
• Scales predictably with CPU cores
• Isolates CPU overhead from other variables like network or storage

This makes it ideal for measuring virtualization overhead. Any performance difference between Edera zones and standard containers directly reflects the CPU overhead of Edera’s VM-based isolation.

sysbench metrics

• Events per second: Prime number calculations completed per second (higher is better)
• Latency: Time per calculation in milliseconds (lower is better)

The test runs with 2 CPU threads for 60 seconds to get stable, comparable results.

Run sysbench comparison

make sysbench-compare

This will:

1. Deploy sysbench jobs (Edera and baseline)
2. Run 60-second benchmarks on each
3. Output a comparison table

Example output:

┌─────────────┬─────────────────┬─────────────┬──────────────┐
│ Test        │ Events/sec      │ Avg Latency │ % of Baseline│
├─────────────┼─────────────────┼─────────────┼──────────────┤
│ Edera       │     762.25      │   2.62 ms   │  98.4%       │
│ Baseline    │     774.13      │   2.58 ms   │ 100.0%       │
└─────────────┴─────────────────┴─────────────┴──────────────┘

Run sysbench individually

Edera:

make sysbench

Baseline:

make sysbench-baseline

Success criteria:
Compare your Edera events/sec against baseline. CPU performance should be very close to baseline as this is a pure compute workload with minimal virtualization overhead.

Cleanup:

kubectl delete job sysbench-edera sysbench-baseline

Test 3: Establish performance baselines

Document your specific environment’s performance characteristics for future reference.

Create a performance report

Run the benchmarks and record results:

# Create results directory
mkdir -p pov-results

# Run iperf tests (JSON output)
POD_IP=$(kubectl get pod iperf-edera -o jsonpath='{.status.podIP}')
kubectl exec iperf-edera -c iperf-client -- iperf3 -c $POD_IP -t 30 -J > pov-results/iperf-edera.json
kubectl exec iperf-baseline -c iperf-client -- iperf3 -c localhost -t 30 -J > pov-results/iperf-baseline.json

# Run sysbench comparison
make sysbench-compare > pov-results/sysbench-results.txt

Performance validation checklist

• Network throughput comparable to baseline
• CPU performance comparable to baseline

Summary

Cleanup

Remove all performance test resources:

make clean-performance

Next steps

• Operations integration suite - Verify observability and automation
• Security demonstration suite - Validate container isolation