Edge Delta Benchmarks

Benchmarks

Our team ran a series of tests to measure the performance and resource utilization of the Edge Delta agent compared to other log collection agents. To do so, we ran popular agents on Amazon EC2, collecting varying data volumes.

The Edge Delta agent v3 is responsible for more tasks than other log collection agents. In addition to collecting and forwarding data, Edge Delta’s agent:

• Processes log data (e.g., filter, shape/transform, logs-to-metrics, and more)
• Runs AI/ML to detect anomalies
• Clusters recurring loglines into Patterns

Note: The performance benchmarks may change between minor Agent versions. These fluctuations might not be explicitly documented in the Agent’s changelog.

Test Specifications

• Instance size: Amazon EC2 c5d.4xlarge
• Pipeline configuration version: v3

Measurements

CPU Utilization: The percentage of the instance’s central processing unit (CPU) actively executing tasks to understand system efficiency.

Memory Utilization: The proportion of a computer’s random access memory (RAM) actively in use. Efficient management ensures optimal system performance and responsiveness.

Performance Benchmarks

CPU Utilization %

Memory Utilization %

PII Masking Performance

This section documents the mask processor’s throughput when applying multiple regex patterns simultaneously.

Test Configuration

PII Patterns Tested

The following regex patterns were applied simultaneously:

• Email addresses
• Credit card numbers (Visa, Mastercard, American Express, Discover)
• IPv4 and IPv6 addresses
• MAC addresses
• IBAN (International Bank Account Numbers)

Reproduce this test

The test uses httploggen, an open-source tool that generates realistic nginx logs with embedded PII data.

./httploggen \
    --endpoint http://localhost:8080 \
    --format nginx_log \
    --number 1 \
    --workers 80 \
    --period 1ms \
    --total-time 1m

This configuration creates 80 concurrent workers, each sending one log event per millisecond for 60 seconds, producing a maximum potential load of 80,000 events/second.

Configure the Edge Delta agent with a mask processor that includes the PII patterns listed above. Disable downstream exporters to isolate masking performance from network latency.

Results