Flow Difference

Algorithm

The basic idea of the Flow Plugin is to detect data leaks within service layers. It is calculating the difference of two time series and evaluates the deviations. If the deviation is seen as an anomaly an error flag is set.

Evaluation

Error Types

error_norm = metric1 - metric2

Error Handling

Tukeys (see [here](Error Handling))

Configuration

Example:

{  
    "FlowDifference": {
        "scheduler_options": {},
        "worker_options": {
            "requests": {
                "in_metric": "service1.requests.out*",
                "out_metric": "service2.requests.in*",
                "stale": 600,
                "error_params": {
                    "scaling": 2,
                    "minimal_upper_threshold": 0.1,
                    "minimal_lower_threshold": -0.1
                }
            }
        }
    }
}

• in_metric Regex for Metric1
• out_metric Regex for Metric2
• stale stale period for metric data in seconds - if exceeded set error flag
• [optional/default: 1.5] error_params:scaling IQR Scaling Factor
• [optional] error_params:minimal_lower_threshold minimum threshold which needs to be fallen below before setting the error flag (both methods)
• [optional] error_params:minimal_upper_threshold minimum threshold which has to exceeded before setting the error flag (both methods)