Probabilistic analysis of failures

gistfile1.txt

import yaml
import sys
from decimal import Decimal, getcontext
from collections import defaultdict

# Set the precision
getcontext().prec = 100

def calculate_probability(N: int, M: int) -> float:
    return 1 - (Decimal(1)/Decimal(M))**N

# Read YAML input from stdin
input_yaml = yaml.safe_load(sys.stdin)

# Specify the number of nodes
M = 429

# Create a defaultdict of lists to store the Deployment groups by number of replicas
deployments_by_replicas = defaultdict(list)

# Group the Deployments by the number of replicas and store them in the defaultdict
for deployment in input_yaml['items']:
    if deployment['kind'] == "Deployment":
        replicas = deployment['spec']['replicas']
        # Ignore Deployments with zero replicas
        if replicas > 0:
            deployments_by_replicas[replicas].append(deployment['metadata']['name'])

# Print the header line of the histogram
print("Number of Replicas,Number of Services,Probability,Joint Probability No Failure")

# Calculate the no-failure probability for each group of Deployments, sorted by the number of replicas
total_no_failure_prob = Decimal(1.0)
for replicas in sorted(deployments_by_replicas.keys()):
    number_of_services = len(deployments_by_replicas[replicas])
    probability = calculate_probability(replicas, M)
    joint_probability_no_failure = probability ** number_of_services
    print(f"{replicas},{number_of_services},{float(probability)},{float(joint_probability_no_failure)}")
    total_no_failure_prob *= joint_probability_no_failure

# Calculate the total probability that a single node failure causes the loss of availability of at least one service
total_prob = 1 - total_no_failure_prob
print(f"Total probability that a single node failure causes the loss of availability of at least one service is {float(total_prob)}")