Small script to simulate rack-aware Cassandra reads/writes.

simulate_cassandra_load.py

import random

original = [
        ["RAC1"],
        ["az-A-0", "az-A-1"],
        ["az-B-0", "az-B-1"],
        ["az-C-0", "az-C-1"],
        ]
remove_rac1 = [
        ["az-A-0", "az-A-1"],
        ["az-B-0", "az-B-1"],
        ["az-C-0", "az-C-1"],
        ]
add_new_node = [
        ["RAC1"],
        ["az-A-0", "az-A-1", "az-A-2"],
        ["az-B-0", "az-B-1"],
        ["az-C-0", "az-C-1"],
        ]

nodes = add_new_node

a = dict()
consistency = 1
N = 10000

for i in range(N):
    random.shuffle(nodes)
    racks = nodes[:consistency]
    for rack in racks:
        node = random.choice(rack)
        if node not in a:
            a[node] = 0
        a[node] += 1


print "Consistency:", consistency
print
for k, v in a.items():
    print "{0}: {1}%".format(k, 100.*v/(consistency*N))