ch_part2_slices.py

from collections import defaultdict


def compute_slices(
    n: int,
    k: int,
    seed: Optional[int] = None,
) -> List[float]:
    np.random.seed(seed)
    tokens = np.random.rand(n * k)
    servers = np.random.choice(range(n), n * k)
    idxs = np.argsort(tokens)
    fractions = defaultdict(int)
    for idx_prev, idx in zip(idxs, idxs[1:]):
        server_prev = servers[idx_prev]
        diff = tokens[idx] - tokens[idx_prev]
        fractions[server_prev] += diff
        
    last = 1 - tokens[idxs[-1]]
    last += tokens[idxs[0]]
    fractions[servers[-1]] += last
    assert sum(fractions.values()) == 1
    return fractions

df_slices = pd.DataFrame(
    dict(
        k=k,
        exp_idx=exp_idx,
        server1=compute_slices(n=5, k=k, seed=exp_idx)[0],
    )
    for k in (1, 5, 10, 100, 1000, 10_000)
    for exp_idx in range(100)
).fillna(0)