sharding_vs_replication_efficiency.py

import math
import matplotlib.pyplot as plt

#%%

# each tuple contains a mix of read workload and read workload
xs = [(18, 12), (18, 24), (10, 50), (50, 10)]

#%%
def model(s_w, s_r, c, rf, f):
  return math.ceil(s_w * rf / c + s_r * rf / (rf - f) / c)

#%%
from matplotlib.ticker import MaxNLocator
fig, ax = plt.subplots()  # Create a figure containing a single axes.
fig.set_size_inches(8, 5)
ax.set_xlabel('RF')
ax.set_ylabel('N')
ax.xaxis.set_major_locator(MaxNLocator(integer=True))
ax.yaxis.set_major_locator(MaxNLocator(integer=True))

for s_w, s_r in xs:
  px = []
  py = []
  for rf in range(1, 10):
    px.append(rf)
    py.append(model(s_w, s_r, 11, rf, 0))

  ax.plot(px, py, '-o', label="$\Sigma_w = {} \quad \Sigma_r = {}$)".format(s_w, s_r))

ax.legend()

plt.savefig("sharding_vs_replication_no_f.svg")

#%%
from matplotlib.ticker import MaxNLocator
fig, ax = plt.subplots()  # Create a figure containing a single axes.
fig.set_size_inches(8, 5)
ax.set_xlabel('RF')
ax.set_ylabel('N')
ax.xaxis.set_major_locator(MaxNLocator(integer=True))
ax.yaxis.set_major_locator(MaxNLocator(integer=True))

f = 1

for s_w, s_r in xs:
  px = []
  py = []
  for rf in range(f + 1, 10):
    px.append(rf)
    py.append(model(s_w, s_r, 11, rf, f))

  ax.plot(px, py, '-o', label="$\Sigma_w = {} \quad \Sigma_r = {} \quad F = {}$)".format(s_w, s_r, f))

ax.legend()

plt.savefig("sharding_vs_replication_f_1.svg")



#%%
from matplotlib.ticker import MaxNLocator
fig, ax = plt.subplots()  # Create a figure containing a single axes.
fig.set_size_inches(8, 5)
ax.set_xlabel('RF')
ax.set_ylabel('N')
ax.xaxis.set_major_locator(MaxNLocator(integer=True))
ax.yaxis.set_major_locator(MaxNLocator(integer=True))

f = 3

for s_w, s_r in xs:
  px = []
  py = []
  for rf in range(f + 1, 10):
    px.append(rf)
    py.append(model(s_w, s_r, 11, rf, f))

  ax.plot(px, py, '-o', label="$\Sigma_w = {} \quad \Sigma_r = {} \quad F = {}$)".format(s_w, s_r, f))

ax.legend()

plt.savefig("sharding_vs_replication_f_3.svg")