def __sample_day(time, signal, sampling_frequency, period_duration=hour(24), period=1):
# Remove Weekends/Sundays
tv = np.vectorize(__to_weekday)
time = tv(time)
indices = np.where(time < 5)
time = np.take(time, indices)
signal = np.ravel(np.take(signal, indices))
# Counters
elements_per_period = period_duration / sampling_frequency
periods = len(signal) / elements_per_period
signal = np.resize(signal, periods * elements_per_period)
# Filter extreme values using a sliding window forecast
outlier = np.ravel(np.where(signal > np.percentile(signal, 99)))
for i in outlier:
window = signal[max(i - 5, 0) : i]
if len(window) < 5:
window = np.append(window, signal[-5:])
signal[i] = np.mean(window)
# Reshape the signal (each profile in its on row)
signal = np.reshape(signal, (-1, elements_per_period))
# Extract signal of one profile
profile = signal[period]
# Establish frequency of 5 minutes
target_bucket_count = period_duration / minu(5)
factor = target_bucket_count / elements_per_period
profile = np.ravel(np.array(zip(*[profile for _ in xrange(factor)])))
# Single exponential signal smother
# profile = forecasting.single_exponential_smoother(profile, 0.9)[1]
# Calculate frequency
frequency = period_duration / len(profile)
return frequency, profile
def __to_weekday(value):
day = long(long(value) / hour(24)) % 7
return day
frequency = period_duration / len(profile)
return frequency, profile
def process_trace(connection, name, cycle_time, day):
timeSeries = connection.load(name)
time, demand = util.to_array(timeSeries)
return __sample_day(time, demand, timeSeries.frequency, cycle_time, period=day)
if __name__ == "__main__":
import matplotlib.pyplot as plt
# Connect with Times
connection = times_client.connect()
# Download TS
timeSeries = connection.load("RAW_SIS_299_cpu")
print timeSeries
time, demand = util.to_array(timeSeries)
signal, freq = __sample_day(time, demand, timeSeries.frequency, hour(24), 4)
# Plotting
fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot(signal)
plt.show()
print freq
def main():
tc.connect()
createBenchmarkProfile('PUSER_BENCHMARK_SMALL', 100, sec(3), hour(6))
createBenchmarkProfile('PUSER_BENCHMARK_MEDIUM', 150, sec(3), hour(6))
createBenchmarkProfile('PUSER_BENCHMARK_LARGE', 200, sec(3), hour(6))
tc.close()