def get_hourly_task_resource_request()

in cluster-trace-gpu-v2020/analysis/utils.py [0:0]

• 18 lines of code
• 7 McCabe index (conditional complexity)

def get_hourly_task_resource_request(df, metrics='cpu'): # df = dftjkix
    sum_df_list = []
    df = add_hour_date(df)
    if metrics == 'cpu':
        df['plan_resource'] = df.plan_cpu.apply(lambda x: x/100)
    elif metrics == 'gpu':
        df['plan_resource'] = df.plan_gpu.apply(lambda x: x/100)
    elif metrics == 'mem':
        df['plan_resource'] = df.plan_mem.apply(lambda x: x/1000)
    else:
        exit()
    # for day in sorted(df.dayofyear.unique()):
    for date in sorted(df.date.unique()):
        # tempdf = df[df.dayofyear==day]
        tempdf = df[df.date==date]
        res_df = tempdf.groupby('hour').sum()[['plan_resource']]
        res_df.rename(columns={'job_name':date}, inplace=True)
        sum_df_list.append(res_df.T)
    out_df = pd.DataFrame().append(sum_df_list)
    return out_df.dropna() # if a day contains hours of NaN, it is not a typical day