def run_taro_experiments():
experiments = get_taro_experiments()
completed_experiment_procs = []
logging.info('Going to run {} experiments.'.format(len(experiments)))
running_experiment = 1
for experiment in experiments.values():
try:
print('Running experiments {}/{}'.format(running_experiment,
len(experiments)))
cctestbed.run_local_command(
'/opt/bess/bessctl/bessctl daemon stop')
proc = experiment.run()
completed_experiment_procs.append(proc)
running_experiment += 1
except Exception as e:
print('ERROR RUNNING EXPERIMENT: {}'.format(e))
for proc in completed_experiment_procs:
logging.info('Waiting for subprocess to finish PID={}'.format(
proc.pid))
proc.wait()
if proc.returncode != 0:
logging.warning('Error running cmd PID={}'.format(proc.pid))
def start_bess_for_local_video(exp, duration):
with ExitStack() as stack:
exp._run_tcpdump('server', stack)
exp._run_tcpdump('server', stack, capture_http=True)
cctestbed.stop_bess()
stack.enter_context(exp._run_bess(
ping_source='client',
skip_ping=False,
bess_config_name='active-middlebox-pmd-fairness'))
# give bess time to start
time.sleep(5)
exp._show_bess_pipeline()
stack.enter_context(exp._run_bess_monitor())
util.start_apache_server(exp.flows[0])
video_flow = start_single_local_video_flow(exp.flows[0], exp, stack)
logging.info('Waiting for flow to finish')
# wait for flow to finish
video_flow._wait()
# add add a time buffer before finishing up experiment
logging.info('Video flow finished')
# add add a time buffer before finishing up experiment
time.sleep(5)
exp._show_bess_pipeline()
cmd = '/opt/bess/bessctl/bessctl command module queue0 get_status EmptyArg'
print(cctestbed.run_local_command(cmd))
util.stop_local_server_and_cleanup(exp)
def get_ping_rtt(instance_ip):
#cmd = "nping --icmp -v-1 -H -c 5 {} | grep -oP 'Avg rtt:\s+\K.*(?=ms)'".format(instance_ip)
cmd = 'ping -c 5 {} | tail -1 | awk "{{print $4}}" '.format(instance_ip),
line = cctestbed.run_local_command(cmd, shell=True)
print("get ping rtt: ", line)
rtt = float(line.split('=')[-1].split('/')[1])
return rtt
def get_website_ip(url):
url_parts = list(urlsplit(url.strip()))
hostname = url_parts[1]
ip_addrs = cctestbed.run_local_command(
"nslookup {} | awk '/^Address: / {{ print $2 ; exit }}'".format(hostname), shell=True)
ip_addr = ip_addrs.split('\n')[0]
if ip_addr.strip() == '':
raise ValueError('Could not find IP addr for {}'.format(url))
return ip_addr
def get_video_server_host(url):
videoUrls = cctestbed.run_local_command(
"youtube-dl --youtube-skip-dash-manifest -g {}".format(url),
shell=True)
# print(videoUrls)
video_url = videoUrls.split('\n')[0]
audio_url = videoUrls.split('\n')[1]
print("Video URL: ", video_url)
print("Audio URL: ", audio_url)
url_parts = list(urlsplit(video_url.strip()))
hostname = url_parts[1]
print("video hostname {}", hostname)
return hostname
def get_video_server_ip(hostname):
video_ip_addrs = cctestbed.run_local_command(
"nslookup {} | awk '/^Address: / {{ print $2 ; exit }}'".format(
hostname),
shell=True)
print("Video IPs: ", video_ip_addrs)
video_ip_addr = video_ip_addrs.split('\n')[0]
print("ipaddress of video server {}", video_ip_addr)
if video_ip_addr.strip() == '':
raise ValueError('Could not find IP addr for {}'.format(video_url))
return video_ip_addr
def start_bess_for_combinatory_flows(experiment, includeWebsite, duration):
with ExitStack() as stack:
util.prerequisite_for_combinatory_tests(experiment, stack,
includeWebsite)
ssh_client = util.get_ssh_client_for_server_node(experiment)
with ssh_client as ssh_client:
local_web_service_required, single_local_flow_details = util.isApacheNeeded(
experiment)
print("local_web_service_required & single_local_flow_details>",
local_web_service_required, single_local_flow_details)
if local_web_service_required:
print("Start Apache")
util.start_apache_server(single_local_flow_details)
#TODO:Measure RTT of third party services again after remaining delay is set
#iperf
flowImpl.start_iperf_flows(experiment, stack)
#Web video
contains_webvideo_flows = flowImpl.start_web_video_flows(
experiment, stack)
#Local video
contains_localvideo_flows = flowImpl.start_local_video_flows(
experiment, stack)
#Local website
contains_local_website_flows = flowImpl.start_local_website_flows(
ssh_client, experiment, stack)
#Website
contains_website_flows = flowImpl.start_website_flows(
ssh_client, experiment, stack)
time.sleep(duration + 5)
#Save website info onto a file
if contains_webvideo_flows:
clean_up_web_video(experiment, duration)
if contains_webvideo_flows or contains_website_flows:
util.write_webdata_to_log(experiment, duration)
if contains_localvideo_flows or contains_local_website_flows:
util.stop_local_server_and_cleanup(experiment)
experiment._show_bess_pipeline()
cmd = '/opt/bess/bessctl/bessctl command module queue0 get_status EmptyArg'
print(cctestbed.run_local_command(cmd))
def start_bess_for_iperf(exp, duration):
with ExitStack() as stack:
exp._run_tcpdump('server', stack)
cctestbed.stop_bess()
stack.enter_context(exp._run_bess(ping_source='client',
skip_ping=False,
bess_config_name='active-middlebox-pmd-fairness'))
# give bess time to start
time.sleep(5)
exp._show_bess_pipeline()
stack.enter_context(exp._run_bess_monitor())
start_iperf_flows(exp, stack)
time.sleep(duration+5)
exp._show_bess_pipeline()
cmd = '/opt/bess/bessctl/bessctl command module queue0 get_status EmptyArg'
print(cctestbed.run_local_command(cmd))
def start_bess_for_website(exp, duration, web_data):
with ExitStack() as stack:
print(web_data)
stack.enter_context(util.add_dnat_rule(exp, web_data['url_ip']))
stack.enter_context(util.add_route(exp, web_data['url_ip']))
stack.enter_context(util.add_dns_rule(exp, web_data['website'], web_data['url_ip']))
exp._run_tcpdump('server', stack)
# run the flow
# turns out there is a bug when using subprocess and Popen in Python 3.5
# so skip ping needs to be true
# https://bugs.python.org/issue27122
cctestbed.stop_bess()
stack.enter_context(exp._run_bess(ping_source='server', skip_ping=False, bess_config_name='active-middlebox-pmd-fairness'))
# give bess some time to start
time.sleep(5)
exp._show_bess_pipeline()
stack.enter_context(exp._run_bess_monitor())
stack.enter_context(exp._run_rtt_monitor())
ssh_client = cctestbed.get_ssh_client(exp.server.ip_wan, exp.server.username, key_filename=exp.server.key_filename)
with ssh_client as ssh_client:
start_website_flows(ssh_client, exp, stack)
# exit_status = stdout.channel.recv_exit_status()
time.sleep(duration+5)
# flow_end_time = time.time()
logging.info('Flow ran for {} seconds'.format(duration+5))
exp._show_bess_pipeline()
cmd = '/opt/bess/bessctl/bessctl command module queue0 get_status EmptyArg'
print(cctestbed.run_local_command(cmd))
logging.info('Dumping website data to log: {}'.format(exp.logs['website_log']))
with open(exp.logs['website_log'], 'w') as f:
website_info = {}
website_info['website'] = web_data['website']
website_info['url'] = web_data['url']
website_info['website_rtt'] = web_data['website_rtt']
website_info['experiment_rtt'] = web_data['experiment_rtt']
website_info['delay'] = web_data['delay']
website_info['url_ip'] = web_data['url_ip']
website_info['flow_runtime'] = duration+5
json.dump(website_info, f)
def run_experiment(website, url, btlbw=10, queue_size=128, rtt=35, force=False):
experiment_name = '{}bw-{}rtt-{}q-{}'.format(btlbw, rtt, queue_size, website)
if not force and is_completed_experiment(experiment_name):
return (None, '')
logging.info('Creating experiment for website: {}'.format(website))
url_ip = get_website_ip(url)
logging.info('Got website IP: {}'.format(url_ip))
website_rtt = int(float(get_nping_rtt(url_ip)))
logging.info('Got website RTT: {}'.format(website_rtt))
if website_rtt >= rtt:
logging.warning('Skipping experiment with website RTT {} >= {}'.format(
website_rtt, rtt))
return (-1, '')
client = HOST_CLIENT_TEMPLATE
client['ip_wan'] = url_ip
client = cctestbed.Host(**client)
server = HOST_SERVER
server_nat_ip = HOST_CLIENT.ip_wan #'128.104.222.182' taro
server_port = 5201
client_port = 5555
flow = {'ccalg': 'reno',
'end_time': 60,
'rtt': rtt - website_rtt,
'start_time': 0}
flows = [cctestbed.Flow(ccalg=flow['ccalg'], start_time=flow['start_time'],
end_time=flow['end_time'], rtt=flow['rtt'],
server_port=server_port, client_port=client_port,
client_log=None, server_log=None, kind='website',
client=client)]
exp = cctestbed.Experiment(name=experiment_name,
btlbw=btlbw,
queue_size=queue_size,
flows=flows, server=server, client=client,
config_filename='experiments-all-ccalgs-aws.yaml',
server_nat_ip=server_nat_ip)
logging.info('Running experiment: {}'.format(exp.name))
# make sure tcpdump cleaned up
logging.info('Making sure tcpdump is cleaned up')
with cctestbed.get_ssh_client(
exp.server.ip_wan,
username=exp.server.username,
key_filename=exp.server.key_filename) as ssh_client:
cctestbed.exec_command(
ssh_client,
exp.client.ip_wan,
'sudo pkill -9 tcpdump')
with ExitStack() as stack:
# add DNAT rule
stack.enter_context(add_dnat_rule(exp, url_ip))
# add route to URL
stack.enter_context(add_route(exp, url_ip))
# add dns entry
stack.enter_context(add_dns_rule(exp, website, url_ip))
exp._run_tcpdump('server', stack)
# run the flow
# turns out there is a bug when using subprocess and Popen in Python 3.5
# so skip ping needs to be true
# https://bugs.python.org/issue27122
cctestbed.stop_bess()
stack.enter_context(exp._run_bess(ping_source='server', skip_ping=False))
# give bess some time to start
time.sleep(5)
exp._show_bess_pipeline()
stack.enter_context(exp._run_bess_monitor())
stack.enter_context(exp._run_rtt_monitor())
with cctestbed.get_ssh_client(exp.server.ip_wan,
exp.server.username,
key_filename=exp.server.key_filename) as ssh_client:
filename = os.path.basename(url)
if filename.strip() == '':
logging.warning('Could not get filename from URL')
start_flow_cmd = 'timeout 65s wget --no-check-certificate --no-cache --delete-after --connect-timeout=10 --tries=3 --bind-address {} -P /tmp/ "{}" || rm -f /tmp/{}.tmp*'.format(exp.server.ip_lan, url, filename)
# won't return until flow is done
flow_start_time = time.time()
_, stdout, _ = cctestbed.exec_command(ssh_client, exp.server.ip_wan, start_flow_cmd)
exit_status = stdout.channel.recv_exit_status()
flow_end_time = time.time()
logging.info('Flow ran for {} seconds'.format(flow_end_time - flow_start_time))
exp._show_bess_pipeline()
cmd = '/opt/bess/bessctl/bessctl command module queue0 get_status EmptyArg'
print(cctestbed.run_local_command(cmd))
logging.info('Dumping website data to log: {}'.format(exp.logs['website_log']))
with open(exp.logs['website_log'], 'w') as f:
website_info = {}
website_info['website'] = website
website_info['url'] = url
website_info['website_rtt'] = website_rtt
website_info['url_ip'] = url_ip
website_info['flow_runtime'] = flow_end_time - flow_start_time
json.dump(website_info, f)
if exit_status != 0:
if exit_status == 124: # timeout exit status
print('Timeout. Flow longer than 65s.')
logging.warning('Timeout. Flow longer than 65s.')
else:
logging.error(stdout.read())
raise RuntimeError('Error running flow.')
proc = exp._compress_logs_url()
return (proc, '{}-{}'.format(experiment_name, exp.exp_time))
def run_rtt_monitor(url_ip):
cmd = "nping --delay 5s {} > {} &".format(url_ip, '')
rtt = cctestbed.run_local_command(cmd, shell=True)
return rtt
def get_nping_rtt(url_ip):
cmd = "nping -v-1 -H -c 5 {} | grep -oP 'Avg rtt:\s+\K.*(?=ms)'".format(url_ip)
rtt = cctestbed.run_local_command(cmd, shell=True)
return rtt
def run_experiment(website1,
url1,
website2,
url2,
btlbw=10,
queue_size=128,
rtt=35,
force=False):
experiment_name = '{}bw-{}rtt-{}q-{}-{}'.format(btlbw, rtt, queue_size,
website1, website2)
if not force and is_completed_experiment(experiment_name):
return
else:
if ran_experiment_today(experiment_name):
return
logging.info('Creating experiment for website1: {} website2: {}'.format(
website1, website2))
url_ip1 = get_website_ip(url1)
url_ip2 = get_website_ip(url2)
logging.info('Got website1 IP: {} website2 IP: {}'.format(
url_ip1, url_ip2))
website_rtt1 = int(float(get_nping_rtt(url_ip1)))
website_rtt2 = int(float(get_nping_rtt(url_ip2)))
logging.info('Got website1 RTT: {} website2 RTT: {}'.format(
website_rtt1, website_rtt2))
if website_rtt1 >= rtt:
logging.warning(
'Skipping experiment with website1 RTT {} >= {}'.format(
website_rtt1, rtt))
return -1
elif website_rtt2 >= rtt:
logging.warning(
'Skipping experiment with website2 RTT {} >= {}'.format(
website_rtt2, rtt))
return -1
client = HOST_CLIENT_TEMPLATE
# TODO: Which IP should be used for client?
client['ip_wan'] = url_ip1
client = cctestbed.Host(**client)
server = HOST_SERVER
server_nat_ip = HOST_CLIENT.ip_wan #'128.104.222.182' taro
server_port = 5201
client_port = 5555
flow1 = {
'ccalg': 'reno',
'end_time': 60,
'rtt': rtt - website_rtt1,
'start_time': 0
}
flow2 = {
'ccalg': 'reno',
'end_time': 60,
'rtt': rtt - website_rtt2,
'start_time': 0
}
flows = [
cctestbed.Flow(ccalg=flow1['ccalg'],
start_time=flow1['start_time'],
end_time=flow1['end_time'],
rtt=flow1['rtt'],
server_port=server_port,
client_port=client_port,
client_log=None,
server_log=None),
cctestbed.Flow(ccalg=flow2['ccalg'],
start_time=flow2['start_time'],
end_time=flow2['end_time'],
rtt=flow2['rtt'],
server_port=server_port,
client_port=client_port,
client_log=None,
server_log=None),
]
exp = cctestbed.Experiment(
name=experiment_name,
btlbw=btlbw,
queue_size=queue_size,
flows=flows,
server=server,
client=client,
config_filename='experiments-all-ccalgs-aws.yaml',
server_nat_ip=server_nat_ip)
logging.info('Running experiment: {}'.format(exp.name))
# make sure tcpdump cleaned up
logging.info('Making sure tcpdump is cleaned up')
with cctestbed.get_ssh_client(
exp.server.ip_wan,
username=exp.server.username,
key_filename=exp.server.key_filename) as ssh_client:
cctestbed.exec_command(ssh_client, exp.client.ip_wan,
'sudo pkill -9 tcpdump')
with ExitStack() as stack:
# add DNAT rule
stack.enter_context(add_dnat_rule(exp, url_ip1))
stack.enter_context(add_dnat_rule(exp, url_ip2))
# add route to URL
stack.enter_context(add_route(exp, url_ip1))
stack.enter_context(add_route(exp, url_ip2))
# add dns entry
stack.enter_context(add_dns_rule(exp, website1, url_ip1))
stack.enter_context(add_dns_rule(exp, website2, url_ip2))
exp._run_tcpdump('server', stack)
# run the flow
# turns out there is a bug when using subprocess and Popen in Python 3.5
# so skip ping needs to be true
# https://bugs.python.org/issue27122
cctestbed.stop_bess()
stack.enter_context(
exp._run_bess(ping_source='server', skip_ping=False))
# give bess some time to start
time.sleep(5)
exp._show_bess_pipeline()
stack.enter_context(exp._run_bess_monitor())
stack.enter_context(exp._run_rtt_monitor())
with cctestbed.get_ssh_client(
exp.server.ip_wan,
exp.server.username,
key_filename=exp.server.key_filename) as ssh_client:
filename1 = os.path.basename(url1)
filename2 = os.path.basename(url2)
if filename1.strip() == '':
logging.warning('Could not get filename from URL 1')
if filename2.strip() == '':
logging.warning('Could not get filename from URL 2')
# Start first flow in background and second in foreground
start_flow_cmd1 = 'timeout 65s wget --no-cache --delete-after --connect-timeout=10 --tries=3 --bind-address {} -P /tmp/ {} || rm -f /tmp/{}.tmp* &'.format(
exp.server.ip_lan, url1, filename1)
start_flow_cmd2 = 'timeout 65s wget --no-cache --delete-after --connect-timeout=10 --tries=3 --bind-address {} -P /tmp/ {} || rm -f /tmp/{}.tmp*'.format(
exp.server.ip_lan, url2, filename2)
# won't return until flow is done
flow_start_time = time.time()
_, _, _ = cctestbed.exec_command(ssh_client, exp.server.ip_wan,
start_flow_cmd1)
_, stdout, _ = cctestbed.exec_command(ssh_client,
exp.server.ip_wan,
start_flow_cmd2)
exit_status = stdout.channel.recv_exit_status()
flow_end_time = time.time()
logging.info('Flow ran for {} seconds'.format(flow_end_time -
flow_start_time))
exp._show_bess_pipeline()
cmd = '/opt/bess/bessctl/bessctl command module queue0 get_status EmptyArg'
print(cctestbed.run_local_command(cmd))
if exit_status != 0:
if exit_status == 124: # timeout exit status
print('Timeout. Flow longer than 65s.')
logging.warning('Timeout. Flow longer than 65s.')
else:
logging.error(stdout.read())
raise RuntimeError('Error running flow.')
proc = exp._compress_logs_url()
return (proc, exp.tar_filename, experiment_name)
def load_experiments(experiment_name_patterns, remote=True, force_local=False,
remote_username=REMOTE_USERNAME, remote_ip=REMOTE_IP_ADDR,
load_queue=False, clean=False, parallel=True,
min_num_files=0, min_date=None, remove_duplicates=True):
"""Load all experiments into experiment analyzers
experiment_name_pattern : list of str
Should be a pattern that will be called
with '{}.tar.gz'.format(experiment_name_pattern)
remote : bool, (default: True)
If True, look for experiments remotely.
If False, don't look for experiments remotely,
only locally.
force_local : bool, (default: False)
If True, always look for local experiments.
If False, only look for local experiments,
if no remote experiments are found.
clean: bool
If True, delete all local files matching this exp_name_pattern
before downloading again.
parallel: bool
If True, run download for experiments in parallel
min_num_files: int
If greater than 0, then expected to get atleast this number of files
min_date: string
Only return experiments with equal to or large than the expected date
remove_duplicates: bool
Remove experiments with the same name, keeping the most recent one
"""
assert(type(experiment_name_patterns) is list)
tarfile_remotepaths = []
# i feel like this code is too dangerous since there is a rm command ...
if clean:
for experiment_name_pattern in experiment_name_patterns:
print('Deleting local files matching experiment pattern: {}'.format(experiment_name_pattern))
run_local_command('rm {}.h5'.format(os.path.join(DATAPATH_PROCESSED, experiment_name_pattern)))
if remote:
print('Searching for experiments on remote machine: {}'.format(remote_ip))
with get_ssh_client(ip_addr=remote_ip, username=remote_username) as ssh_client:
'''
for experiment_name_pattern in experiment_name_patterns:
_, stdout, _ = ssh_client.exec_command(
'ls -1 /tmp/{}.tar.gz'.format(experiment_name_pattern))
tarfile_remotepaths += [filename.strip()
for filename in stdout.readlines()]
'''
cmd = 'ls -1 ' + ' '.join(['/tmp/{}.tar.gz']*len(experiment_name_patterns)).format(*experiment_name_patterns)
print(cmd)
_, stdout, _ = ssh_client.exec_command(cmd)
tarfile_remotepaths += [filename.strip() for filename in stdout.readlines()]
print('Found {} experiment(s) on remote machine: {}'.format(
len(tarfile_remotepaths), tarfile_remotepaths))
else:
print('Not searching remote machine for experiments.')
if force_local or len(tarfile_remotepaths) == 0:
num_local_files = 0
for experiment_name_pattern in experiment_name_patterns:
local_filepaths = glob.glob(os.path.join(DATAPATH_RAW,
experiment_name_pattern +'.tar.gz'))
tarfile_remotepaths += local_filepaths
num_local_files += len(local_filepaths)
if len(tarfile_remotepaths) == 0:
raise ValueError(('Found no experiments on remote or local machine '
'{} with name pattern {}').format(
remote_ip, experiment_name_pattern))
if num_local_files > 0:
print('Found {} experiment(s) on local machine: {}'.format(num_local_files,
tarfile_remotepaths[-num_local_files:]))
else:
print('Found 0 experiment(s) on local machines.')
if min_date is not None:
# copy file so we iterate over list and modify it
remotepaths = tarfile_remotepaths[:]
num_wrong_date = 0
for remotepath in remotepaths:
date = os.path.basename(remotepath).split('-')[-1]
if date < min_date:
num_wrong_date += 1
tarfile_remotepaths.remove(remotepath)
if num_wrong_date > 0:
print('Found {} experiment(s) with date smaller than {}.'.format(num_wrong_date, min_date))
if remove_duplicates:
# keep only most recent experiments with the same name
tmp = pd.DataFrame(tarfile_remotepaths)
num_duplicates = len(tmp)
tarfile_remotepaths = tmp.loc[tmp[0].sort_values().apply(lambda x: '-'.join(x.split('-')[:-1])).drop_duplicates(keep='last').index][0].tolist()
num_duplicates = num_duplicates - len(tarfile_remotepaths)
if num_duplicates > 0:
print('Found {} experiment(s) with duplicate prefixes.'.format(num_duplicates))
if min_num_files > 0:
if len(tarfile_remotepaths) < min_num_files:
print('Wanted min number of {} experiment(s), but only found {}.'.format(min_num_files, len(tarfile_remotepaths)))
tarfile_remotepaths = []
#experiments = {}
num_proc = 10
num_tarfiles = len(tarfile_remotepaths)
num_tarfiles_per_process = int(num_tarfiles / num_proc) + 1
if parallel and num_tarfiles > 1:
with mp.Pool(num_proc) as pool:
analyzers = pool.starmap(get_experiment, zip(tarfile_remotepaths,
it.repeat(remote_ip, num_tarfiles),
it.repeat(remote_username, num_tarfiles),
it.repeat(load_queue, num_tarfiles)),
chunksize=num_tarfiles_per_process)
else:
analyzers = [get_experiment(tarfile_remotepath, remote_ip, remote_username, load_queue) for tarfile_remotepath in tarfile_remotepaths]
experiment_analyzers = ExperimentAnalyzers()
for analyzer in analyzers:
experiment_analyzers['{}-{}'.format(analyzer.experiment.name,
analyzer.experiment.exp_time)] = analyzer
return experiment_analyzers
def _create_hdf_queue(self, raw_queue_log_tarpath, raw_queue_log_localpath, processed_queue_log_localpath):
# haven't created HDF5 store yet; create it now
find_bad_lines_cmd = 'grep ^.*,.*,.*,.*,.*,.*,.*,.*,.*$ {} -v -n'.format(raw_queue_log_localpath)
badlines = run_local_command(find_bad_lines_cmd, shell=False).split('\n')
if len(badlines) >= 1 and badlines[0] != '':
sort_cmd = 'sort -k 2 -o {} {}'.format(raw_queue_log_localpath, raw_queue_log_localpath)
print('Found {} bad lines:\n {}'.format(len(badlines), badlines))
else:
tmp_queue_filename = raw_queue_log_localpath + '.tmp'
sort_cmd = 'sort -k 2 -o {} {} && grep ^.*,.*,.*,.*,.*,.*,.*,.*,.*$ {} > {} && mv {} {} '.format(
raw_queue_log_localpath, raw_queue_log_localpath, raw_queue_log_localpath, tmp_queue_filename, tmp_queue_filename, raw_queue_log_localpath)
with untarfile(self.experiment.tarfile_localpath, raw_queue_log_tarpath, postprocess_cmd=sort_cmd) as f:
with pd.HDFStore(processed_queue_log_localpath, mode='w') as store:
df = pd.read_csv(f, names = ['dequeued',
'time',
'src',
'seq',
'datalen',
'size',
'dropped',
'queued',
'batch'],
converters = {'seq': tohex,
'src': tohex},
dtype={'dequeued': bool,
'time': np.uint64,
'datalen': np.uint16,
'size': np.uint32,
'dropped':bool,
'queued': np.uint16,
'batch': np.uint16}, skip_blank_lines=True)
df['seq'] = df['seq'].astype( np.uint32)
df['src'] = df['src'].astype( np.uint16)
#chunk['time'] = pd.to_datetime(chunk['time'], infer_datetime_format=True, unit='ns')
df['lineno'] = df.index + 1
df = df.set_index('time')
df_enq = (pd
.get_dummies(df[(df.dequeued==0) & (df.dropped==0)]['src'])
.astype(np.uint8))
df_deq = (pd
.get_dummies(df[df.dequeued==1]['src'])
.replace(1,-1)
.astype(np.int8))
df_flows = (df_enq
.append(df_deq)
.sort_index()
.cumsum()
.fillna(0)
.astype(np.uint32)) #cumsum may return negative numbers and if we convert that to uint32, it will return large numbers
df = (df
#.reset_index()
.join(df_flows)
.sort_index()
.ffill())
df.index = pd.to_datetime(df.index,
infer_datetime_format=True,
unit='ns')
#df = df.set_index('time')
store.append('df_queue',
df,
format='table',
data_columns=['src', 'dropped', 'dequeued'])
