def data_generation_post():
global coordinator_db, username, parallel_ssh_cmd, generate_command,\
generate_data_size
print "Data generation post"
if bottle.request.POST.generate_data:
print "Generating data"
data_size = int(bottle.request.POST.data_size)
unit = bottle.request.POST.unit
print "Generating %d%s of data..." % (data_size, unit)
files_per_disk = int(bottle.request.POST.files_per_disk)
generate_data_size = unitconversion.convert(data_size, unit, "B")
generate_command = parallel_ssh_cmd[
"%s --no_sudo -g -n%d %d%s gensort" % (
os.path.join(SCRIPT_DIR, os.pardir,
"generate_graysort_inputs.py"),
files_per_disk, data_size, unit)] & BG
elif bottle.request.POST.abort_data_generation:
print "Aborting data generation"
pkill["-f"]["generate_graysort_inputs.py"]()
generate_command = None
elif bottle.request.POST.wipe_input:
allowed = bottle.request.POST.getall('allow_wipe_data')
if len(allowed) == 1 and allowed[0] == "True":
print "Wiping input data"
parallel_ssh_cmd[
"%s input %s" % (
os.path.join(SCRIPT_DIR, os.pardir,
"clear_disks.sh"), disk_mountpoint)]()
elif bottle.request.POST.wipe_intermediate_output:
allowed = bottle.request.POST.getall('allow_wipe_data')
if len(allowed) == 1 and allowed[0] == "True":
print "Wiping intermediate/output data"
parallel_ssh_cmd[
"%s noinput %s" % (
os.path.join(SCRIPT_DIR, os.pardir,
"clear_disks.sh"), disk_mountpoint)]()
elif bottle.request.POST.wipe_all:
allowed = bottle.request.POST.getall('allow_wipe_data')
if len(allowed) == 1 and allowed[0] == "True":
print "Wiping all data"
parallel_ssh_cmd[
"%s all %s" % (
os.path.join(SCRIPT_DIR, os.pardir,
"clear_disks.sh"), disk_mountpoint)]()
# Call the normal /data_generation method.
return data_generation()
def data_generation_post():
global coordinator_db, username, parallel_ssh_cmd, generate_command,\
generate_data_size
print "Data generation post"
if bottle.request.POST.generate_data:
print "Generating data"
data_size = int(bottle.request.POST.data_size)
unit = bottle.request.POST.unit
print "Generating %d%s of data..." % (data_size, unit)
files_per_disk = int(bottle.request.POST.files_per_disk)
generate_data_size = unitconversion.convert(data_size, unit, "B")
skewed_data = bool(bottle.request.POST.skewed_data)
replica_data = bool(bottle.request.POST.replica_data)
extra_args = ""
if skewed_data:
extra_args += "-s "
if replica_data:
extra_args += "-r 2 "
generate_command = parallel_ssh_cmd[
"%s --no_sudo %s-g -n%d %d%s gensort" %
(os.path.join(SCRIPT_DIR, os.pardir, "generate_graysort_inputs.py"
), extra_args, files_per_disk, data_size, unit)] & BG
elif bottle.request.POST.abort_data_generation:
print "Aborting data generation"
pkill["-f"]["generate_graysort_inputs.py"]()
generate_command = None
elif bottle.request.POST.wipe_input:
allowed = bottle.request.POST.getall('allow_wipe_data')
if len(allowed) == 1 and allowed[0] == "True":
print "Wiping input data"
parallel_ssh_cmd["%s input %s" % (os.path.join(
SCRIPT_DIR, os.pardir, "clear_disks.sh"), disk_mountpoint)]()
elif bottle.request.POST.wipe_intermediate_output:
allowed = bottle.request.POST.getall('allow_wipe_data')
if len(allowed) == 1 and allowed[0] == "True":
print "Wiping intermediate/output data"
parallel_ssh_cmd["%s noinput %s" % (os.path.join(
SCRIPT_DIR, os.pardir, "clear_disks.sh"), disk_mountpoint)]()
elif bottle.request.POST.wipe_all:
allowed = bottle.request.POST.getall('allow_wipe_data')
if len(allowed) == 1 and allowed[0] == "True":
print "Wiping all data"
parallel_ssh_cmd["%s all %s" % (os.path.join(
SCRIPT_DIR, os.pardir, "clear_disks.sh"), disk_mountpoint)]()
# Call the normal /data_generation method.
return data_generation()
def load_constraints(filename):
with open(filename, 'r') as fp:
constraints = json.load(fp)
for phase in constraints:
for stage in constraints[phase][consts.STAGES_KEY]:
stage_info = constraints[phase][consts.STAGES_KEY][stage]
if consts.IO_BOUND_KEY in stage_info:
(rate, sep, unit) = stage_info[
consts.MAX_RATE_PER_WORKER_KEY].partition(' ')
rate = float(rate)
rate_Bps = convert(rate, unit, "Bps")
stage_info[consts.MAX_RATE_PER_WORKER_KEY] = rate_Bps
return constraints
def load_constraints(filename):
with open(filename, 'r') as fp:
constraints = json.load(fp)
for phase in constraints:
for stage in constraints[phase][consts.STAGES_KEY]:
stage_info = constraints[phase][consts.STAGES_KEY][stage]
if consts.IO_BOUND_KEY in stage_info:
(rate, sep, unit
) = stage_info[consts.MAX_RATE_PER_WORKER_KEY].partition(' ')
rate = float(rate)
rate_Bps = convert(rate, unit, "Bps")
stage_info[consts.MAX_RATE_PER_WORKER_KEY] = rate_Bps
return constraints
def _make_cell_readable(self, cell_value):
# Make sure we don't try to convert non-numeric types
if type(cell_value) not in [int, float]:
return cell_value
units = ["B", "KiB", "MiB", "GiB", "TiB"]
if cell_value > 0:
# floor(log_2(X)) gives the closest integer power of 2 to
# X. Dividing it by 10 gives the unit (2^0 = bytes, 2^10 =
# kibibytes, and so on, so 0 = bytes, 1 = kibibytes, etc.)
cell_value_dest_unit = units[min(
int(math.floor(math.log(cell_value, 2) / 10)),
len(units) - 1)]
cell_value = "%.2f %s" % (convert(
cell_value, "B", cell_value_dest_unit), cell_value_dest_unit)
else:
cell_value = "0 B"
return cell_value
def _make_cell_readable(self, cell_value):
# Make sure we don't try to convert non-numeric types
if type(cell_value) not in [int, float]:
return cell_value
units = ["B", "KiB", "MiB", "GiB", "TiB"]
if cell_value > 0:
# floor(log_2(X)) gives the closest integer power of 2 to
# X. Dividing it by 10 gives the unit (2^0 = bytes, 2^10 =
# kibibytes, and so on, so 0 = bytes, 1 = kibibytes, etc.)
cell_value_dest_unit = units[
min(int(math.floor(math.log(cell_value, 2) / 10)),
len(units) - 1)]
cell_value = "%.2f %s" % (
convert(cell_value, "B", cell_value_dest_unit),
cell_value_dest_unit)
else:
cell_value = "0 B"
return cell_value
def postprocess_rate_data(data):
display_data = {}
disk_data = {}
for data_key in sorted(data.keys()):
(job, phase, epoch) = data_key
epoch_data = data[data_key]
if phase is None and epoch is None:
disk_data = epoch_data["disks"]
for key, value in disk_data.items():
disk_data[key] = value
continue
epoch_display_data = {}
epoch_display_data["job"] = job
epoch_display_data["phase"] = phase
epoch_display_data["epoch"] = epoch
host_data = epoch_data["hosts"]
stats_data = epoch_data["stats"]
epoch_display_data["intermediate_disks"] = sum(
(disk_data[host]["num_intermediate_disks"] for host in disk_data))
epoch_display_data["num_nodes"] = len(host_data)
input_size = stats_data["input_size"]
if "output_size" in stats_data:
output_size = stats_data["output_size"]
else:
output_size = 0
max_size = max(input_size, output_size)
for (size_name, size) in [("input_size", input_size),
("output_size", output_size),
("max_size", max_size)]:
epoch_display_data[size_name] = {}
for unit in ["B", "MB", "MiB", "GB", "Gb", "GiB", "TB"]:
epoch_display_data[size_name][unit] = convert(size, "B", unit)
epoch_display_data["total_time"] = {}
if "timestamps" in stats_data:
smallest_start_time = stats_data["timestamps"]["start"].min()
largest_stop_time = stats_data["timestamps"]["stop"].max()
total_time = largest_stop_time - smallest_start_time
for unit in ["s", "min"]:
epoch_display_data["total_time"][unit] = convert(
total_time, "us", unit)
else:
for unit in ["s", "min"]:
epoch_display_data["total_time"][unit] = 0
display_data[data_key] = epoch_display_data
return display_data
def postprocess_rate_data(data):
display_data = {}
disk_data = {}
for data_key in sorted(data.keys()):
(job, phase, epoch) = data_key
epoch_data = data[data_key]
if phase is None and epoch is None:
disk_data = epoch_data["disks"]
for key, value in disk_data.items():
disk_data[key] = value
continue
epoch_display_data = {}
epoch_display_data["job"] = job
epoch_display_data["phase"] = phase
epoch_display_data["epoch"] = epoch
host_data = epoch_data["hosts"]
stats_data = epoch_data["stats"]
epoch_display_data["intermediate_disks"] = sum(
(disk_data[host]["num_intermediate_disks"] for host in disk_data))
epoch_display_data["num_nodes"] = len(host_data)
input_size = stats_data["input_size"]
if "output_size" in stats_data:
output_size = stats_data["output_size"]
else:
output_size = 0
max_size = max(input_size, output_size)
for (size_name, size) in [("input_size", input_size),
("output_size", output_size),
("max_size", max_size)]:
epoch_display_data[size_name] = {}
for unit in ["B", "MB", "MiB", "GB", "Gb", "GiB", "TB"]:
epoch_display_data[size_name][unit] = convert(
size, "B", unit)
epoch_display_data["total_time"] = {}
if "timestamps" in stats_data:
smallest_start_time = stats_data["timestamps"]["start"].min()
largest_stop_time = stats_data["timestamps"]["stop"].max()
total_time = largest_stop_time - smallest_start_time
for unit in ["s", "min"]:
epoch_display_data["total_time"][unit] = convert(
total_time, "us", unit)
else:
for unit in ["s", "min"]:
epoch_display_data["total_time"][unit] = 0
display_data[data_key] = epoch_display_data
return display_data
