Revamp test result display

Instead of showing CLAT percentiles, which are very hard to interpret
and understand, instead use the main latency buckets.
This commit is contained in:
Joshua Boniface 2021-10-03 15:49:01 -04:00
parent bba73980de
commit 7c8ba56561
1 changed files with 103 additions and 73 deletions

View File

@ -1939,14 +1939,14 @@ def format_info_benchmark_json(config, benchmark_information):
ainformation.append('{}Storage Benchmark details:{}'.format(ansiprint.bold(), ansiprint.end()))
nice_test_name_map = {
"seq_read": "Sequential Read (4M blocks)",
"seq_write": "Sequential Write (4M blocks)",
"rand_read_4M": "Random Read (4M blocks)",
"rand_write_4M": "Random Write (4M blocks)",
"rand_read_4K": "Random Read (4K blocks)",
"rand_write_4K": "Random Write (4K blocks)",
"rand_read_4K_lowdepth": "Random Read (4K blocks, single-queue)",
"rand_write_4K_lowdepth": "Random Write (4K blocks, single-queue)",
"seq_read": "Sequential Read (4M blocks, queue depth 64)",
"seq_write": "Sequential Write (4M blocks, queue depth 64)",
"rand_read_4M": "Random Read (4M blocks, queue depth 64)",
"rand_write_4M": "Random Write (4M blocks queue depth 64)",
"rand_read_4K": "Random Read (4K blocks, queue depth 64)",
"rand_write_4K": "Random Write (4K blocks, queue depth 64)",
"rand_read_4K_lowdepth": "Random Read (4K blocks, queue depth 1)",
"rand_write_4K_lowdepth": "Random Write (4K blocks, queue depth 1)",
}
for test in benchmark_details:
@ -1961,29 +1961,37 @@ def format_info_benchmark_json(config, benchmark_information):
job_details = benchmark_details[test]['jobs'][0]
overall_label_length = 0
overall_column_length = 0
bandwidth_label_length = 0
bandwidth_column_length = 9
iops_column_length = 4
latency_column_length = 12
latency_percentile_label_length = 6
latency_percentile_column_length = 12
# Calculate the unified latency categories (in us)
latency_tree = list()
for field in job_details['latency_ns']:
bucket = str(int(field) / 1000)
latency_tree.append((bucket, job_details['latency_ns'][field]))
for field in job_details['latency_us']:
bucket = field
latency_tree.append((bucket, job_details['latency_us'][field]))
for field in job_details['latency_ms']:
# That one annoying one
if field == '>=2000':
bucket = '>=2000000'
else:
bucket = str(int(field) * 1000)
latency_tree.append((bucket, job_details['latency_ms'][field]))
# Column layout:
# General Bandwidth IOPS Latency CLAT Percentile
# --------- ---------- -------- -------- ---------------
# Size Min Min Min 1.00
# BW Max Max Max 5.00
# IOPS Mean Mean Mean 10.00
# Runtime StdDev StdDev StdDev 50.00
# UsrCPU Samples Samples 90.00
# SysCPU 99.50
# CtxSw 99.90
# MajFault 99.95
# MinFault 99.99
# Find the minimum entry without a zero
useful_latency_tree = list()
for element in latency_tree:
if element[1] != 0:
useful_latency_tree.append(element)
overall_label = [ 'Overall BW:',
max_rows = 9
if len(useful_latency_tree) > 9:
max_rows = len(useful_latency_tree)
elif len(useful_latency_tree) < 9:
while len(useful_latency_tree) < 9:
useful_latency_tree.append(('', ''))
# Format the static data
overall_label = [ 'Overall BW/s:',
'Overall IOPS:',
'Total I/O:',
'Runtime (s):',
@ -1992,6 +2000,9 @@ def format_info_benchmark_json(config, benchmark_information):
'Ctx Switches:',
'Major Faults:',
'Minor Faults:' ]
while len(overall_label) < max_rows:
overall_label.append('')
overall_data = [ format_bytes_tohuman(int(job_details[io_class]['bw_bytes'])),
format_ops_tohuman(int(job_details[io_class]['iops'])),
format_bytes_tohuman(int(job_details[io_class]['io_bytes'])),
@ -2001,6 +2012,9 @@ def format_info_benchmark_json(config, benchmark_information):
job_details['ctx'],
job_details['majf'],
job_details['minf'] ]
while len(overall_data) < max_rows:
overall_data.append('')
bandwidth_label = [ 'Min:',
'Max:',
'Mean:',
@ -2010,6 +2024,9 @@ def format_info_benchmark_json(config, benchmark_information):
'',
'',
'' ]
while len(bandwidth_label) < max_rows:
bandwidth_label.append('')
bandwidth_data = [ format_bytes_tohuman(int(job_details[io_class]['bw_min']) * 1024),
format_bytes_tohuman(int(job_details[io_class]['bw_max']) * 1024),
format_bytes_tohuman(int(job_details[io_class]['bw_mean']) * 1024),
@ -2019,6 +2036,9 @@ def format_info_benchmark_json(config, benchmark_information):
'',
'',
'' ]
while len(bandwidth_data) < max_rows:
bandwidth_data.append('')
iops_data = [ format_ops_tohuman(int(job_details[io_class]['iops_min'])),
format_ops_tohuman(int(job_details[io_class]['iops_max'])),
format_ops_tohuman(int(job_details[io_class]['iops_mean'])),
@ -2028,6 +2048,9 @@ def format_info_benchmark_json(config, benchmark_information):
'',
'',
'' ]
while len(iops_data) < max_rows:
iops_data.append('')
lat_data = [ job_details[io_class]['lat_ns']['min'],
job_details[io_class]['lat_ns']['max'],
job_details[io_class]['lat_ns']['mean'],
@ -2037,25 +2060,39 @@ def format_info_benchmark_json(config, benchmark_information):
'',
'',
'' ]
lat_percentile_label = [ '99.99%:',
'99.95%:',
'99.9%:',
'99.5%:',
'99%:',
'90%:',
'50%:',
'10%:',
'1%:' ]
lat_percentile_data = [ job_details[io_class]['clat_ns']['percentile']['99.990000'],
job_details[io_class]['clat_ns']['percentile']['99.950000'],
job_details[io_class]['clat_ns']['percentile']['99.900000'],
job_details[io_class]['clat_ns']['percentile']['99.500000'],
job_details[io_class]['clat_ns']['percentile']['99.000000'],
job_details[io_class]['clat_ns']['percentile']['90.000000'],
job_details[io_class]['clat_ns']['percentile']['50.000000'],
job_details[io_class]['clat_ns']['percentile']['10.000000'],
job_details[io_class]['clat_ns']['percentile']['1.000000'] ]
while len(lat_data) < max_rows:
lat_data.append('')
# Format the dynamic buckets
lat_bucket_label = list()
lat_bucket_data = list()
for element in useful_latency_tree:
lat_bucket_label.append(element[0])
lat_bucket_data.append(element[1])
# Column default widths
overall_label_length = 0
overall_column_length = 0
bandwidth_label_length = 0
bandwidth_column_length = 11
iops_column_length = 4
latency_column_length = 12
latency_bucket_label_length = 0
# Column layout:
# General Bandwidth IOPS Latency Percentiles
# --------- ---------- -------- -------- ---------------
# Size Min Min Min A
# BW Max Max Max B
# IOPS Mean Mean Mean ...
# Runtime StdDev StdDev StdDev Z
# UsrCPU Samples Samples
# SysCPU
# CtxSw
# MajFault
# MinFault
# Set column widths
for item in overall_label:
_item_length = len(str(item))
if _item_length > overall_label_length:
@ -2091,25 +2128,20 @@ def format_info_benchmark_json(config, benchmark_information):
if _item_length > latency_column_length:
latency_column_length = _item_length
for item in lat_percentile_label:
for item in lat_bucket_label:
_item_length = len(str(item))
if _item_length > latency_percentile_label_length:
latency_percentile_label_length = _item_length
for item in lat_percentile_data:
_item_length = len(str(item))
if _item_length > latency_percentile_column_length:
latency_percentile_column_length = _item_length
if _item_length > latency_bucket_label_length:
latency_bucket_label_length = _item_length
# Top row (Headers)
ainformation.append('{bold}\
{overall_label: <{overall_label_length}} \
{bandwidth_label: <{bandwidth_label_length}} \
{bandwidth: <{bandwidth_length}} \
{iops: <{iops_length}} \
{bandwidth: <{bandwidth_length}} \
{iops: <{iops_length}} \
{latency: <{latency_length}} \
{latency_percentile_label: <{latency_percentile_label_length}} \
{latency_percentile: <{latency_percentile_length}} \
{latency_bucket_label: <{latency_bucket_label_length}} \
{latency_bucket} \
{end_bold}'.format(
bold=ansiprint.bold(),
end_bold=ansiprint.end(),
@ -2117,29 +2149,28 @@ def format_info_benchmark_json(config, benchmark_information):
overall_label_length=overall_label_length,
bandwidth_label='',
bandwidth_label_length=bandwidth_label_length,
bandwidth='Bandwidth',
bandwidth='Bandwidth/s',
bandwidth_length=bandwidth_column_length,
iops='IOPS',
iops_length=iops_column_length,
latency='Latency (μs)',
latency_length=latency_column_length,
latency_percentile_label='CLAT Percentiles (μs)',
latency_percentile_label_length=latency_percentile_label_length,
latency_percentile='',
latency_percentile_length=latency_percentile_column_length,
latency_bucket_label='Latency Buckets (μs/%)',
latency_bucket_label_length=latency_bucket_label_length,
latency_bucket='',
))
for idx, _ in enumerate(overall_data):
for idx in range(0, max_rows):
# Top row (Headers)
ainformation.append('{bold}\
{overall_label: >{overall_label_length}} \
{overall: <{overall_length}} \
{bandwidth_label: >{bandwidth_label_length}} \
{bandwidth: <{bandwidth_length}} \
{iops: <{iops_length}} \
{bandwidth: <{bandwidth_length}} \
{iops: <{iops_length}} \
{latency: <{latency_length}} \
{latency_percentile_label: >{latency_percentile_label_length}} \
{latency_percentile: <{latency_percentile_length}} \
{latency_bucket_label: >{latency_bucket_label_length}} \
{latency_bucket} \
{end_bold}'.format(
bold='',
end_bold='',
@ -2155,10 +2186,9 @@ def format_info_benchmark_json(config, benchmark_information):
iops_length=iops_column_length,
latency=lat_data[idx],
latency_length=latency_column_length,
latency_percentile_label=lat_percentile_label[idx],
latency_percentile_label_length=latency_percentile_label_length,
latency_percentile=lat_percentile_data[idx],
latency_percentile_length=latency_percentile_column_length,
latency_bucket_label=lat_bucket_label[idx],
latency_bucket_label_length=latency_bucket_label_length,
latency_bucket=lat_bucket_data[idx],
))
return '\n'.join(ainformation)