Merge pull request GoogleCloudPlatform#3083 from SMU-ATT-Center-for-V…

…irtualization:netperf_aggregate

PiperOrigin-RevId: 393386857

perfkitbenchmarker/data/netperf.patch

@@ -307,7 +307,8 @@ diff -u -r netperf-2.6.0/doc/examples/runemomniaggdemo.sh netperf-2.6.0-patch/do
  # net.core.[r|w]mem_max are sufficiently large
 -DO_BIDIR=1;
 +DO_BIDIR=0;
- DO_RRAGG=1;
+-DO_RRAGG=1;
++DO_RRAGG=0;
 -DO_RR=1;
 -DO_ANCILLARY=1;
 +DO_RR=0;
@@ -339,3 +340,94 @@ diff -u -r netperf-2.7.0/config.guess  netperf-2.7.0-patch/config.guess
      alpha:Linux:*:*)
         case `sed -n '/^cpu model/s/^.*: \(.*\)/\1/p' < /proc/cpuinfo` in
           EV5)   UNAME_MACHINE=alphaev5 ;;
+
+diff -u -r netperf-2.7.0/doc/examples/post_proc.py netperf-2.7.0-patch/doc/examples/post_proc.py
+--- netperf-2.7.0/doc/examples/post_proc.py 2015-07-20 12:39:35.000000000 -0500
++++ netperf-2.7.0-patch/doc/examples/post_proc.py   2021-07-16 10:22:34.642546090 -0500
+@@ -125,7 +125,7 @@
+             ksink[key] += float(result[0])
+         else:
+             if result[0]:
+-                print "Key %d not in ksink" % key
++                print("Key %d not in ksink" % key)
+
+ def process_result(basename, raw_results, end_time, ksink):
+     first_result = True
+@@ -209,7 +209,7 @@
+             try:
+                 update_rrd(basename,interim_result,interim_end)
+             except Exception as e:
+-                print "Update to %s with %s at %s failed with %s" % (basename,interim_result,interim_end,e)
++                print("Update to %s with %s at %s failed with %s" % (basename,interim_result,interim_end,e))
+             have_result = False
+             had_results = True
+
+@@ -221,7 +221,7 @@
+         return 0, 0
+
+ def process_result_files(prefix,start_time,end_time,ksink):
+-    print "Prefix is %s" % prefix
++    print("Prefix is %s" % prefix)
+     min_timestamp = 9999999999.9
+     results_list = glob.glob(prefix+"*.out")
+
+@@ -254,13 +254,13 @@
+     # find that key in the kitchen sink, we will use the value to
+     # update the overall rrd.
+     prevkey = -1
+-    for key in xrange(int(start_time),int(end_time)+1):
++    for key in range(int(start_time),int(end_time)+1):
+         if key in ksink:
+             try:
+                 update_rrd(overall,ksink[key],key)
+                 prevkey = key
+             except Exception as e:
+-                print "Update to %s failed for %d, previous %d %s" % (overall, key, prevkey, e)
++                print("Update to %s failed for %d, previous %d %s" % (overall, key, prevkey, e))
+
+ def overall_min_max_avg(prefix,start_time,end_time,intervals):
+
+@@ -312,7 +312,7 @@
+         imax = float(result[2].strip('"'))
+         results_list.append((iavg, imin, imax, start, end))
+
+-        for time in xrange(start,end+1):
++        for time in range(start,end+1):
+             rrdtool.update(prefix + "_intervals.rrd",
+                            '%d:%f:%f:%f' % (time, iavg, imin, imax))
+         if iavg > max_average:
+@@ -444,13 +444,13 @@
+     # the time being I will preallocate the entire dict in one fell
+     # swoop until I can modify add_to_ksink() accordingly
+     length = int(end_time + 1) - int(start_time)
+-    ksink=dict(zip(xrange(int(start_time),
++    ksink=dict(zip(range(int(start_time),
+                           int(end_time)+1),
+                    [0.0] * length))
+
+     min_timestamp = process_result_files(prefix,start_time,end_time,ksink)
+     if min_timestamp == 9999999999.9:
+-        print "There were no valid results for this prefix!"
++        print("There were no valid results for this prefix!")
+         exit()
+
+ #    print "Min timestamp for %s is %s start time is %s end_time is %s" % (prefix,min_timestamp,start_time,end_time)
+@@ -471,12 +471,12 @@
+                          annotation=args.annotation,override=args.title)
+
+     units, multiplier, direction = units_et_al_by_prefix(prefix)
+-    print "Average of peak interval is %.3f %s from %d to %d" % (results_list[0][0] * float(multiplier), units, peak_start, peak_end)
+-    print "Minimum of peak interval is %.3f %s from %d to %d" % (peak_minimum * float(multiplier), units, peak_start, peak_end)
+-    print "Maximum of peak interval is %.3f %s from %d to %d" % (peak_maximum * float(multiplier), units, peak_start, peak_end)
++    print("Average of peak interval is %.3f %s from %d to %d" % (results_list[0][0] * float(multiplier), units, peak_start, peak_end))
++    print("Minimum of peak interval is %.3f %s from %d to %d" % (peak_minimum * float(multiplier), units, peak_start, peak_end))
++    print("Maximum of peak interval is %.3f %s from %d to %d" % (peak_maximum * float(multiplier), units, peak_start, peak_end))
+
+     if args.intervals:
+         for id, interval in enumerate(results_list[1:]):
+-            print "Average of interval %d is %.3f %s from %d to %d" % (id, interval[0] * float(multiplier), units, interval[3], interval[4])
+-            print "Minimum of interval %d is %.3f %s from %d to %d" % (id, interval[1] * float(multiplier), units, interval[3], interval[4])
+-            print "Maximum of interval %d is %.3f %s from %d to %d" % (id, interval[2] * float(multiplier), units, interval[3], interval[4])
++            print("Average of interval %d is %.3f %s from %d to %d" % (id, interval[0] * float(multiplier), units, interval[3], interval[4]))
++            print("Minimum of interval %d is %.3f %s from %d to %d" % (id, interval[1] * float(multiplier), units, interval[3], interval[4]))
++            print("Maximum of interval %d is %.3f %s from %d to %d" % (id, interval[2] * float(multiplier), units, interval[3], interval[4]))

perfkitbenchmarker/linux_benchmarks/netperf_pps_benchmark.py → perfkitbenchmarker/linux_benchmarks/netperf_aggregate_benchmark.py

@@ -1,4 +1,4 @@
-# Copyright 2019 PerfKitBenchmarker Authors. All rights reserved.
+# Copyright 2021 PerfKitBenchmarker Authors. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -18,25 +18,38 @@
 https://hewlettpackard.github.io/netperf/doc/netperf.html
 manpage: http://manpages.ubuntu.com/manpages/maverick/man1/netperf.1.html
 
-Runs UDP_RR in script between one source machine and two target machines
-to test packets per second
+Runs multiple tests in script between one source machine and two target machines
+to test packets per second and inbound and outbound throughput
 
 """
 
+import collections
 import logging
+import os
 import re
 from absl import flags
 from perfkitbenchmarker import configs
 from perfkitbenchmarker import sample
 from perfkitbenchmarker import vm_util
 from perfkitbenchmarker.linux_packages import netperf
 
+ALL_BENCHMARKS = ['STREAM', 'MAERTS', 'BIDIR', 'RRAGG']
+flags.DEFINE_list('netperf_aggregate_benchmarks', ALL_BENCHMARKS,
+                  'The netperf aggregate benchmark(s) to run. '
+                  'STREAM measures outbound throughput. '
+                  'MAERTS measures inbound throughput. '
+                  'RRAGG measures packets per second.'
+                  'BIDIR measure bidirectional bulk throughput')
+flags.register_validator(
+    'netperf_aggregate_benchmarks',
+    lambda benchmarks: benchmarks and set(benchmarks).issubset(ALL_BENCHMARKS))
+
 FLAGS = flags.FLAGS
 
-BENCHMARK_NAME = 'netperf_pps'
+BENCHMARK_NAME = 'netperf_aggregate'
 BENCHMARK_CONFIG = """
-netperf_pps:
-  description: test packets per second performance using netperf
+netperf_aggregate:
+  description: simultaneous netperf to multiple endpoints
   vm_groups:
     servers:
       vm_spec: *default_single_core
@@ -61,10 +74,19 @@ def GetConfig(user_config):
 def PrepareNetperfAggregate(vm):
   """Installs netperf on a single vm."""
 
+  vm.Install('python3')
+  vm.Install('pip3')
+  vm.RemoteCommand('sudo pip3 install --upgrade pip')
   vm.Install('texinfo')
   vm.Install('python_rrdtool')
   vm.Install('netperf')
 
+  # Enable test types in the script runemomniaggdemo.sh
+  for benchmark in FLAGS.netperf_aggregate_benchmarks:
+    vm.RemoteCommand(
+        f'sed -i "s/DO_{benchmark}=0;/DO_{benchmark}=1;/g" /opt/pkb/netperf-netperf-2.7.0/doc/examples/runemomniaggdemo.sh'
+    )
+
   port_end = PORT_START
 
   if vm_util.ShouldRunOnExternalIpAddress():
@@ -75,9 +97,6 @@ def PrepareNetperfAggregate(vm):
       netserver_path=netperf.NETSERVER_PATH)
   vm.RemoteCommand(netserver_cmd)
 
-  remote_path = netperf.NETPERF_EXAMPLE_DIR + REMOTE_SCRIPT
-  vm.RemoteCommand('chmod +x %s' % (remote_path))
-
 
 def Prepare(benchmark_spec):
   """Install netperf on the target vm.
@@ -88,14 +107,20 @@ def Prepare(benchmark_spec):
   """
 
   vms = benchmark_spec.vms
+  client_vm = benchmark_spec.vm_groups['client'][0]
+
   vm_util.RunThreaded(PrepareNetperfAggregate, vms)
+  client_vm.RemoteCommand(
+      f'sudo chmod 777 {os.path.join(netperf.NETPERF_EXAMPLE_DIR, REMOTE_SCRIPT)}'
+  )
 
 
-def ParseNetperfAggregateOutput(stdout):
+def ParseNetperfAggregateOutput(stdout, test_type):
   """Parses the stdout of a single netperf process.
 
   Args:
     stdout: the stdout of the netperf process
+    test_type: the type of test
 
   Returns:
     A tuple containing (throughput_sample, latency_samples, latency_histogram)
@@ -110,7 +135,7 @@ def ParseNetperfAggregateOutput(stdout):
     match = re.search('peak interval', line)
     if match:
       line_split = line.split()
-      metric = line_split[0] + ' ' + line_split[6]
+      metric = f'{test_type} {line_split[0]} {line_split[6]}'
       value = float(line_split[5])
       unit = line_split[6]
       aggregate_samples.append(sample.Sample(
@@ -149,7 +174,6 @@ def RunNetperfAggregate(vm, server_ips):
   vm.RemoteCommand(f"echo 'NUM_REMOTE_HOSTS={len(server_ips)}' >> "
                    f"{netperf.NETPERF_EXAMPLE_DIR}/remote_hosts")
 
-  # allow script to be executed and run script
   vm.RemoteCommand(
       f'cd {netperf.NETPERF_EXAMPLE_DIR} && '
       'export PATH=$PATH:. && '
@@ -160,24 +184,25 @@ def RunNetperfAggregate(vm, server_ips):
       login_shell=False,
       timeout=1200)
 
-  # print out netperf_tps.log to log
-  stdout_1, stderr_1 = vm.RemoteCommand(
-      f'cat {netperf.NETPERF_EXAMPLE_DIR}/netperf_tps.log',
-      ignore_failure=True,
-      should_log=True,
-      login_shell=False,
-      timeout=1200)
-
-  logging.info(stdout_1)
-  logging.info(stderr_1)
-
-  # do post processing step
-  proc_stdout, _ = vm.RemoteCommand(
-      f'cd {netperf.NETPERF_EXAMPLE_DIR} && ./post_proc.py '
-      '--intervals netperf_tps.log',
-      ignore_failure=True)
-
-  samples = ParseNetperfAggregateOutput(proc_stdout)
+  interval_naming = collections.namedtuple(
+      'IntervalNaming', 'output_file parse_name')
+  benchmark_interval_mapping = {
+      'STREAM': interval_naming('netperf_outbound', 'Outbound'),
+      'MAERTS': interval_naming('netperf_inbound', 'Inbound'),
+      'RRAGG': interval_naming('netperf_tps', 'Request/Response Aggregate'),
+      'BIDIR': interval_naming('netperf_bidirectional', 'Bidirectional')
+  }
+
+  samples = []
+  for benchmark in FLAGS.netperf_aggregate_benchmarks:
+    output_file = benchmark_interval_mapping[benchmark].output_file
+    parse_name = benchmark_interval_mapping[benchmark].parse_name
+    proc_stdout, _ = vm.RemoteCommand(
+        f'cd {netperf.NETPERF_EXAMPLE_DIR} && python3 post_proc.py '
+        f'--intervals {output_file}.log',
+        ignore_failure=False)
+    vm.RemoteCommand(f'cd {netperf.NETPERF_EXAMPLE_DIR} && rm {output_file}*')
+    samples.extend(ParseNetperfAggregateOutput(proc_stdout, f'{parse_name}'))
 
   return samples
 

perfkitbenchmarker/linux_packages/python_rrdtool.py

@@ -1,4 +1,4 @@
-# Copyright 2014 PerfKitBenchmarker Authors. All rights reserved.
+# Copyright 2021 PerfKitBenchmarker Authors. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -17,10 +17,14 @@
 
 
 def YumInstall(vm):
-  """Installs build tools on the VM."""
-  vm.InstallPackages('python-rrdtool')
+  """Installs python rrdtool on the VM."""
+  vm.InstallPackages('rrdtool')
+  vm.RemoteCommand('sudo dnf --enablerepo=powertools install -y rrdtool-devel')
+  vm.RemoteCommand('sudo pip3 install rrdtool')
 
 
 def AptInstall(vm):
-  """Installs build tools on the VM."""
-  vm.InstallPackages('python-rrdtool')
+  """Installs python rrdtool on the VM."""
+  vm.InstallPackages('librrd-dev')
+  vm.InstallPackages('libpython3-dev')
+  vm.RemoteCommand('sudo pip install rrdtool')

perfkitbenchmarker/linux_packages/texinfo.py

@@ -16,9 +16,9 @@
 """Module containing build tools installation and cleanup functions."""
 
 
-def YumInstall(vm):
+def YumInstall(_):
   """Installs build tools on the VM."""
-  vm.InstallPackages('texinfo')
+  pass
 
 
 def AptInstall(vm):