diff options
author | Tibor Frank <tifrank@cisco.com> | 2019-08-16 09:37:31 +0200 |
---|---|---|
committer | Tibor Frank <tifrank@cisco.com> | 2019-08-19 11:35:12 +0000 |
commit | 1fac7da55c5b6d5a5651bc3ecac8aa945de49ef8 (patch) | |
tree | 9e1b5db16a513b452c7e61f49df26cceb0adce2c /resources | |
parent | 5ed11380df697a0f1565a1545337e1fd7bf0660e (diff) |
PAL: Select tests by name, functions
- throughput
- latency
- speedup
Change-Id: Ia8b33815f5bbb7b0fb50b23e7655068928733832
Signed-off-by: Tibor Frank <tifrank@cisco.com>
(cherry picked from commit 144ebec26ff0a32000283eddfac393e497e01164)
Diffstat (limited to 'resources')
-rw-r--r-- | resources/tools/presentation/generator_plots.py | 606 | ||||
-rw-r--r-- | resources/tools/presentation/input_data_parser.py | 90 | ||||
-rw-r--r-- | resources/tools/presentation/specification.yaml | 55 |
3 files changed, 751 insertions, 0 deletions
diff --git a/resources/tools/presentation/generator_plots.py b/resources/tools/presentation/generator_plots.py index 7c8a9f8628..f5bcb0abfa 100644 --- a/resources/tools/presentation/generator_plots.py +++ b/resources/tools/presentation/generator_plots.py @@ -61,10 +61,612 @@ def generate_plots(spec, data): logging.info("Done.") +def plot_performance_name_box(plot, input_data): + """Generate the plot(s) with algorithm: plot_performance_name_box + specified in the specification file. + + :param plot: Plot to generate. + :param input_data: Data to process. + :type plot: pandas.Series + :type input_data: InputData + """ + + # Transform the data + plot_title = plot.get("title", "") + logging.info(" Creating the data set for the {0} '{1}'.". + format(plot.get("type", ""), plot_title)) + data = input_data.filter_tests_by_name( + plot, params=["throughput", "parent", "tags", "type"]) + if data is None: + logging.error("No data.") + return + + # Prepare the data for the plot + y_vals = OrderedDict() + for job in data: + for build in job: + for test in build: + if y_vals.get(test["parent"], None) is None: + y_vals[test["parent"]] = list() + try: + if test["type"] in ("NDRPDR", ): + if "-pdr" in plot_title.lower(): + y_vals[test["parent"]].\ + append(test["throughput"]["PDR"]["LOWER"]) + elif "-ndr" in plot_title.lower(): + y_vals[test["parent"]]. \ + append(test["throughput"]["NDR"]["LOWER"]) + else: + continue + elif test["type"] in ("SOAK", ): + y_vals[test["parent"]].\ + append(test["throughput"]["LOWER"]) + else: + continue + except (KeyError, TypeError): + y_vals[test["parent"]].append(None) + + # Add None to the lists with missing data + max_len = 0 + nr_of_samples = list() + for val in y_vals.values(): + if len(val) > max_len: + max_len = len(val) + nr_of_samples.append(len(val)) + for key, val in y_vals.items(): + if len(val) < max_len: + val.extend([None for _ in range(max_len - len(val))]) + + # Add plot traces + traces = list() + df = pd.DataFrame(y_vals) + df.head() + y_max = list() + for i, col in enumerate(df.columns): + tst_name = re.sub(REGEX_NIC, "", + col.lower().replace('-ndrpdr', ''). + replace('2n1l-', '')) + name = "{nr}. ({samples:02d} run{plural}) {name}".\ + format(nr=(i + 1), + samples=nr_of_samples[i], + plural='s' if nr_of_samples[i] > 1 else '', + name=tst_name) + + logging.debug(name) + traces.append(plgo.Box(x=[str(i + 1) + '.'] * len(df[col]), + y=[y / 1000000 if y else None for y in df[col]], + name=name, + hoverinfo="x+y", + boxpoints="outliers", + whiskerwidth=0)) + try: + val_max = max(df[col]) + except ValueError as err: + logging.error(repr(err)) + continue + if val_max: + y_max.append(int(val_max / 1000000) + 2) + + try: + # Create plot + layout = deepcopy(plot["layout"]) + if layout.get("title", None): + layout["title"] = "<b>Throughput:</b> {0}". \ + format(layout["title"]) + if y_max: + layout["yaxis"]["range"] = [0, max(y_max)] + plpl = plgo.Figure(data=traces, layout=layout) + + # Export Plot + file_type = plot.get("output-file-type", ".html") + logging.info(" Writing file '{0}{1}'.". + format(plot["output-file"], file_type)) + ploff.plot(plpl, show_link=False, auto_open=False, + filename='{0}{1}'.format(plot["output-file"], file_type)) + except PlotlyError as err: + logging.error(" Finished with error: {}". + format(repr(err).replace("\n", " "))) + return + + +def plot_latency_error_bars_name(plot, input_data): + """Generate the plot(s) with algorithm: plot_latency_error_bars_name + specified in the specification file. + + :param plot: Plot to generate. + :param input_data: Data to process. + :type plot: pandas.Series + :type input_data: InputData + """ + + # Transform the data + plot_title = plot.get("title", "") + logging.info(" Creating the data set for the {0} '{1}'.". + format(plot.get("type", ""), plot_title)) + data = input_data.filter_tests_by_name( + plot, params=["latency", "parent", "tags", "type"]) + if data is None: + logging.error("No data.") + return + + # Prepare the data for the plot + y_tmp_vals = OrderedDict() + for job in data: + for build in job: + for test in build: + try: + logging.debug("test['latency']: {0}\n". + format(test["latency"])) + except ValueError as err: + logging.warning(repr(err)) + if y_tmp_vals.get(test["parent"], None) is None: + y_tmp_vals[test["parent"]] = [ + list(), # direction1, min + list(), # direction1, avg + list(), # direction1, max + list(), # direction2, min + list(), # direction2, avg + list() # direction2, max + ] + try: + if test["type"] in ("NDRPDR", ): + if "-pdr" in plot_title.lower(): + ttype = "PDR" + elif "-ndr" in plot_title.lower(): + ttype = "NDR" + else: + logging.warning("Invalid test type: {0}". + format(test["type"])) + continue + y_tmp_vals[test["parent"]][0].append( + test["latency"][ttype]["direction1"]["min"]) + y_tmp_vals[test["parent"]][1].append( + test["latency"][ttype]["direction1"]["avg"]) + y_tmp_vals[test["parent"]][2].append( + test["latency"][ttype]["direction1"]["max"]) + y_tmp_vals[test["parent"]][3].append( + test["latency"][ttype]["direction2"]["min"]) + y_tmp_vals[test["parent"]][4].append( + test["latency"][ttype]["direction2"]["avg"]) + y_tmp_vals[test["parent"]][5].append( + test["latency"][ttype]["direction2"]["max"]) + else: + logging.warning("Invalid test type: {0}". + format(test["type"])) + continue + except (KeyError, TypeError) as err: + logging.warning(repr(err)) + + x_vals = list() + y_vals = list() + y_mins = list() + y_maxs = list() + nr_of_samples = list() + for key, val in y_tmp_vals.items(): + name = re.sub(REGEX_NIC, "", key.replace('-ndrpdr', ''). + replace('2n1l-', '')) + x_vals.append(name) # dir 1 + y_vals.append(mean(val[1]) if val[1] else None) + y_mins.append(mean(val[0]) if val[0] else None) + y_maxs.append(mean(val[2]) if val[2] else None) + nr_of_samples.append(len(val[1]) if val[1] else 0) + x_vals.append(name) # dir 2 + y_vals.append(mean(val[4]) if val[4] else None) + y_mins.append(mean(val[3]) if val[3] else None) + y_maxs.append(mean(val[5]) if val[5] else None) + nr_of_samples.append(len(val[3]) if val[3] else 0) + + traces = list() + annotations = list() + + for idx in range(len(x_vals)): + if not bool(int(idx % 2)): + direction = "West-East" + else: + direction = "East-West" + hovertext = ("No. of Runs: {nr}<br>" + "Test: {test}<br>" + "Direction: {dir}<br>".format(test=x_vals[idx], + dir=direction, + nr=nr_of_samples[idx])) + if isinstance(y_maxs[idx], float): + hovertext += "Max: {max:.2f}uSec<br>".format(max=y_maxs[idx]) + if isinstance(y_vals[idx], float): + hovertext += "Mean: {avg:.2f}uSec<br>".format(avg=y_vals[idx]) + if isinstance(y_mins[idx], float): + hovertext += "Min: {min:.2f}uSec".format(min=y_mins[idx]) + + if isinstance(y_maxs[idx], float) and isinstance(y_vals[idx], float): + array = [y_maxs[idx] - y_vals[idx], ] + else: + array = [None, ] + if isinstance(y_mins[idx], float) and isinstance(y_vals[idx], float): + arrayminus = [y_vals[idx] - y_mins[idx], ] + else: + arrayminus = [None, ] + traces.append(plgo.Scatter( + x=[idx, ], + y=[y_vals[idx], ], + name=x_vals[idx], + legendgroup=x_vals[idx], + showlegend=bool(int(idx % 2)), + mode="markers", + error_y=dict( + type='data', + symmetric=False, + array=array, + arrayminus=arrayminus, + color=COLORS[int(idx / 2)] + ), + marker=dict( + size=10, + color=COLORS[int(idx / 2)], + ), + text=hovertext, + hoverinfo="text", + )) + annotations.append(dict( + x=idx, + y=0, + xref="x", + yref="y", + xanchor="center", + yanchor="top", + text="E-W" if bool(int(idx % 2)) else "W-E", + font=dict( + size=16, + ), + align="center", + showarrow=False + )) + + try: + # Create plot + file_type = plot.get("output-file-type", ".html") + logging.info(" Writing file '{0}{1}'.". + format(plot["output-file"], file_type)) + layout = deepcopy(plot["layout"]) + if layout.get("title", None): + layout["title"] = "<b>Latency:</b> {0}".\ + format(layout["title"]) + layout["annotations"] = annotations + plpl = plgo.Figure(data=traces, layout=layout) + + # Export Plot + ploff.plot(plpl, + show_link=False, auto_open=False, + filename='{0}{1}'.format(plot["output-file"], file_type)) + except PlotlyError as err: + logging.error(" Finished with error: {}". + format(str(err).replace("\n", " "))) + return + + +def plot_throughput_speedup_analysis_name(plot, input_data): + """Generate the plot(s) with algorithm: + plot_throughput_speedup_analysis_name + specified in the specification file. + + :param plot: Plot to generate. + :param input_data: Data to process. + :type plot: pandas.Series + :type input_data: InputData + """ + + # Transform the data + plot_title = plot.get("title", "") + logging.info(" Creating the data set for the {0} '{1}'.". + format(plot.get("type", ""), plot_title)) + data = input_data.filter_tests_by_name( + plot, params=["throughput", "parent", "tags", "type"]) + if data is None: + logging.error("No data.") + return + + y_vals = OrderedDict() + for job in data: + for build in job: + for test in build: + if y_vals.get(test["parent"], None) is None: + y_vals[test["parent"]] = {"1": list(), + "2": list(), + "4": list()} + try: + if test["type"] in ("NDRPDR",): + if "-pdr" in plot_title.lower(): + ttype = "PDR" + elif "-ndr" in plot_title.lower(): + ttype = "NDR" + else: + continue + if "1C" in test["tags"]: + y_vals[test["parent"]]["1"]. \ + append(test["throughput"][ttype]["LOWER"]) + elif "2C" in test["tags"]: + y_vals[test["parent"]]["2"]. \ + append(test["throughput"][ttype]["LOWER"]) + elif "4C" in test["tags"]: + y_vals[test["parent"]]["4"]. \ + append(test["throughput"][ttype]["LOWER"]) + except (KeyError, TypeError): + pass + + if not y_vals: + logging.warning("No data for the plot '{}'". + format(plot.get("title", ""))) + return + + y_1c_max = dict() + for test_name, test_vals in y_vals.items(): + for key, test_val in test_vals.items(): + if test_val: + avg_val = sum(test_val) / len(test_val) + y_vals[test_name][key] = (avg_val, len(test_val)) + ideal = avg_val / (int(key) * 1000000.0) + if test_name not in y_1c_max or ideal > y_1c_max[test_name]: + y_1c_max[test_name] = ideal + + vals = OrderedDict() + y_max = list() + nic_limit = 0 + lnk_limit = 0 + pci_limit = plot["limits"]["pci"]["pci-g3-x8"] + for test_name, test_vals in y_vals.items(): + try: + if test_vals["1"][1]: + name = re.sub(REGEX_NIC, "", test_name.replace('-ndrpdr', ''). + replace('2n1l-', '')) + vals[name] = OrderedDict() + y_val_1 = test_vals["1"][0] / 1000000.0 + y_val_2 = test_vals["2"][0] / 1000000.0 if test_vals["2"][0] \ + else None + y_val_4 = test_vals["4"][0] / 1000000.0 if test_vals["4"][0] \ + else None + + vals[name]["val"] = [y_val_1, y_val_2, y_val_4] + vals[name]["rel"] = [1.0, None, None] + vals[name]["ideal"] = [y_1c_max[test_name], + y_1c_max[test_name] * 2, + y_1c_max[test_name] * 4] + vals[name]["diff"] = [(y_val_1 - y_1c_max[test_name]) * 100 / + y_val_1, None, None] + vals[name]["count"] = [test_vals["1"][1], + test_vals["2"][1], + test_vals["4"][1]] + + try: + val_max = max(vals[name]["val"]) + except ValueError as err: + logging.error(repr(err)) + continue + if val_max: + y_max.append(val_max) + + if y_val_2: + vals[name]["rel"][1] = round(y_val_2 / y_val_1, 2) + vals[name]["diff"][1] = \ + (y_val_2 - vals[name]["ideal"][1]) * 100 / y_val_2 + if y_val_4: + vals[name]["rel"][2] = round(y_val_4 / y_val_1, 2) + vals[name]["diff"][2] = \ + (y_val_4 - vals[name]["ideal"][2]) * 100 / y_val_4 + except IndexError as err: + logging.warning("No data for '{0}'".format(test_name)) + logging.warning(repr(err)) + + # Limits: + if "x520" in test_name: + limit = plot["limits"]["nic"]["x520"] + elif "x710" in test_name: + limit = plot["limits"]["nic"]["x710"] + elif "xxv710" in test_name: + limit = plot["limits"]["nic"]["xxv710"] + elif "xl710" in test_name: + limit = plot["limits"]["nic"]["xl710"] + elif "x553" in test_name: + limit = plot["limits"]["nic"]["x553"] + else: + limit = 0 + if limit > nic_limit: + nic_limit = limit + + mul = 2 if "ge2p" in test_name else 1 + if "10ge" in test_name: + limit = plot["limits"]["link"]["10ge"] * mul + elif "25ge" in test_name: + limit = plot["limits"]["link"]["25ge"] * mul + elif "40ge" in test_name: + limit = plot["limits"]["link"]["40ge"] * mul + elif "100ge" in test_name: + limit = plot["limits"]["link"]["100ge"] * mul + else: + limit = 0 + if limit > lnk_limit: + lnk_limit = limit + + traces = list() + annotations = list() + x_vals = [1, 2, 4] + + # Limits: + try: + threshold = 1.1 * max(y_max) # 10% + except ValueError as err: + logging.error(err) + return + nic_limit /= 1000000.0 + traces.append(plgo.Scatter( + x=x_vals, + y=[nic_limit, ] * len(x_vals), + name="NIC: {0:.2f}Mpps".format(nic_limit), + showlegend=False, + mode="lines", + line=dict( + dash="dot", + color=COLORS[-1], + width=1), + hoverinfo="none" + )) + annotations.append(dict( + x=1, + y=nic_limit, + xref="x", + yref="y", + xanchor="left", + yanchor="bottom", + text="NIC: {0:.2f}Mpps".format(nic_limit), + font=dict( + size=14, + color=COLORS[-1], + ), + align="left", + showarrow=False + )) + y_max.append(nic_limit) + + lnk_limit /= 1000000.0 + if lnk_limit < threshold: + traces.append(plgo.Scatter( + x=x_vals, + y=[lnk_limit, ] * len(x_vals), + name="Link: {0:.2f}Mpps".format(lnk_limit), + showlegend=False, + mode="lines", + line=dict( + dash="dot", + color=COLORS[-2], + width=1), + hoverinfo="none" + )) + annotations.append(dict( + x=1, + y=lnk_limit, + xref="x", + yref="y", + xanchor="left", + yanchor="bottom", + text="Link: {0:.2f}Mpps".format(lnk_limit), + font=dict( + size=14, + color=COLORS[-2], + ), + align="left", + showarrow=False + )) + y_max.append(lnk_limit) + + pci_limit /= 1000000.0 + if (pci_limit < threshold and + (pci_limit < lnk_limit * 0.95 or lnk_limit > lnk_limit * 1.05)): + traces.append(plgo.Scatter( + x=x_vals, + y=[pci_limit, ] * len(x_vals), + name="PCIe: {0:.2f}Mpps".format(pci_limit), + showlegend=False, + mode="lines", + line=dict( + dash="dot", + color=COLORS[-3], + width=1), + hoverinfo="none" + )) + annotations.append(dict( + x=1, + y=pci_limit, + xref="x", + yref="y", + xanchor="left", + yanchor="bottom", + text="PCIe: {0:.2f}Mpps".format(pci_limit), + font=dict( + size=14, + color=COLORS[-3], + ), + align="left", + showarrow=False + )) + y_max.append(pci_limit) + + # Perfect and measured: + cidx = 0 + for name, val in vals.iteritems(): + hovertext = list() + try: + for idx in range(len(val["val"])): + htext = "" + if isinstance(val["val"][idx], float): + htext += "No. of Runs: {1}<br>" \ + "Mean: {0:.2f}Mpps<br>".format(val["val"][idx], + val["count"][idx]) + if isinstance(val["diff"][idx], float): + htext += "Diff: {0:.0f}%<br>".format( + round(val["diff"][idx])) + if isinstance(val["rel"][idx], float): + htext += "Speedup: {0:.2f}".format(val["rel"][idx]) + hovertext.append(htext) + traces.append(plgo.Scatter(x=x_vals, + y=val["val"], + name=name, + legendgroup=name, + mode="lines+markers", + line=dict( + color=COLORS[cidx], + width=2), + marker=dict( + symbol="circle", + size=10 + ), + text=hovertext, + hoverinfo="text+name" + )) + traces.append(plgo.Scatter(x=x_vals, + y=val["ideal"], + name="{0} perfect".format(name), + legendgroup=name, + showlegend=False, + mode="lines", + line=dict( + color=COLORS[cidx], + width=2, + dash="dash"), + text=["Perfect: {0:.2f}Mpps".format(y) + for y in val["ideal"]], + hoverinfo="text" + )) + cidx += 1 + except (IndexError, ValueError, KeyError) as err: + logging.warning("No data for '{0}'".format(name)) + logging.warning(repr(err)) + + try: + # Create plot + file_type = plot.get("output-file-type", ".html") + logging.info(" Writing file '{0}{1}'.". + format(plot["output-file"], file_type)) + layout = deepcopy(plot["layout"]) + if layout.get("title", None): + layout["title"] = "<b>Speedup Multi-core:</b> {0}". \ + format(layout["title"]) + layout["yaxis"]["range"] = [0, int(max(y_max) * 1.1)] + layout["annotations"].extend(annotations) + plpl = plgo.Figure(data=traces, layout=layout) + + # Export Plot + ploff.plot(plpl, + show_link=False, auto_open=False, + filename='{0}{1}'.format(plot["output-file"], file_type)) + except PlotlyError as err: + logging.error(" Finished with error: {}". + format(repr(err).replace("\n", " "))) + return + + def plot_performance_box(plot, input_data): """Generate the plot(s) with algorithm: plot_performance_box specified in the specification file. + TODO: Remove when not needed. + :param plot: Plot to generate. :param input_data: Data to process. :type plot: pandas.Series @@ -439,6 +1041,8 @@ def plot_latency_error_bars(plot, input_data): """Generate the plot(s) with algorithm: plot_latency_error_bars specified in the specification file. + TODO: Remove when not needed. + :param plot: Plot to generate. :param input_data: Data to process. :type plot: pandas.Series @@ -650,6 +1254,8 @@ def plot_throughput_speedup_analysis(plot, input_data): plot_throughput_speedup_analysis specified in the specification file. + TODO: Remove when not needed. + :param plot: Plot to generate. :param input_data: Data to process. :type plot: pandas.Series diff --git a/resources/tools/presentation/input_data_parser.py b/resources/tools/presentation/input_data_parser.py index bbbf0a9ae0..f47f1bc6df 100644 --- a/resources/tools/presentation/input_data_parser.py +++ b/resources/tools/presentation/input_data_parser.py @@ -1497,6 +1497,96 @@ class InputData(object): "tags are enclosed by apostrophes.".format(cond)) return None + def filter_tests_by_name(self, element, params=None, data_set="tests", + continue_on_error=False): + """Filter required data from the given jobs and builds. + + The output data structure is: + + - job 1 + - build 1 + - test (or suite) 1 ID: + - param 1 + - param 2 + ... + - param n + ... + - test (or suite) n ID: + ... + ... + - build n + ... + - job n + + :param element: Element which will use the filtered data. + :param params: Parameters which will be included in the output. If None, + all parameters are included. + :param data_set: The set of data to be filtered: tests, suites, + metadata. + :param continue_on_error: Continue if there is error while reading the + data. The Item will be empty then + :type element: pandas.Series + :type params: list + :type data_set: str + :type continue_on_error: bool + :returns: Filtered data. + :rtype pandas.Series + """ + + include = element.get("include", None) + if not include: + logging.warning("No tests to include, skipping the element.") + return None + + if params is None: + params = element.get("parameters", None) + if params: + params.append("type") + + data = pd.Series() + try: + for job, builds in element["data"].items(): + data[job] = pd.Series() + for build in builds: + data[job][str(build)] = pd.Series() + for test in include: + try: + reg_ex = re.compile(str(test).lower()) + for test_ID in self.data[job][str(build)]\ + [data_set].keys(): + if re.match(reg_ex, str(test_ID).lower()): + test_data = self.data[job][str(build)]\ + [data_set][test_ID] + data[job][str(build)][test_ID] = pd.Series() + if params is None: + for param, val in test_data.items(): + data[job][str(build)][test_ID]\ + [param] = val + else: + for param in params: + try: + data[job][str(build)][test_ID]\ + [param] = test_data[param] + except KeyError: + data[job][str(build)][test_ID]\ + [param] = "No Data" + except KeyError as err: + logging.error("{err!r}".format(err=err)) + if continue_on_error: + continue + else: + return None + return data + + except (KeyError, IndexError, ValueError) as err: + logging.error("Missing mandatory parameter in the element " + "specification: {err!r}".format(err=err)) + return None + except AttributeError as err: + logging.error("{err!r}".format(err=err)) + return None + + @staticmethod def merge_data(data): """Merge data from more jobs and builds to a simple data structure. diff --git a/resources/tools/presentation/specification.yaml b/resources/tools/presentation/specification.yaml index 21ce8542bf..02141ac162 100644 --- a/resources/tools/presentation/specification.yaml +++ b/resources/tools/presentation/specification.yaml @@ -6300,6 +6300,61 @@ ################################################################################ +# Example plots + +# Packet Throughput - VPP L2 3n-skx-x710 base and scale +- type: "plot" + title: "Throughput: l2sw-3n-skx-x710-64b-2t1c-base_and_scale-ndr" + algorithm: "plot_performance_name_box" + output-file: "{DIR[STATIC,VPP]}/l2sw-3n-skx-x710-64b-2t1c-base_and_scale-ndr" + data: "plot-vpp-throughput-lat-tsa-3n-skx" + include: + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Patch-Ndrpdr.64B-2t1c-eth-l2patch-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Xcbase-Ndrpdr.64B-2t1c-eth-l2xcbase-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdbasemaclrn-Ndrpdr.64B-2t1c-eth-l2bdbasemaclrn-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdscale10Kmaclrn-Ndrpdr.64B-2t1c-eth-l2bdscale10kmaclrn-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdscale100Kmaclrn-Ndrpdr.64B-2t1c-eth-l2bdscale100kmaclrn-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdscale1Mmaclrn-Ndrpdr.64B-2t1c-eth-l2bdscale1mmaclrn-ndrpdr" + layout: + title: "l2sw-3n-skx-x710-64b-2t1c-base_and_scale-ndr" + layout: "plot-throughput" + +# Packet Latency - VPP L2 3n-skx-x710 base and scale +- type: "plot" + title: "Latency: l2sw-3n-skx-x710-64b-2t1c-base_and_scale-ndr-lat" + algorithm: "plot_latency_error_bars_name" + output-file: "{DIR[STATIC,VPP]}/l2sw-3n-skx-x710-64b-2t1c-base_and_scale-ndr-lat" + data: "plot-vpp-throughput-lat-tsa-3n-skx" + include: + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Patch-Ndrpdr.64B-2t1c-eth-l2patch-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Xcbase-Ndrpdr.64B-2t1c-eth-l2xcbase-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdbasemaclrn-Ndrpdr.64B-2t1c-eth-l2bdbasemaclrn-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdscale10Kmaclrn-Ndrpdr.64B-2t1c-eth-l2bdscale10kmaclrn-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdscale100Kmaclrn-Ndrpdr.64B-2t1c-eth-l2bdscale100kmaclrn-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdscale1Mmaclrn-Ndrpdr.64B-2t1c-eth-l2bdscale1mmaclrn-ndrpdr" + layout: + title: "l2sw-3n-skx-x710-64b-2t1c-base_and_scale-ndr" + layout: "plot-latency" + +# Speedup - VPP L2 3n-skx-x710 base and scale +- type: "plot" + title: "Speedup: l2sw-3n-skx-x710-64b-base_and_scale-ndr-tsa" + algorithm: "plot_throughput_speedup_analysis_name" + output-file: "{DIR[STATIC,VPP]}/l2sw-3n-skx-x710-64b-base_and_scale-ndr-tsa" + data: "plot-vpp-throughput-lat-tsa-3n-skx" + include: + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Patch-Ndrpdr.64B-.t.c-eth-l2patch-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Xcbase-Ndrpdr.64B-.t.c-eth-l2xcbase-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdbasemaclrn-Ndrpdr.64B-.t.c-eth-l2bdbasemaclrn-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdscale10Kmaclrn-Ndrpdr.64B-.t.c-eth-l2bdscale10kmaclrn-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdscale100Kmaclrn-Ndrpdr.64B-.t.c-eth-l2bdscale100kmaclrn-ndrpdr" + - "Tests.Vpp.Perf.L2.10Ge2P1X710-Eth-L2Bdscale1Mmaclrn-Ndrpdr.64B-.t.c-eth-l2bdscale1mmaclrn-ndrpdr" + layout: + title: "l2sw-3n-skx-x710-64b-base_and_scale-ndr" + layout: "plot-throughput-speedup-analysis" + +################################################################################ + # Packet Throughput - VPP L2 3n-hsw-x520 base and scale - type: "plot" title: "VPP Throughput: l2sw-3n-hsw-x520-64b-1t1c-base_and_scale-ndr" |