Create an user adjustable table for Perf metrics (#1000)

- create packages for BenchmarkMathCalculation.js. It can be shared by both client and server code
- remove TestResultSummaryService/perf/BenchmarkMathCalculation.js
- create user adjustable MetricTable

resolves: automation/issues/235

Signed-off-by: Lan Xia <[email protected]>

Author: llxia

TestResultSummaryService/package-lock.json

@@ -38,6 +38,7 @@
         "mathjs": "^7.6.0",
         "mongodb": "^3.6.10",
         "request": "^2.88.2",
+        "utils": "file:../packages/utils",
         "winston": "^2.4.5"
     },
     "devDependencies": {

TestResultSummaryService/package.json

@@ -1,4 +1,4 @@
-const BenchmarkMath = require('./BenchmarkMathCalculation');
+const BenchmarkMath = require('utils/BenchmarkMathCalculation');
 const math = require('mathjs');
 const ObjectID = require('mongodb').ObjectID;
 

TestResultSummaryService/perf/BenchmarkMathCalculation.js

@@ -4,7 +4,7 @@ const DefaultParser = require('../parsers/Default');
 const DataManagerAggregate = require('../perf/DataManagerAggregate');
 const ArgParser = require('../ArgParser');
 const math = require('mathjs');
-const BenchmarkMath = require('../perf/BenchmarkMathCalculation');
+const BenchmarkMath = require('utils/BenchmarkMathCalculation');
 
 module.exports = async (req, res) => {
     const { user, password, server } =
@@ -54,14 +54,16 @@ module.exports = async (req, res) => {
                     }
                     const results = JSON.parse(body.body);
                     // extract worker name from timeline
-                    const agentName = results.records.find(
-                        (job) =>
-                            job.name.startsWith(`Run ${buildInfo.templateParameters.suite}`)
+                    const agentName = results.records.find((job) =>
+                        job.name.startsWith(
+                            `Run ${buildInfo.templateParameters.suite}`
+                        )
                     ).workerName;
                     // find the log with the name of the benchmark
-                    const logURL = results.records.find(
-                        (job) =>
-                            job.name.startsWith(`Run ${buildInfo.templateParameters.suite}`)
+                    const logURL = results.records.find((job) =>
+                        job.name.startsWith(
+                            `Run ${buildInfo.templateParameters.suite}`
+                        )
                     ).log.url;
                     // fetch the log
                     request.get(

TestResultSummaryService/perf/DataManagerAggregate.js

@@ -7,6 +7,5 @@ module.exports = async (req, res) => {
     if (query.parentId) query.parentId = new ObjectID(query.parentId);
     const db = new TestResultsDB();
     const result = await db.getData({ _id: new ObjectID(query._id) }).toArray();
-    console.log(result);
     res.send(result);
 };

TestResultSummaryService/routes/getAzDoRun.js

@@ -0,0 +1,2 @@
+# dependencies
+node_modules

TestResultSummaryService/routes/getData.js

@@ -0,0 +1,84 @@
+const math = require("mathjs");
+
+class BenchmarkMathCalculation {
+  // Taken from Perffarm/perfsite/benchmarks.php
+  /*
+   * This function calculates the confidence interval of a set
+   * of results. We are limited to testing for 5%
+   * convergence.
+   */
+  static confidence_interval(scores) {
+    let scores_sum = 0;
+    scores.forEach((x) => {
+      if (x !== null) {
+        scores_sum += x;
+      }
+    });
+
+    // First get the std deviation and other
+    // useful values as floats.
+    if (scores_sum !== 0) {
+      let stddev = math.std(scores);
+      let count = scores.length;
+      let mean = scores_sum / count;
+
+      // Do the convergence calculations.
+      let ci = stddev * this.t_dist05(count - 1);
+      ci /= mean;
+      ci /= Math.sqrt(count);
+
+      return ci;
+    } else {
+      return 0;
+    }
+  }
+
+  // Taken from Perffarm/perfsite/benchmarks.php
+  /*
+   * This is a lookup function for the t-distribution.
+   * It is based on the statistical tests code Dave
+   * Siegwart wrote for RAJ.
+   * It only does probability of 0.05.
+   */
+  static t_dist05(N) {
+    // Constants for t-dist calculations.
+    let Student_t_05 = [
+      -1.0, 12.706, 4.303, 3.182, 2.776, 2.571, 2.447, 2.365, 2.306, 2.262,
+      2.228,
+
+      2.201, 2.179, 2.16, 2.145, 2.131, 2.12, 2.11, 2.101, 2.093, 2.086,
+
+      2.08, 2.074, 2.069, 2.064, 2.06, 2.056, 2.052, 2.048, 2.045, 2.042,
+    ];
+    let Student_t_05_40 = 2.021;
+    let Student_t_05_60 = 2.0;
+    let Student_t_05_120 = 1.98;
+    let Student_t_05_2000 = 1.96;
+
+    let P = 0.0;
+
+    if (N <= 30) {
+      P = Student_t_05[N];
+    } else if (N <= 40) {
+      P = this.interp(Student_t_05[30], Student_t_05_40, 30, 40, N);
+    } else if (N <= 60) {
+      P = this.interp(Student_t_05_40, Student_t_05_60, 40, 60, N);
+    } else if (N <= 120) {
+      P = this.interp(Student_t_05_60, Student_t_05_120, 60, 120, N);
+    } else if (N <= 2000) {
+      P = this.interp(Student_t_05_120, Student_t_05_2000, 120, 2000, N);
+    } else {
+      P = Student_t_05_2000;
+    }
+    return P;
+  }
+
+  // Taken from Perffarm/perfsite/benchmarks.php
+  // Support function for t_dist05
+  static interp(a, b, aN, bN, N) {
+    let mu = (N - aN) / (bN - aN);
+    let v = mu * (b - a) + a;
+    return v;
+  }
+}
+module.exports = BenchmarkMathCalculation;