public function mloutliersAction()
{
$jsonData = $jsonWarns = $jsonOuts = array();
$message = $instance = $jsonHeader = $jsonTable = $model_html = $config = $model_info = '';
$possible_models = $possible_models_id = $other_models = array();
$jsonResolutions = $jsonResolutionsHeader = '[]';
$max_x = $max_y = 0;
$must_wait = 'NO';
try {
$dbml = new \PDO($this->container->get('config')['db_conn_chain'], $this->container->get('config')['mysql_user'], $this->container->get('config')['mysql_pwd']);
$dbml->setAttribute(\PDO::ATTR_ERRMODE, \PDO::ERRMODE_EXCEPTION);
$dbml->setAttribute(\PDO::ATTR_EMULATE_PREPARES, false);
$db = $this->container->getDBUtils();
// FIXME - This must be counted BEFORE building filters, as filters inject rubbish in GET when there are no parameters...
$instructions = count($_GET) <= 1;
if (array_key_exists('dump', $_GET)) {
$dump = $_GET["dump"];
unset($_GET["dump"]);
}
if (array_key_exists('register', $_GET)) {
$register = $_GET["register"];
unset($_GET["register"]);
}
$this->buildFilters(array('current_model' => array('type' => 'selectOne', 'default' => null, 'label' => 'Model to use: ', 'generateChoices' => function () {
return array();
}, 'parseFunction' => function () {
$choice = isset($_GET['current_model']) ? $_GET['current_model'] : array("");
return array('whereClause' => '', 'currentChoice' => $choice);
}, 'filterGroup' => 'MLearning'), 'sigma' => array('type' => 'inputNumber', 'default' => 1, 'label' => 'Sigmas: ', 'parseFunction' => function () {
$choice = isset($_GET['sigma']) ? $_GET['sigma'] : 1;
return array('whereClause' => '', 'currentChoice' => $choice);
}, 'max' => 3, 'min' => 1, 'filterGroup' => 'MLearning'), 'minexetime' => array('default' => 0), 'valid' => array('default' => 0), 'filter' => array('default' => 0), 'prepares' => array('default' => 1)));
$this->buildFilterGroups(array('MLearning' => array('label' => 'Machine Learning', 'tabOpenDefault' => true, 'filters' => array('current_model', 'sigma'))));
$params = array();
$param_names = array('bench', 'net', 'disk', 'maps', 'iosf', 'replication', 'iofilebuf', 'comp', 'blk_size', 'id_cluster', 'datanodes', 'vm_OS', 'vm_cores', 'vm_RAM', 'provider', 'vm_size', 'type', 'bench_type', 'hadoop_version');
// Order is important
$params = $this->filters->getFiltersSelectedChoices($param_names);
foreach ($param_names as $p) {
if (!is_null($params[$p]) && is_array($params[$p])) {
sort($params[$p]);
}
}
$params_additional = array();
$param_names_additional = array('datefrom', 'dateto', 'minexetime', 'maxexetime', 'valid', 'filter');
// Order is important
$params_additional = $this->filters->getFiltersSelectedChoices($param_names_additional);
$param_variables = $this->filters->getFiltersSelectedChoices(array('current_model', 'sigma'));
$param_current_model = $param_variables['current_model'];
$sigma_param = $param_variables['sigma'];
$where_configs = $this->filters->getWhereclause();
$where_configs = str_replace("AND .", "AND ", $where_configs);
// compose instance
$instance = MLUtils::generateSimpleInstance($this->filters, $param_names, $params, true);
$model_info = MLUtils::generateModelInfo($this->filters, $param_names, $params, true);
$slice_info = MLUtils::generateDatasliceInfo($this->filters, $param_names_additional, $params_additional);
// model for filling
MLUtils::findMatchingModels($model_info, $possible_models, $possible_models_id, $dbml);
$current_model = '';
if (!is_null($possible_models_id) && in_array($param_current_model, $possible_models_id)) {
$current_model = $param_current_model;
}
// Other models for filling
$where_models = '';
if (!empty($possible_models_id)) {
$where_models = " WHERE id_learner NOT IN ('" . implode("','", $possible_models_id) . "')";
}
$result = $dbml->query("SELECT id_learner FROM aloja_ml.learners" . $where_models);
foreach ($result as $row) {
$other_models[] = $row['id_learner'];
}
if ($instructions) {
$result = $dbml->query("SELECT id_learner, model, algorithm FROM aloja_ml.learners");
foreach ($result as $row) {
$model_html = $model_html . "<li>" . $row['id_learner'] . " => " . $row['algorithm'] . " : " . $row['model'] . "</li>";
}
MLUtils::getIndexOutExps($jsonResolutions, $jsonResolutionsHeader, $dbml);
$this->filters->setCurrentChoices('current_model', array_merge($possible_models_id, array('---Other models---'), $other_models));
return $this->render('mltemplate/mloutliers.html.twig', array('outexps' => $jsonResolutions, 'header_outexps' => $jsonResolutionsHeader, 'jsonData' => '[]', 'jsonWarns' => '[]', 'jsonOuts' => '[]', 'jsonHeader' => '[]', 'jsonTable' => '[]', 'max_p' => 0, 'models' => $model_html, 'instructions' => 'YES'));
}
if (!empty($possible_models_id)) {
$result = $dbml->query("SELECT id_learner, model, algorithm, CASE WHEN `id_learner` IN ('" . implode("','", $possible_models_id) . "') THEN 'COMPATIBLE' ELSE 'NOT MATCHED' END AS compatible FROM aloja_ml.learners");
foreach ($result as $row) {
$model_html = $model_html . "<li>" . $row['id_learner'] . " => " . $row['algorithm'] . " : " . $row['compatible'] . " : " . $row['model'] . "</li>";
}
if ($current_model == "") {
$query = "SELECT AVG(ABS(exe_time - pred_time)) AS MAE, AVG(ABS(exe_time - pred_time)/exe_time) AS RAE, p.id_learner FROM aloja_ml.predictions p, aloja_ml.learners l WHERE l.id_learner = p.id_learner AND p.id_learner IN ('" . implode("','", $possible_models_id) . "') AND predict_code > 0 ORDER BY MAE LIMIT 1";
$result = $dbml->query($query);
$row = $result->fetch();
$current_model = $row['id_learner'];
}
$config = $instance . '-' . $current_model . '-' . $sigma_param . ' ' . $slice_info . '-outliers';
$is_cached_mysql = $dbml->query("SELECT count(*) as total FROM aloja_ml.resolutions WHERE id_resolution = '" . md5($config) . "'");
$tmp_result = $is_cached_mysql->fetch();
$is_cached = $tmp_result['total'] > 0;
$cache_ds = getcwd() . '/cache/query/' . md5($config) . '-cache.csv';
$in_process = file_exists(getcwd() . '/cache/query/' . md5($config) . '.lock');
$finished_process = file_exists(getcwd() . '/cache/query/' . md5($config) . '-resolutions.csv');
if (!$is_cached && !$in_process && !$finished_process) {
// get headers for csv
$header_names = array('id_exec' => 'ID', 'bench' => 'Benchmark', 'exe_time' => 'Exe.Time', 'net' => 'Net', 'disk' => 'Disk', 'maps' => 'Maps', 'iosf' => 'IO.SFac', 'replication' => 'Rep', 'iofilebuf' => 'IO.FBuf', 'comp' => 'Comp', 'blk_size' => 'Blk.size', 'e.id_cluster' => 'Cluster', 'datanodes' => 'Datanodes', 'vm_OS' => 'VM.OS', 'vm_cores' => 'VM.Cores', 'vm_RAM' => 'VM.RAM', 'provider' => 'Provider', 'vm_size' => 'VM.Size', 'type' => 'Type', 'bench_type' => 'Bench.Type', 'hadoop_version' => 'Hadoop.Version');
$headers = array_keys($header_names);
$names = array_values($header_names);
// dump the result to csv
$query = "SELECT " . implode(",", $headers) . " FROM aloja2.execs e LEFT JOIN aloja2.clusters c ON e.id_cluster = c.id_cluster WHERE hadoop_version IS NOT NULL" . $where_configs . ";";
$rows = $db->get_rows($query);
if (empty($rows)) {
throw new \Exception('No data matches with your critteria.');
}
$fp = fopen($cache_ds, 'w');
fputcsv($fp, $names, ',', '"');
foreach ($rows as $row) {
$row['id_cluster'] = "Cl" . $row['id_cluster'];
// Cluster is numerically codified...
$row['comp'] = "Cmp" . $row['comp'];
// Compression is numerically codified...
fputcsv($fp, array_values($row), ',', '"');
}
// Retrieve file model from DB
$query = "SELECT file FROM aloja_ml.model_storage WHERE id_hash='" . $current_model . "' AND type='learner';";
$result = $dbml->query($query);
$row = $result->fetch();
$content = $row['file'];
$filemodel = getcwd() . '/cache/query/' . $current_model . '-object.rds';
$fp = fopen($filemodel, 'w');
fwrite($fp, $content);
fclose($fp);
// launch query
exec('cd ' . getcwd() . '/cache/query ; touch ' . md5($config) . '.lock');
exec(getcwd() . '/resources/queue -c "cd ' . getcwd() . '/cache/query ; ' . getcwd() . '/resources/aloja_cli.r -m aloja_outlier_dataset -d ' . $cache_ds . ' -l ' . $current_model . ' -p sigma=' . $sigma_param . ':hdistance=3:saveall=' . md5($config) . ' > /dev/null 2>&1 ; rm -f ' . md5($config) . '.lock" > /dev/null 2>&1 &');
}
$finished_process = file_exists(getcwd() . '/cache/query/' . md5($config) . '-resolutions.csv');
if ($finished_process && !$is_cached) {
if (($handle = fopen(getcwd() . '/cache/query/' . md5($config) . '-resolutions.csv', 'r')) !== FALSE) {
$header = fgetcsv($handle, 1000, ",");
$token = 0;
$query = "REPLACE INTO aloja_ml.resolutions (id_resolution,id_learner,id_exec,instance,model,dataslice,sigma,outlier_code,predicted,observed) VALUES ";
while (($data = fgetcsv($handle, 1000, ",")) !== FALSE) {
$resolution = $data[0];
$pred_value = (int) $data[1] >= 100 ? (int) $data[1] : 100;
$exec_value = (int) $data[2];
$selected_instance_pre = preg_replace('/\\s+/', '', $data[3]);
$selected_instance_pre = str_replace(':', ',', $selected_instance_pre);
$specific_id = $data[4];
if ($token > 0) {
$query = $query . ",";
}
$token = 1;
$query = $query . "('" . md5($config) . "','" . $current_model . "','" . $specific_id . "','" . $selected_instance_pre . "','" . $model_info . "','" . $slice_info . "','" . $sigma_param . "','" . $resolution . "','" . $pred_value . "','" . $exec_value . "') ";
}
if ($dbml->query($query) === FALSE) {
throw new \Exception('Error when saving tree into DB');
}
}
// Store file model to DB
$filemodel = getcwd() . '/cache/query/' . md5($config) . '-object.rds';
$fp = fopen($filemodel, 'r');
$content = fread($fp, filesize($filemodel));
$content = addslashes($content);
fclose($fp);
$query = "INSERT INTO aloja_ml.model_storage (id_hash,type,file) VALUES ('" . md5($config) . "','resolution','" . $content . "');";
if ($dbml->query($query) === FALSE) {
throw new \Exception('Error when saving file resolution into DB');
}
// Remove temporary files
$output = shell_exec('rm -f ' . getcwd() . '/cache/query/' . md5($config) . '-*.csv');
$is_cached = true;
}
if (!$is_cached) {
$jsonData = $jsonOuts = $jsonWarns = $jsonHeader = $jsonTable = '[]';
$must_wait = 'YES';
if (isset($dump)) {
echo "1";
exit(0);
}
} else {
$must_wait = 'NO';
$query = "SELECT predicted, observed, outlier_code, id_exec, instance FROM aloja_ml.resolutions WHERE id_resolution = '" . md5($config) . "' LIMIT 5000";
// FIXME - CLUMSY PATCH FOR BYPASS THE BUG FROM HIGHCHARTS... REMEMBER TO ERASE THIS LINE WHEN THE BUG IS SOLVED
$result = $dbml->query($query);
foreach ($result as $row) {
$entry = array('x' => (int) $row['predicted'], 'y' => (int) $row['observed'], 'name' => $row['instance'], 'id' => (int) $row['id_exec']);
if ($row['outlier_code'] == 0) {
$jsonData[] = $entry;
}
if ($row['outlier_code'] == 1) {
$jsonWarns[] = $entry;
}
if ($row['outlier_code'] == 2) {
$jsonOuts[] = $entry;
}
$jsonTable .= ($jsonTable == '' ? '' : ',') . '["' . ($row['outlier_code'] == 0 ? 'Legitimate' : ($row['outlier_code'] == 1 ? 'Warning' : 'Outlier')) . '","' . $row['predicted'] . '","' . $row['observed'] . '","' . str_replace(",", "\",\"", $row['instance']) . '","' . $row['id_exec'] . '"]';
}
$query_var = "SELECT MAX(predicted) as max_x, MAX(observed) as max_y FROM aloja_ml.resolutions WHERE id_resolution = '" . md5($config) . "' LIMIT 5000";
$result = $dbml->query($query_var);
$row = $result->fetch();
$max_x = $row['max_x'];
$max_y = $row['max_y'];
$header = array('Prediction', 'Observed', 'Benchmark', 'Net', 'Disk', 'Maps', 'IO.SFS', 'Rep', 'IO.FBuf', 'Comp', 'Blk.Size', 'Cluster', 'Datanodes', 'VM.OS', 'VM.Cores', 'VM.RAM', 'Provider', 'VM.Size', 'Type', 'Bench.Type', 'Version', 'ID');
$jsonHeader = '[{title:""}';
foreach ($header as $title) {
$jsonHeader = $jsonHeader . ',{title:"' . $title . '"}';
}
$jsonHeader = $jsonHeader . ']';
$jsonData = json_encode($jsonData);
$jsonWarns = json_encode($jsonWarns);
$jsonOuts = json_encode($jsonOuts);
$jsonTable = '[' . $jsonTable . ']';
// Dump case
if (isset($dump)) {
echo str_replace(array("[", "]", "{title:\"", "\"}"), array('', '', ''), $jsonHeader) . "\n";
echo str_replace(array('],[', '[[', ']]'), array("\n", '', ''), $jsonOuts);
echo str_replace(array('],[', '[[', ']]'), array("\n", '', ''), $jsonWarns);
echo str_replace(array('],[', '[[', ']]'), array("\n", '', ''), $jsonData);
exit(0);
}
// Register case
if (isset($register)) {
// Update the predictions table
$query_var = "UPDATE aloja_ml.predictions as p, aloja_ml.resolutions as r\n\t\t\t\t\t\t\t\tSET p.outlier = r.outlier_code\n\t\t\t\t\t\t\t\tWHERE r.id_exec = p.id_exec\n\t\t\t\t\t\t\t\t\tAND r.id_resolution = '" . md5($config) . "'\n\t\t\t\t\t\t\t\t\tAND p.id_learner = '" . $current_model . "'";
if ($dbml->query($query_var) === FALSE) {
throw new \Exception('Error when updating aloja_ml.predictions in DB');
}
}
}
} else {
throw new \Exception('There are no prediction models trained for such parameters. Train at least one model in "ML Prediction" section.');
}
$dbml = null;
} catch (\Exception $e) {
$this->container->getTwig()->addGlobal('message', $e->getMessage());
$jsonData = $jsonOuts = $jsonWarns = $jsonHeader = $jsonTable = '[]';
$must_wait = "NO";
$dbml = null;
}
$return_params = array('jsonData' => $jsonData, 'jsonWarns' => $jsonWarns, 'jsonOuts' => $jsonOuts, 'jsonHeader' => $jsonHeader, 'jsonTable' => $jsonTable, 'max_p' => min(array($max_x, $max_y)), 'outexps' => $jsonResolutions, 'header_outexps' => $jsonResolutionsHeader, 'must_wait' => $must_wait, 'models' => $model_html, 'models_id' => $possible_models_id, 'other_models_id' => $other_models, 'current_model' => $current_model, 'resolution_id' => md5($config), 'model_info' => $model_info, 'slice_info' => $slice_info, 'sigma' => $sigma_param, 'message' => $message, 'instance' => $instance);
$this->filters->setCurrentChoices('current_model', array_merge($possible_models_id, array('---Other models---'), $other_models));
return $this->render('mltemplate/mloutliers.html.twig', $return_params);
}
public function mlparamEvaluationAction()
{
$rows = $categories = $series = '';
$must_wait = 'NO';
try {
$dbml = new \PDO($this->container->get('config')['db_conn_chain_ml'], $this->container->get('config')['mysql_user'], $this->container->get('config')['mysql_pwd']);
$dbml->setAttribute(\PDO::ATTR_ERRMODE, \PDO::ERRMODE_EXCEPTION);
$dbml->setAttribute(\PDO::ATTR_EMULATE_PREPARES, false);
$db = $this->container->getDBUtils();
$where_configs = '';
$preset = null;
if (count($_GET) <= 1 || count($_GET) == 2 && array_key_exists('parameval', $_GET) || count($_GET) == 2 && array_key_exists('current_model', $_GET)) {
$preset = Utils::setDefaultPreset($db, 'mlparameval');
}
$selPreset = isset($_GET['presets']) ? $_GET['presets'] : "none";
$params = array();
$param_names = array('benchs', 'nets', 'disks', 'mapss', 'iosfs', 'replications', 'iofilebufs', 'comps', 'blk_sizes', 'id_clusters', 'datanodess', 'bench_types', 'vm_sizes', 'vm_coress', 'vm_RAMs', 'types');
// Order is important
foreach ($param_names as $p) {
$params[$p] = Utils::read_params($p, $where_configs, FALSE);
sort($params[$p]);
}
$money = Utils::read_params('money', $where_configs);
$paramEval = isset($_GET['parameval']) && $_GET['parameval'] != '' ? $_GET['parameval'] : 'maps';
$minExecs = isset($_GET['minexecs']) ? $_GET['minexecs'] : -1;
$minExecsFilter = "";
// FIXME PATCH FOR PARAM LIBRARIES WITHOUT LEGACY
$where_configs = str_replace("AND .", "AND ", $where_configs);
$where_configs = str_replace("`id_cluster`", "e.`id_cluster`", $where_configs);
if ($minExecs > 0) {
$minExecsFilter = "HAVING COUNT(*) > {$minExecs}";
}
$filter_execs = DBUtils::getFilterExecs();
$options = Utils::getFilterOptions($db);
$paramOptions = array();
foreach ($options[$paramEval] as $option) {
if ($paramEval == 'id_cluster') {
$paramOptions[] = $option['name'];
} else {
if ($paramEval == 'comp') {
$paramOptions[] = Utils::getCompressionName($option[$paramEval]);
} else {
if ($paramEval == 'net') {
$paramOptions[] = Utils::getNetworkName($option[$paramEval]);
} else {
if ($paramEval == 'disk') {
$paramOptions[] = Utils::getDisksName($option[$paramEval]);
} else {
$paramOptions[] = $option[$paramEval];
}
}
}
}
}
$benchOptions = $db->get_rows("SELECT DISTINCT bench FROM execs e LEFT JOIN clusters c ON e.id_cluster = c.id_cluster WHERE 1 {$filter_execs} {$where_configs} GROUP BY {$paramEval}, bench order by {$paramEval}");
// get the result rows
$query = "SELECT count(*) as count, {$paramEval}, e.id_exec, exec as conf, bench, " . "exe_time, avg(exe_time) avg_exe_time, min(exe_time) min_exe_time " . "from execs e LEFT JOIN clusters c ON e.id_cluster = c.id_cluster WHERE 1 {$filter_execs} {$where_configs}" . "GROUP BY {$paramEval}, bench {$minExecsFilter} order by bench,{$paramEval}";
$rows = $db->get_rows($query);
if (!$rows) {
throw new \Exception("No results for query!");
}
$arrayBenchs = array();
foreach ($paramOptions as $param) {
foreach ($benchOptions as $bench) {
$arrayBenchs[$bench['bench']][$param] = null;
$arrayBenchs[$bench['bench']][$param]['y'] = 0;
$arrayBenchs[$bench['bench']][$param]['count'] = 0;
}
}
$series = array();
$bench = '';
foreach ($rows as $row) {
if ($paramEval == 'comp') {
$row[$paramEval] = Utils::getCompressionName($row['comp']);
} else {
if ($paramEval == 'id_cluster') {
$row[$paramEval] = Utils::getClusterName($row[$paramEval], $db);
} else {
if ($paramEval == 'net') {
$row[$paramEval] = Utils::getNetworkName($row['net']);
} else {
if ($paramEval == 'disk') {
$row[$paramEval] = Utils::getDisksName($row['disk']);
} else {
if ($paramEval == 'iofilebuf') {
$row[$paramEval] /= 1024;
}
}
}
}
}
$arrayBenchs[$row['bench']][$row[$paramEval]]['y'] = round((int) $row['avg_exe_time'], 2);
$arrayBenchs[$row['bench']][$row[$paramEval]]['count'] = (int) $row['count'];
}
// ----------------------------------------------------
// Add predictions to the series
// ----------------------------------------------------
$jsonData = $jsonHeader = "[]";
$instance = "";
$arrayBenchs_pred = array();
// FIXME PATCH FOR PARAM LIBRARIES WITHOUT LEGACY
$where_configs = str_replace("AND .", "AND ", $where_configs);
// compose instance
$instance = MLUtils::generateSimpleInstance($param_names, $params, true, $db);
$model_info = MLUtils::generateModelInfo($param_names, $params, true, $db);
$instances = MLUtils::generateInstances($param_names, $params, true, $db);
// model for filling
$possible_models = $possible_models_id = array();
MLUtils::findMatchingModels($model_info, $possible_models, $possible_models_id, $dbml);
$current_model = "";
if (array_key_exists('current_model', $_GET) && in_array($_GET['current_model'], $possible_models_id)) {
$current_model = $_GET['current_model'];
}
if (!empty($possible_models_id)) {
if ($current_model == "") {
$query = "SELECT AVG(ABS(exe_time - pred_time)) AS MAE, AVG(ABS(exe_time - pred_time)/exe_time) AS RAE, p.id_learner FROM predictions p, learners l WHERE l.id_learner = p.id_learner AND p.id_learner IN ('" . implode("','", $possible_models_id) . "') AND predict_code > 0 ORDER BY MAE LIMIT 1";
$result = $dbml->query($query);
$row = $result->fetch();
$current_model = $row['id_learner'];
}
$config = $instance . '-' . $current_model . "-parameval";
$query_cache = "SELECT count(*) as total FROM trees WHERE id_learner = '" . $current_model . "' AND model = '" . $model_info . "'";
$is_cached_mysql = $dbml->query($query_cache);
$tmp_result = $is_cached_mysql->fetch();
$is_cached = $tmp_result['total'] > 0;
$ret_data = null;
if (!$is_cached) {
// Call to MLFindAttributes, to fetch data
$_GET['pass'] = 2;
$_GET['unseen'] = 1;
$_GET['current_model'] = $current_model;
$mlfa1 = new MLFindAttributesController();
$mlfa1->container = $this->container;
$ret_data = $mlfa1->mlfindattributesAction();
if ($ret_data == 1) {
$must_wait = "YES";
$jsonData = $jsonHeader = '[]';
} else {
$is_cached_mysql = $dbml->query($query_cache);
$tmp_result = $is_cached_mysql->fetch();
$is_cached = $tmp_result['total'] > 0;
}
}
if ($is_cached) {
$must_wait = 'NO';
$query = "SELECT count(*) as count, {$paramEval}, bench, exe_time, avg(pred_time) avg_pred_time, min(pred_time) min_pred_time " . "FROM predictions e WHERE e.id_learner = '" . $current_model . "' {$filter_execs} {$where_configs}" . "GROUP BY {$paramEval}, bench {$minExecsFilter} order by bench, {$paramEval}";
$result = $dbml->query($query);
// Initialize array
foreach ($paramOptions as $param) {
foreach ($benchOptions as $bench) {
$arrayBenchs_pred[$bench['bench'] . '_pred'][$param] = null;
$arrayBenchs_pred[$bench['bench'] . '_pred'][$param]['y'] = 0;
$arrayBenchs_pred[$bench['bench'] . '_pred'][$param]['count'] = 0;
}
}
foreach ($result as $row) {
$bench_n = $row['bench'] . '_pred';
$class = $row[$paramEval];
if ($paramEval == 'comp') {
$value = Utils::getCompressionName($class);
} else {
if ($paramEval == 'id_cluster') {
$value = Utils::getClusterName($class, $db);
} else {
if ($paramEval == 'net') {
$value = Utils::getNetworkName($class);
} else {
if ($paramEval == 'disk') {
$value = Utils::getDisksName($class);
} else {
if ($paramEval == 'iofilebuf') {
$value = $class / 1024;
} else {
$value = $class;
}
}
}
}
}
if (!in_array($value, $paramOptions)) {
$paramOptions[] = $value;
foreach ($benchOptions as $bench) {
$arrayBenchs_pred[$bench['bench'] . '_pred'][$value] = null;
$arrayBenchs_pred[$bench['bench'] . '_pred'][$value]['y'] = 0;
$arrayBenchs_pred[$bench['bench'] . '_pred'][$value]['count'] = 0;
$arrayBenchs[$bench['bench']][$value] = null;
$arrayBenchs[$bench['bench']][$value]['y'] = 0;
$arrayBenchs[$bench['bench']][$value]['count'] = 0;
}
}
$arrayBenchs_pred[$bench_n][$value]['y'] = (int) $row['avg_pred_time'];
$arrayBenchs_pred[$bench_n][$value]['count'] = (int) $row['count'];
}
}
}
// ----------------------------------------------------
// END - Add predictions to the series
// ----------------------------------------------------
asort($paramOptions);
foreach ($arrayBenchs as $key => $arrayBench) {
$caregories = '';
$data_a = null;
$data_p = null;
foreach ($paramOptions as $param) {
if ($arrayBenchs[$key][$param]['count'] > 0 && empty($arrayBenchs_pred) || !empty($arrayBenchs_pred) && ($arrayBenchs_pred[$key . '_pred'][$param]['count'] > 0 || $arrayBenchs[$key][$param]['count'] > 0)) {
$data_a[] = $arrayBenchs[$key][$param];
if (!empty($arrayBenchs_pred)) {
$data_p[] = $arrayBenchs_pred[$key . '_pred'][$param];
}
$categories = $categories . "'{$param} " . Utils::getParamevalUnit($paramEval) . "',";
// FIXME - Redundant n times performed... don't care now
}
}
$series[] = array('name' => $key, 'data' => $data_a);
if (!empty($arrayBenchs_pred)) {
$series[] = array('name' => $key . '_pred', 'data' => $data_p);
}
}
$series = json_encode($series);
if (!empty($arrayBenchs_pred)) {
$colors = "['#7cb5ec','#9cd5fc','#434348','#636368','#90ed7d','#b0fd9d','#f7a35c','#f7c37c','#8085e9','#a0a5f9','#f15c80','#f17ca0','#e4d354','#f4f374','#8085e8','#a0a5f8','#8d4653','#ad6673','#91e8e1','#b1f8f1']";
} else {
$colors = "['#7cb5ec','#434348','#90ed7d','#f7a35c','#8085e9','#f15c80','#e4d354','#8085e8','#8d4653','#91e8e1']";
}
} catch (\Exception $e) {
$this->container->getTwig()->addGlobal('message', $e->getMessage() . "\n");
$series = $jsonHeader = $colors = '[]';
$instance = $current_model = '';
$possible_models = $possible_models_id = array();
$must_wait = 'NO';
}
echo $this->container->getTwig()->render('mltemplate/mlconfigperf.html.twig', array('selected' => 'mlparameval', 'title' => 'Improvement of Hadoop Execution by SW and HW Configurations', 'categories' => $categories, 'series' => $series, 'benchs' => $params['benchs'], 'nets' => $params['nets'], 'disks' => $params['disks'], 'blk_sizes' => $params['blk_sizes'], 'comps' => $params['comps'], 'id_clusters' => $params['id_clusters'], 'mapss' => $params['mapss'], 'replications' => $params['replications'], 'iosfs' => $params['iosfs'], 'iofilebufs' => $params['iofilebufs'], 'datanodess' => $params['datanodess'], 'bench_types' => $params['bench_types'], 'vm_sizes' => $params['vm_sizes'], 'vm_coress' => $params['vm_coress'], 'vm_RAMs' => $params['vm_RAMs'], 'types' => $params['types'], 'money' => $money, 'paramEval' => $paramEval, 'instance' => $instance, 'models' => '<li>' . implode('</li><li>', $possible_models) . '</li>', 'models_id' => $possible_models_id, 'current_model' => $current_model, 'gammacolors' => $colors, 'must_wait' => $must_wait, 'preset' => $preset, 'selPreset' => $selPreset, 'options' => Utils::getFilterOptions($db)));
}
public function mlfindattributesAction()
{
$instance = $instances = $message = $tree_descriptor = $model_html = $config = '';
$possible_models = $possible_models_id = $other_models = array();
$must_wait = 'NO';
try {
$dbml = new \PDO($this->container->get('config')['db_conn_chain'], $this->container->get('config')['mysql_user'], $this->container->get('config')['mysql_pwd']);
$dbml->setAttribute(\PDO::ATTR_ERRMODE, \PDO::ERRMODE_EXCEPTION);
$dbml->setAttribute(\PDO::ATTR_EMULATE_PREPARES, false);
$db = $this->container->getDBUtils();
if (array_key_exists('dump', $_GET)) {
$dump = $_GET["dump"];
unset($_GET["dump"]);
}
if (array_key_exists('pass', $_GET)) {
$pass = $_GET["pass"];
unset($_GET["pass"]);
}
$this->buildFilters(array('current_model' => array('type' => 'selectOne', 'default' => null, 'label' => 'Model to use: ', 'generateChoices' => function () {
return array();
}, 'parseFunction' => function () {
$choice = isset($_GET['current_model']) ? $_GET['current_model'] : array("");
return array('whereClause' => '', 'currentChoice' => $choice);
}, 'filterGroup' => 'MLearning'), 'unseen' => array('type' => 'checkbox', 'default' => 1, 'label' => 'Predict with unseen atributes ⚠', 'parseFunction' => function () {
$choice = isset($_GET['unseen']) && !isset($_GET['unseen']) ? 0 : 1;
return array('whereClause' => '', 'currentChoice' => $choice);
}, 'filterGroup' => 'MLearning'), 'minexetime' => array('default' => 0), 'valid' => array('default' => 0), 'filter' => array('default' => 0), 'prepares' => array('default' => 1)));
$this->buildFilterGroups(array('MLearning' => array('label' => 'Machine Learning', 'tabOpenDefault' => true, 'filters' => array('current_model', 'unseen'))));
$where_configs = $this->filters->getWhereClause();
$param_names = array('bench', 'net', 'disk', 'maps', 'iosf', 'replication', 'iofilebuf', 'comp', 'blk_size', 'id_cluster', 'datanodes', 'vm_OS', 'vm_cores', 'vm_RAM', 'provider', 'vm_size', 'type', 'bench_type', 'hadoop_version');
// Order is important
$params = $this->filters->getFiltersSelectedChoices($param_names);
foreach ($param_names as $p) {
if (!is_null($params[$p]) && is_array($params[$p])) {
sort($params[$p]);
}
}
$learnParams = $this->filters->getFiltersSelectedChoices(array('current_model', 'unseen'));
$param_current_model = $learnParams['current_model'];
$unseen = $learnParams['unseen'] ? true : false;
// FIXME PATCH FOR PARAM LIBRARIES WITHOUT LEGACY
$where_configs = str_replace("AND .", "AND ", $where_configs);
$jsonData = $jsonHeader = "[]";
$mae = $rae = 0;
// compose instance
$model_info = MLUtils::generateModelInfo($this->filters, $param_names, $params, $unseen);
$instance = MLUtils::generateSimpleInstance($this->filters, $param_names, $params, $unseen);
$instances = MLUtils::generateInstances($this->filters, $param_names, $params, $unseen, $db);
// Model for filling
MLUtils::findMatchingModels($model_info, $possible_models, $possible_models_id, $dbml);
$current_model = '';
if (!is_null($possible_models_id) && in_array($param_current_model, $possible_models_id)) {
$current_model = $param_current_model;
}
// Other models for filling
$where_models = '';
if (!empty($possible_models_id)) {
$where_models = " WHERE id_learner NOT IN ('" . implode("','", $possible_models_id) . "')";
}
$result = $dbml->query("SELECT id_learner FROM aloja_ml.learners" . $where_models);
foreach ($result as $row) {
$other_models[] = $row['id_learner'];
}
if (!empty($possible_models_id) || $current_model != "") {
$result = $dbml->query("SELECT id_learner, model, algorithm, CASE WHEN `id_learner` IN ('" . implode("','", $possible_models_id) . "') THEN 'COMPATIBLE' ELSE 'NOT MATCHED' END AS compatible FROM aloja_ml.learners");
foreach ($result as $row) {
$model_html = $model_html . "<li>" . $row['id_learner'] . " => " . $row['algorithm'] . " : " . $row['compatible'] . " : " . $row['model'] . "</li>";
}
if ($current_model == "") {
$query = "SELECT AVG(ABS(exe_time - pred_time)) AS MAE, AVG(ABS(exe_time - pred_time)/exe_time) AS RAE, p.id_learner FROM aloja_ml.predictions p, aloja_ml.learners l WHERE l.id_learner = p.id_learner AND p.id_learner IN ('" . implode("','", $possible_models_id) . "') AND predict_code > 0 ORDER BY MAE LIMIT 1";
$result = $dbml->query($query);
$row = $result->fetch();
$current_model = $row['id_learner'];
}
$config = $instance . '-' . $current_model . '-' . ($unseen ? 'U' : 'R');
$is_cached_mysql = $dbml->query("SELECT count(*) as total FROM aloja_ml.trees WHERE id_findattrs = '" . md5($config) . "'");
$tmp_result = $is_cached_mysql->fetch();
$is_cached = $tmp_result['total'] > 0;
$tmp_file = md5($config) . '.tmp';
$in_process = file_exists(getcwd() . '/cache/query/' . md5($config) . '.lock');
$finished_process = $in_process && (int) shell_exec('ls ' . getcwd() . '/cache/query/' . md5($config) . '-*.lock | wc -w ') == count($instances);
if (!$in_process && !$finished_process && !$is_cached) {
// Retrieve file model from DB
$query = "SELECT file FROM aloja_ml.model_storage WHERE id_hash='" . $current_model . "' AND type='learner';";
$result = $dbml->query($query);
$row = $result->fetch();
$content = $row['file'];
$filemodel = getcwd() . '/cache/query/' . $current_model . '-object.rds';
$fp = fopen($filemodel, 'w');
fwrite($fp, $content);
fclose($fp);
// Run the predictor
exec('cd ' . getcwd() . '/cache/query ; touch ' . md5($config) . '.lock ; rm -f ' . $tmp_file);
$count = 1;
foreach ($instances as $inst) {
exec(getcwd() . '/resources/queue -d -c "cd ' . getcwd() . '/cache/query ; ../../resources/aloja_cli.r -m aloja_predict_instance -l ' . $current_model . ' -p inst_predict=\'' . $inst . '\' -v | grep -v \'Prediction\' >>' . $tmp_file . ' 2>/dev/null; touch ' . md5($config) . '-' . $count++ . '.lock" >/dev/null 2>&1 &');
}
}
$finished_process = (int) shell_exec('ls ' . getcwd() . '/cache/query/' . md5($config) . '-*.lock | wc -w ') == count($instances);
if ($finished_process && !$is_cached) {
// Read results and dump to DB
$i = 0;
$token = 0;
$token_i = 0;
$query = "INSERT IGNORE INTO aloja_ml.predictions (id_exec,exe_time,bench,net,disk,maps,iosf,replication,iofilebuf,comp,blk_size,id_cluster,datanodes,vm_OS,vm_cores,vm_RAM,provider,vm_size,type,bench_type,hadoop_version,pred_time,id_learner,instance,predict_code) VALUES ";
if (($handle = fopen(getcwd() . '/cache/query/' . $tmp_file, "r")) !== FALSE) {
while (($line = fgets($handle, 1000)) !== FALSE && $i < 1000) {
if ($line == '') {
break;
}
// Fetch Real Value
$inst_aux = preg_split("/\\s+/", $line);
$query_var = "SELECT AVG(exe_time) as AVG, id_exec, outlier FROM aloja_ml.predictions WHERE instance = '" . $inst_aux[1] . "' AND predict_code > 0";
$result = $dbml->query($query_var);
$row = $result->fetch();
$realexecval = is_null($row['AVG']) || $row['outlier'] == 2 ? 0 : $row['AVG'];
$realid_exec = is_null($row['id_exec']) || $row['outlier'] == 2 ? 0 : $row['id_exec'];
$query_var = "SELECT count(*) as num FROM aloja_ml.predictions WHERE instance = '" . $inst_aux[1] . "' AND id_learner = '" . $current_model . "'";
$result = $dbml->query($query_var);
$row = $result->fetch();
// Insert instance values
if ($row['num'] == 0) {
$token_i = 1;
$selected_instance = preg_replace('/,Cmp(\\d+),/', ',${1},', $inst_aux[1]);
$selected_instance = preg_replace('/,Cl(\\d+),/', ',${1},', $selected_instance);
if ($token > 0) {
$query = $query . ",";
}
$token = 1;
$query = $query . "('" . $realid_exec . "','" . $realexecval . "','" . str_replace(",", "','", $selected_instance) . "','" . $inst_aux[2] . "','" . $current_model . "','" . $inst_aux[1] . "','0') ";
}
$i++;
if ($i % 100 == 0 && $token_i > 0) {
if ($dbml->query($query) === FALSE) {
throw new \Exception('Error when saving into DB');
}
$query = "INSERT IGNORE INTO aloja_ml.predictions (id_exec,exe_time,bench,net,disk,maps,iosf,replication,iofilebuf,comp,blk_size,id_cluster,datanodes,vm_OS,vm_cores,vm_RAM,provider,vm_size,type,bench_type,hadoop_version,pred_time,id_learner,instance,predict_code) VALUES ";
$token = 0;
$token_i = 0;
}
}
if ($token_i > 0) {
if ($dbml->query($query) === FALSE) {
throw new \Exception('Error when saving into DB');
}
}
// Descriptive Tree
$tree_descriptor = shell_exec(getcwd() . '/resources/aloja_cli.r -m aloja_representative_tree -p method=ordered:dump_file="' . getcwd() . '/cache/query/' . $tmp_file . '":output="html" -v 2> /dev/null');
$tree_descriptor = substr($tree_descriptor, 5, -2);
$query = "INSERT INTO aloja_ml.trees (id_findattrs,id_learner,instance,model,tree_code) VALUES ('" . md5($config) . "','" . $current_model . "','" . $instance . "','" . $model_info . "','" . $tree_descriptor . "')";
if ($dbml->query($query) === FALSE) {
throw new \Exception('Error when saving tree into DB');
}
// remove remaining locks
shell_exec('rm -f ' . getcwd() . '/cache/query/' . md5($config) . '*.lock');
// Remove temporal files
$output = shell_exec('rm -f ' . getcwd() . '/cache/query/' . md5($config) . '.tmp');
$is_cached = true;
}
fclose($handle);
}
if (!$is_cached) {
$jsonData = $jsonHeader = $jsonColumns = $jsonColor = '[]';
$must_wait = 'YES';
if (isset($dump)) {
$dbml = null;
echo "1";
exit(0);
}
if (isset($pass)) {
$dbml = null;
return "1";
}
} else {
if (isset($pass) && $pass == 2) {
$dbml = null;
return "2";
}
// Fetch results and compose JSON
$header = array('Benchmark', 'Net', 'Disk', 'Maps', 'IO.SFS', 'Rep', 'IO.FBuf', 'Comp', 'Blk.Size', 'Cluster', 'Datanodes', 'VM.OS', 'VM.Cores', 'VM.RAM', 'Provider', 'VM.Size', 'Type', 'Bench.Type', 'Version', 'Prediction', 'Observed');
$jsonHeader = '[{title:""}';
foreach ($header as $title) {
$jsonHeader = $jsonHeader . ',{title:"' . $title . '"}';
}
$jsonHeader = $jsonHeader . ']';
$query = "SELECT @i:=@i+1 as num, instance, AVG(pred_time) as pred_time, AVG(exe_time) as exe_time FROM aloja_ml.predictions, (SELECT @i:=0) d WHERE id_learner='" . $current_model . "' " . $where_configs . " GROUP BY instance";
$result = $dbml->query($query);
$jsonData = '[';
foreach ($result as $row) {
if ($jsonData != '[') {
$jsonData = $jsonData . ',';
}
$jsonData = $jsonData . "['" . $row['num'] . "','" . str_replace(",", "','", $row['instance']) . "','" . $row['pred_time'] . "','" . $row['exe_time'] . "']";
}
$jsonData = $jsonData . ']';
foreach (range(1, 33) as $value) {
$jsonData = str_replace('Cmp' . $value, Utils::getCompressionName($value), $jsonData);
}
// Fetch MAE & RAE values
$query = "SELECT AVG(ABS(exe_time - pred_time)) AS MAE, AVG(ABS(exe_time - pred_time)/exe_time) AS RAE FROM aloja_ml.predictions WHERE id_learner='" . md5($config) . "' AND predict_code > 0";
$result = $dbml->query($query);
$row = $result->fetch();
$mae = $row['MAE'];
$rae = $row['RAE'];
// Dump case
if (isset($dump)) {
echo "ID" . str_replace(array("[", "]", "{title:\"", "\"}"), array('', '', ''), $jsonHeader) . "\n";
echo str_replace(array('],[', '[[', ']]'), array("\n", '', ''), $jsonData);
$dbml = null;
exit(0);
}
if (isset($pass) && $pass == 1) {
$retval = "ID" . str_replace(array("[", "]", "{title:\"", "\"}"), array('', '', ''), $jsonHeader) . "\n";
$retval .= str_replace(array('],[', '[[', ']]'), array("\n", '', ''), $jsonData);
$dbml = null;
return $retval;
}
// Display Descriptive Tree
$query = "SELECT tree_code FROM aloja_ml.trees WHERE id_findattrs = '" . md5($config) . "'";
$result = $dbml->query($query);
$row = $result->fetch();
$tree_descriptor = $row['tree_code'];
}
} else {
$message = "There are no prediction models trained for such parameters. Train at least one model in 'ML Prediction' section.";
$must_wait = 'NO';
if (isset($dump)) {
echo "-1";
exit(0);
}
if (isset($pass)) {
return "-1";
}
}
$dbml = null;
} catch (\Exception $e) {
$this->container->getTwig()->addGlobal('message', $e->getMessage() . "\n");
$jsonData = $jsonHeader = "[]";
$must_wait = 'NO';
$mae = $rae = 0;
$dbml = null;
if (isset($pass)) {
return "-2";
}
}
$return_params = array('instance' => $instance, 'jsonData' => $jsonData, 'jsonHeader' => $jsonHeader, 'models' => $model_html, 'models_id' => $possible_models_id, 'other_models_id' => $other_models, 'current_model' => $current_model, 'message' => $message, 'mae' => $mae, 'rae' => $rae, 'must_wait' => $must_wait, 'instance' => $instance, 'instances' => implode("<br/> ", $instances), 'model_info' => $model_info, 'id_findattr' => md5($config), 'unseen' => $unseen, 'tree_descriptor' => $tree_descriptor);
$this->filters->setCurrentChoices('current_model', array_merge($possible_models_id, array('---Other models---'), $other_models));
return $this->render('mltemplate/mlfindattributes.html.twig', $return_params);
}
public function mlparamEvaluationAction()
{
$rows = $categories = $series = $instance = $model_info = $config = $current_model = $slice_info = '';
$arrayBenchs_pred = $possible_models = $possible_models_id = $other_models = array();
$jsonData = $jsonHeader = "[]";
$must_wait = 'NO';
try {
$dbml = new \PDO($this->container->get('config')['db_conn_chain'], $this->container->get('config')['mysql_user'], $this->container->get('config')['mysql_pwd']);
$dbml->setAttribute(\PDO::ATTR_ERRMODE, \PDO::ERRMODE_EXCEPTION);
$dbml->setAttribute(\PDO::ATTR_EMULATE_PREPARES, false);
$db = $this->container->getDBUtils();
if (array_key_exists('parameval', $_GET)) {
$paramEval = isset($_GET['parameval']) && Utils::get_GET_string('parameval') != '' ? Utils::get_GET_string('parameval') : 'maps';
unset($_GET["parameval"]);
}
$this->buildFilters(array('current_model' => array('type' => 'selectOne', 'default' => null, 'label' => 'Model to use: ', 'generateChoices' => function () {
return array();
}, 'parseFunction' => function () {
$choice = isset($_GET['current_model']) ? $_GET['current_model'] : array("");
return array('whereClause' => '', 'currentChoice' => $choice);
}, 'filterGroup' => 'MLearning'), 'minExecs' => array('default' => 0, 'type' => 'inputNumber', 'label' => 'Minimum executions:', 'parseFunction' => function () {
return 0;
}, 'filterGroup' => 'basic'), 'minexetime' => array('default' => 0), 'valid' => array('default' => 0), 'filter' => array('default' => 0), 'prepares' => array('default' => 0)));
$this->buildFilterGroups(array('MLearning' => array('label' => 'Machine Learning', 'tabOpenDefault' => true, 'filters' => array('current_model'))));
$where_configs = $this->filters->getWhereClause();
$params = array();
$param_names = array('bench', 'net', 'disk', 'maps', 'iosf', 'replication', 'iofilebuf', 'comp', 'blk_size', 'id_cluster', 'datanodes', 'vm_OS', 'vm_cores', 'vm_RAM', 'provider', 'vm_size', 'type', 'bench_type', 'hadoop_version');
// Order is important
$params = $this->filters->getFiltersSelectedChoices($param_names);
foreach ($param_names as $p) {
if (!is_null($params[$p]) && is_array($params[$p])) {
sort($params[$p]);
}
}
$params_additional = array();
$param_names_additional = array('datefrom', 'dateto', 'minexetime', 'maxexetime', 'valid', 'filter');
// Order is important
$params_additional = $this->filters->getFiltersSelectedChoices($param_names_additional);
$param_variables = $this->filters->getFiltersSelectedChoices(array('current_model', 'minExecs'));
$param_current_model = $param_variables['current_model'];
$minExecs = $param_variables['minExecs'];
$where_configs = str_replace("AND .", "AND ", $where_configs);
$where_configs = str_replace("id_cluster", "e.id_cluster", $where_configs);
$minExecsFilter = "";
if ($minExecs > 0) {
$minExecsFilter = "HAVING COUNT(*) > {$minExecs}";
}
$filter_execs = DBUtils::getFilterExecs();
$options = $this->filters->getFilterChoices();
$paramOptions = array();
foreach ($options[$paramEval] as $option) {
if ($paramEval == 'comp') {
$paramOptions[] = Utils::getCompressionName($option);
} else {
if ($paramEval == 'net') {
$paramOptions[] = Utils::getNetworkName($option);
} else {
if ($paramEval == 'disk') {
$paramOptions[] = Utils::getDisksName($option);
} else {
$paramOptions[] = $option;
}
}
}
}
$param_eval_query = $paramEval == 'id_cluster' ? 'e.id_cluster' : $paramEval;
$benchOptions = $db->get_rows("SELECT DISTINCT bench FROM aloja2.execs e LEFT JOIN aloja2.clusters c ON e.id_cluster = c.id_cluster WHERE 1 {$filter_execs} {$where_configs} GROUP BY {$param_eval_query}, bench order by {$param_eval_query}");
// get the result rows
$query = "SELECT count(*) as count, {$param_eval_query}, e.id_exec, exec as conf, bench, " . "exe_time, avg(exe_time) avg_exe_time, min(exe_time) min_exe_time " . "from aloja2.execs e LEFT JOIN aloja2.clusters c ON e.id_cluster = c.id_cluster WHERE 1 {$filter_execs} {$where_configs}" . "GROUP BY {$param_eval_query},bench {$minExecsFilter} order by bench,{$param_eval_query}";
$rows = $db->get_rows($query);
if (!$rows) {
throw new \Exception("No results for query!");
}
$arrayBenchs = array();
foreach ($paramOptions as $param) {
foreach ($benchOptions as $bench) {
$arrayBenchs[$bench['bench']][$param] = null;
$arrayBenchs[$bench['bench']][$param]['y'] = 0;
$arrayBenchs[$bench['bench']][$param]['count'] = 0;
}
}
$series = array();
$bench = '';
foreach ($rows as $row) {
if ($paramEval == 'comp') {
$row[$paramEval] = Utils::getCompressionName($row['comp']);
} else {
if ($paramEval == 'net') {
$row[$paramEval] = Utils::getNetworkName($row['net']);
} else {
if ($paramEval == 'disk') {
$row[$paramEval] = Utils::getDisksName($row['disk']);
} else {
if ($paramEval == 'iofilebuf') {
$row[$paramEval] /= 1024;
}
}
}
}
$arrayBenchs[$row['bench']][$row[$paramEval]]['y'] = round((int) $row['avg_exe_time'], 2);
$arrayBenchs[$row['bench']][$row[$paramEval]]['count'] = (int) $row['count'];
}
// ----------------------------------------------------
// Add predictions to the series
// ----------------------------------------------------
$param_variables = $this->filters->getFiltersSelectedChoices(array('current_model'));
$param_current_model = $param_variables['current_model'];
$where_configs = str_replace("AND .", "AND ", $where_configs);
// compose instance
$instance = MLUtils::generateSimpleInstance($this->filters, $param_names, $params, true);
$model_info = MLUtils::generateModelInfo($this->filters, $param_names, $params, true);
$slice_info = MLUtils::generateDatasliceInfo($this->filters, $param_names_additional, $params_additional);
// model for filling
MLUtils::findMatchingModels($model_info, $possible_models, $possible_models_id, $dbml);
$current_model = '';
if (!is_null($possible_models_id) && in_array($param_current_model, $possible_models_id)) {
$current_model = $param_current_model;
}
// Other models for filling
$where_models = '';
if (!empty($possible_models_id)) {
$where_models = " WHERE id_learner NOT IN ('" . implode("','", $possible_models_id) . "')";
}
$result = $dbml->query("SELECT id_learner FROM aloja_ml.learners" . $where_models);
foreach ($result as $row) {
$other_models[] = $row['id_learner'];
}
if (!empty($possible_models_id)) {
if ($current_model == "") {
$query = "SELECT AVG(ABS(exe_time - pred_time)) AS MAE, AVG(ABS(exe_time - pred_time)/exe_time) AS RAE, p.id_learner FROM aloja_ml.predictions p, aloja_ml.learners l WHERE l.id_learner = p.id_learner AND p.id_learner IN ('" . implode("','", $possible_models_id) . "') AND predict_code > 0 ORDER BY MAE LIMIT 1";
$result = $dbml->query($query);
$row = $result->fetch();
$current_model = $row['id_learner'];
}
$config = $instance . '-' . $current_model . ' ' . $slice_info . "-parameval";
$query_cache = "SELECT count(*) as total FROM aloja_ml.trees WHERE id_learner = '" . $current_model . "' AND model = '" . $model_info . "'";
$is_cached_mysql = $dbml->query($query_cache);
$tmp_result = $is_cached_mysql->fetch();
$is_cached = $tmp_result['total'] > 0;
$ret_data = null;
if (!$is_cached) {
// Call to MLFindAttributes, to fetch data
$_GET['pass'] = 2;
$_GET['unseen'] = 1;
$_GET['current_model'] = $current_model;
$mlfa1 = new MLFindAttributesController();
$mlfa1->container = $this->container;
$ret_data = $mlfa1->mlfindattributesAction();
if ($ret_data == 1) {
$must_wait = "YES";
$jsonData = $jsonHeader = '[]';
} else {
$is_cached_mysql = $dbml->query($query_cache);
$tmp_result = $is_cached_mysql->fetch();
$is_cached = $tmp_result['total'] > 0;
}
}
if ($is_cached) {
$must_wait = 'NO';
$query = "SELECT count(*) as count, {$param_eval_query}, bench, exe_time, avg(pred_time) avg_pred_time, min(pred_time) min_pred_time " . "FROM aloja_ml.predictions e WHERE e.id_learner = '" . $current_model . "' {$filter_execs} {$where_configs}" . "GROUP BY {$param_eval_query}, bench {$minExecsFilter} order by bench,{$param_eval_query}";
$result = $dbml->query($query);
// Initialize array
foreach ($paramOptions as $param) {
foreach ($benchOptions as $bench) {
$arrayBenchs_pred[$bench['bench'] . '_pred'][$param] = null;
$arrayBenchs_pred[$bench['bench'] . '_pred'][$param]['y'] = 0;
$arrayBenchs_pred[$bench['bench'] . '_pred'][$param]['count'] = 0;
}
}
foreach ($result as $row) {
$bench_n = $row['bench'] . '_pred';
$class = $row[$paramEval];
if ($paramEval == 'comp') {
$value = Utils::getCompressionName($class);
} else {
if ($paramEval == 'id_cluster') {
$value = Utils::getClusterName($class, $db);
} else {
if ($paramEval == 'net') {
$value = Utils::getNetworkName($class);
} else {
if ($paramEval == 'disk') {
$value = Utils::getDisksName($class);
} else {
if ($paramEval == 'iofilebuf') {
$value = $class / 1024;
} else {
$value = $class;
}
}
}
}
}
if (!in_array($value, $paramOptions)) {
$paramOptions[] = $value;
foreach ($benchOptions as $bench) {
$arrayBenchs_pred[$bench['bench'] . '_pred'][$value] = null;
$arrayBenchs_pred[$bench['bench'] . '_pred'][$value]['y'] = 0;
$arrayBenchs_pred[$bench['bench'] . '_pred'][$value]['count'] = 0;
$arrayBenchs[$bench['bench']][$value] = null;
$arrayBenchs[$bench['bench']][$value]['y'] = 0;
$arrayBenchs[$bench['bench']][$value]['count'] = 0;
}
}
$arrayBenchs_pred[$bench_n][$value]['y'] = (int) $row['avg_pred_time'];
$arrayBenchs_pred[$bench_n][$value]['count'] = (int) $row['count'];
}
}
}
// ----------------------------------------------------
// END - Add predictions to the series
// ----------------------------------------------------
asort($paramOptions);
foreach ($arrayBenchs as $key => $arrayBench) {
$caregories = '';
$data_a = null;
$data_p = null;
foreach ($paramOptions as $param) {
if ($arrayBenchs[$key][$param]['count'] > 0 && empty($arrayBenchs_pred) || !empty($arrayBenchs_pred) && ($arrayBenchs_pred[$key . '_pred'][$param]['count'] > 0 || $arrayBenchs[$key][$param]['count'] > 0)) {
$data_a[] = $arrayBenchs[$key][$param];
if (!empty($arrayBenchs_pred)) {
$data_p[] = $arrayBenchs_pred[$key . '_pred'][$param];
}
$categories = $categories . "'{$param} " . Utils::getParamevalUnit($paramEval) . "',";
// FIXME - Redundant n times performed... don't care now
}
}
$series[] = array('name' => $key, 'data' => $data_a);
if (!empty($arrayBenchs_pred)) {
$series[] = array('name' => $key . '_pred', 'data' => $data_p);
}
}
$series = json_encode($series);
if (!empty($arrayBenchs_pred)) {
$colors = "['#7cb5ec','#9cd5fc','#434348','#636368','#90ed7d','#b0fd9d','#f7a35c','#f7c37c','#8085e9','#a0a5f9','#f15c80','#f17ca0','#e4d354','#f4f374','#8085e8','#a0a5f8','#8d4653','#ad6673','#91e8e1','#b1f8f1']";
} else {
$colors = "['#7cb5ec','#434348','#90ed7d','#f7a35c','#8085e9','#f15c80','#e4d354','#8085e8','#8d4653','#91e8e1']";
}
} catch (\Exception $e) {
$this->container->getTwig()->addGlobal('message', $e->getMessage() . "\n");
$series = $jsonHeader = $colors = '[]';
$must_wait = 'NO';
}
$return_params = array('title' => 'Improvement of Hadoop Execution by SW and HW Configurations', 'categories' => $categories, 'series' => $series, 'paramEval' => $paramEval, 'instance' => $instance, 'models' => '<li>' . implode('</li><li>', $possible_models) . '</li>', 'models_id' => $possible_models_id, 'current_model' => $current_model, 'gammacolors' => $colors, 'model_info' => $model_info, 'slice_info' => $slice_info, 'must_wait' => $must_wait);
$this->filters->setCurrentChoices('current_model', array_merge($possible_models_id, array('---Other models---'), $other_models));
return $this->render('mltemplate/mlparameval.html.twig', $return_params);
}
public function mlcrossvar3dfaAction()
{
$jsonData = array();
$message = $instance = $possible_models_id = '';
$maxx = $minx = $maxy = $miny = $maxz = $minz = 0;
$must_wait = 'NO';
try {
$dbml = new \PDO($this->container->get('config')['db_conn_chain_ml'], $this->container->get('config')['mysql_user'], $this->container->get('config')['mysql_pwd']);
$dbml->setAttribute(\PDO::ATTR_ERRMODE, \PDO::ERRMODE_EXCEPTION);
$dbml->setAttribute(\PDO::ATTR_EMULATE_PREPARES, false);
$db = $this->container->getDBUtils();
$where_configs = '';
$preset = null;
if (count($_GET) <= 1 || count($_GET) == 2 && array_key_exists('current_model', $_GET) || count($_GET) == 3 && array_key_exists('variable1', $_GET) && array_key_exists('variable2', $_GET) || count($_GET) == 4 && array_key_exists('current_model', $_GET) && array_key_exists('variable1', $_GET) && array_key_exists('variable2', $_GET)) {
$preset = Utils::setDefaultPreset($db, 'mlcrossvar3dfa');
}
$selPreset = isset($_GET['presets']) ? $_GET['presets'] : "none";
$params = array();
$param_names = array('benchs', 'nets', 'disks', 'mapss', 'iosfs', 'replications', 'iofilebufs', 'comps', 'blk_sizes', 'id_clusters', 'datanodess', 'bench_types', 'vm_sizes', 'vm_coress', 'vm_RAMs', 'types');
// Order is important
foreach ($param_names as $p) {
$params[$p] = Utils::read_params($p, $where_configs, FALSE);
sort($params[$p]);
}
$cross_var1 = array_key_exists('variable1', $_GET) ? $_GET['variable1'] : 'maps';
$cross_var2 = array_key_exists('variable2', $_GET) ? $_GET['variable2'] : 'net';
$unseen = array_key_exists('unseen', $_GET) && $_GET['unseen'] == 1;
$where_configs = str_replace("AND .", "AND ", $where_configs);
$cross_var1 = str_replace("id_cluster", "e.id_cluster", $cross_var1);
$cross_var2 = str_replace("id_cluster", "e.id_cluster", $cross_var2);
// compose instance
$instance = MLUtils::generateSimpleInstance($param_names, $params, $unseen, $db);
$model_info = MLUtils::generateModelInfo($param_names, $params, $unseen, $db);
// Model for filling
MLUtils::findMatchingModels($model_info, $possible_models, $possible_models_id, $dbml);
$current_model = "";
if (array_key_exists('current_model', $_GET) && in_array($_GET['current_model'], $possible_models_id)) {
$current_model = $_GET['current_model'];
}
// Call to MLFindAttributes, to fetch data
$_GET['pass'] = 1;
$_GET['unseen'] = $unseen;
$mlfa1 = new MLFindAttributesController();
$mlfa1->container = $this->container;
$ret_data = $mlfa1->mlfindattributesAction();
$rows = null;
if ($ret_data == 1) {
$must_wait = "YES";
$jsonData = '[]';
$categories1 = $categories2 = "''";
} else {
if ($ret_data == -1) {
$must_wait = "NO";
$jsonData = '[]';
$categories1 = $categories2 = "''";
$message = "There are no prediction models trained for such parameters. Train at least one model in 'ML Prediction' section. [" . $instance . "]";
} else {
// Get stuff from the DB
$query = "SELECT " . $cross_var1 . " AS V1, " . $cross_var2 . " AS V2, AVG(p.pred_time) as V3, p.instance\n\t\t\t\t\tFROM predictions as p\n\t\t\t\t\tWHERE p.id_learner " . ($current_model != '' ? "='" . $current_model . "'" : "IN (SELECT id_learner FROM trees WHERE model='" . $model_info . "')") . $where_configs . "\n\t\t\t\t\tGROUP BY p.instance\n\t\t\t\t\tORDER BY RAND() LIMIT 5000;";
// FIXME - CLUMPSY PATCH FOR BYPASS THE BUG FROM HIGHCHARTS... REMEMBER TO ERASE THIS LINE WHEN THE BUG IS SOLVED
$rows = $dbml->query($query);
if (empty($rows)) {
throw new \Exception('No data matches with your critteria.');
}
}
}
if ($must_wait == "NO") {
$map_var1 = $map_var2 = array();
$count_var1 = $count_var2 = 0;
$categories1 = $categories2 = '';
$var1_categorical = in_array($cross_var1, array("net", "disk", "bench", "vm_OS", "provider", "vm_size", "type", "bench_type"));
$var2_categorical = in_array($cross_var2, array("net", "disk", "bench", "vm_OS", "provider", "vm_size", "type", "bench_type"));
foreach ($rows as $row) {
$entry = array();
if ($var1_categorical) {
if (!array_key_exists($row['V1'], $map_var1)) {
$map_var1[$row['V1']] = $count_var1++;
$categories1 = $categories1 . ($categories1 != '' ? "," : "") . "\"" . $row['V1'] . "\"";
}
$entry['y'] = $map_var1[$row['V1']] * (rand(990, 1010) / 1000);
} else {
$entry['y'] = (int) $row['V1'] * (rand(990, 1010) / 1000);
}
if ($entry['y'] > $maxy) {
$maxy = $entry['y'];
}
if ($entry['y'] < $miny) {
$miny = $entry['y'];
}
if ($var2_categorical) {
if (!array_key_exists($row['V2'], $map_var2)) {
$map_var2[$row['V2']] = $count_var2++;
$categories2 = $categories2 . ($categories2 != '' ? "," : "") . "\"" . $row['V2'] . "\"";
}
$entry['x'] = $map_var2[$row['V2']] * (rand(990, 1010) / 1000);
} else {
$entry['x'] = (int) $row['V2'] * (rand(990, 1010) / 1000);
}
if ($entry['x'] > $maxx) {
$maxx = $entry['x'];
}
if ($entry['x'] < $minx) {
$minx = $entry['x'];
}
$entry['z'] = -1 * (int) $row['V3'] * (rand(990, 1010) / 1000);
if ($entry['z'] > $maxz) {
$maxz = $entry['z'];
}
if ($entry['z'] < $minz) {
$minz = $entry['z'];
}
$entry['name'] = $row['instance'];
//$row['V1']." - ".$row['V2']." - ".max(100,(int)$row['V3']);
$jsonData[] = $entry;
}
$jsonData = json_encode($jsonData);
if ($categories1 != '') {
$categories1 = "[" . $categories1 . "]";
} else {
$categories1 = "''";
}
if ($categories2 != '') {
$categories2 = "[" . $categories2 . "]";
} else {
$categories2 = "''";
}
}
$dbml = null;
$cross_var1 = str_replace("e.id_cluster", "id_cluster", $cross_var1);
$cross_var2 = str_replace("e.id_cluster", "id_cluster", $cross_var2);
} catch (\Exception $e) {
$this->container->getTwig()->addGlobal('message', $e->getMessage() . "\n");
$jsonData = '[]';
$cross_var1 = $cross_var2 = '';
$categories1 = $categories2 = '';
$maxx = $minx = $maxy = $miny = $maxz = $minz = 0;
$must_wait = "NO";
$dbml = null;
$possible_models = $possible_models_id = array();
}
echo $this->container->getTwig()->render('mltemplate/mlcrossvar3dfa.html.twig', array('selected' => 'mlcrossvar3dfa', 'jsonData' => $jsonData, 'variable1' => $cross_var1, 'variable2' => $cross_var2, 'categories1' => $categories1, 'categories2' => $categories2, 'maxx' => $maxx, 'minx' => $minx, 'maxy' => $maxy, 'miny' => $miny, 'maxz' => $maxz, 'minz' => $minz, 'benchs' => $params['benchs'], 'nets' => $params['nets'], 'disks' => $params['disks'], 'blk_sizes' => $params['blk_sizes'], 'comps' => $params['comps'], 'id_clusters' => $params['id_clusters'], 'mapss' => $params['mapss'], 'replications' => $params['replications'], 'iosfs' => $params['iosfs'], 'iofilebufs' => $params['iofilebufs'], 'datanodess' => $params['datanodess'], 'bench_types' => $params['bench_types'], 'vm_sizes' => $params['vm_sizes'], 'vm_coress' => $params['vm_coress'], 'vm_RAMs' => $params['vm_RAMs'], 'types' => $params['types'], 'message' => $message, 'instance' => $instance, 'model_info' => $model_info, 'current_model' => $current_model, 'unseen' => $unseen, 'models' => '<li>' . implode('</li><li>', $possible_models) . '</li>', 'models_id' => $possible_models_id, 'must_wait' => $must_wait, 'preset' => $preset, 'selPreset' => $selPreset, 'options' => Utils::getFilterOptions($db)));
}
public function mldatacollapseAction()
{
try {
$dbml = new \PDO($this->container->get('config')['db_conn_chain'], $this->container->get('config')['mysql_user'], $this->container->get('config')['mysql_pwd']);
$dbml->setAttribute(\PDO::ATTR_ERRMODE, \PDO::ERRMODE_EXCEPTION);
$dbml->setAttribute(\PDO::ATTR_EMULATE_PREPARES, false);
$db = $this->container->getDBUtils();
$where_configs = '';
$preset = null;
if (count($_GET) <= 1 || count($_GET) == 2 && array_key_exists("current_model", $_GET)) {
$preset = Utils::initDefaultPreset($db, 'mldatacollapse');
}
$selPreset = isset($_GET['presets']) ? $_GET['presets'] : "none";
$params = array();
$param_names = array('benchs', 'nets', 'disks', 'mapss', 'iosfs', 'replications', 'iofilebufs', 'comps', 'blk_sizes', 'id_clusters', 'datanodess', 'bench_types', 'vm_sizes', 'vm_coress', 'vm_RAMs', 'types', 'hadoop_versions');
// Order is important
foreach ($param_names as $p) {
$params[$p] = Utils::read_params($p, $where_configs, FALSE);
sort($params[$p]);
}
$params_additional = array();
$param_names_additional = array('datefrom', 'dateto', 'minexetime', 'maxexetime', 'valids', 'filters');
// Order is important
foreach ($param_names_additional as $p) {
$params_additional[$p] = Utils::read_params($p, $where_configs, FALSE);
}
$unseen = array_key_exists('unseen', $_GET) && $_GET['unseen'] == 1;
// FIXME PATCH FOR PARAM LIBRARIES WITHOUT LEGACY
$where_configs = str_replace("AND .", "AND ", $where_configs);
$dims1 = (empty($params['nets']) ? '' : 'Net,') . (empty($params['disks']) ? '' : 'Disk,') . (empty($params['blk_sizes']) ? '' : 'Blk.size,') . (empty($params['comps']) ? '' : 'Comp,');
$dims1 = $dims1 . (empty($params['id_clusters']) ? '' : 'Cluster,') . (empty($params['mapss']) ? '' : 'Maps,') . (empty($params['replications']) ? '' : 'Rep,') . (empty($params['iosfs']) ? '' : 'IO.SFac,') . (empty($params['iofilebufs']) ? '' : 'IO.FBuf,');
$dims1 = $dims1 . (empty($params['hadoop_versionss']) ? '' : 'Version,') . (empty($params['bench_types']) ? '' : 'Bench.Type');
if (substr($dims1, -1) == ',') {
$dims1 = substr($dims1, 0, -1);
}
$dims2 = "Benchmark";
// compose instance
$instance = MLUtils::generateSimpleInstance($this->filters, $param_names, $params, true);
$model_info = MLUtils::generateModelInfo($this->filters, $param_names, $params, true);
$slice_info = MLUtils::generateDatasliceInfo($this->filters, $param_names_additional, $params_additional);
// select model for filling
$possible_models = $possible_models_id = array();
MLUtils::findMatchingModels($model_info, $possible_models, $possible_models_id, $dbml);
$current_model = '';
/* if (array_key_exists('current_model',$_GET) && in_array($_GET['current_model'],$possible_models_id)) $current_model = $_GET['current_model']; // FIXME - Needs re-think logic
if ($current_model == '')
{
$query = "SELECT AVG(ABS(exe_time - pred_time)) AS MAE, AVG(ABS(exe_time - pred_time)/exe_time) AS RAE, p.id_learner FROM predictions p, learners l WHERE l.id_learner = p.id_learner AND p.id_learner IN ('".implode("','",$possible_models_id)."') AND predict_code > 0 ORDER BY MAE LIMIT 1";
$result = $dbml->query($query);
$row = $result->fetch();
$current_model = $row['id_learner'];
}
*/
$learning_model = '';
if ($current_model != '' && file_exists(getcwd() . '/cache/query/' . $current_model . '-object.rds')) {
$learning_model = ':model_name=' . $current_model . ':inst_general="' . $instance . '"';
}
$config = $dims1 . '-' . $dims2 . '-' . $current_model . '-' . $model_info . '-' . $slice_info;
// get headers for csv
$header_names = array('id_exec' => 'ID', 'bench' => 'Benchmark', 'exe_time' => 'Exe.Time', 'net' => 'Net', 'disk' => 'Disk', 'maps' => 'Maps', 'iosf' => 'IO.SFac', 'replication' => 'Rep', 'iofilebuf' => 'IO.FBuf', 'comp' => 'Comp', 'blk_size' => 'Blk.size', 'e.id_cluster' => 'Cluster', 'name' => 'Cl.Name', 'datanodes' => 'Datanodes', 'headnodes' => 'Headnodes', 'vm_OS' => 'VM.OS', 'vm_cores' => 'VM.Cores', 'vm_RAM' => 'VM.RAM', 'provider' => 'Provider', 'vm_size' => 'VM.Size', 'type' => 'Type', 'bench_type' => 'Bench.Type', 'hadoop_version' => 'Hadoop.Version');
$headers = array_keys($header_names);
$names = array_values($header_names);
$dims1_array = explode(",", $dims1);
$dims1_query = '';
$dims1_title = $dims1_concat = '';
foreach ($dims1_array as $d1value) {
$dims1_query = $dims1_query . ($dims1_query == '' ? '' : ',') . array_search($d1value, $header_names);
$dims1_title = $dims1_title . ($dims1_title == '' ? '' : ':') . array_search($d1value, $header_names);
$dims1_concat = $dims1_concat . ($dims1_concat == '' ? '' : ',":",') . array_search($d1value, $header_names);
}
$query = "SELECT distinct bench FROM aloja2.execs e LEFT JOIN aloja2.clusters c ON e.id_cluster = c.id_cluster WHERE hadoop_version IS NOT NULL" . $where_configs . " ORDER BY bench;";
$rows = $db->get_rows($query);
if (empty($rows)) {
throw new \Exception('No data matches with your critteria.');
}
$table = array();
$jsonHeader = '[{title:"' . $dims1_title . '"}';
foreach ($rows as $row) {
$jsonHeader = $jsonHeader . ',{title:"' . $row['bench'] . '"}';
$table[$row['bench']] = array();
}
$jsonHeader = $jsonHeader . ']';
$query = "SELECT CONCAT(" . $dims1_concat . ") as dim1, bench, avg(exe_time) as avg_exe_time FROM aloja2.execs e LEFT JOIN aloja2.clusters c ON e.id_cluster = c.id_cluster WHERE hadoop_version IS NOT NULL" . $where_configs . " GROUP BY bench," . $dims1_query . " ORDER BY dim1,bench;";
$rows = $db->get_rows($query);
if (empty($rows)) {
throw new \Exception('No data matches with your critteria.');
}
foreach ($rows as $row) {
$table[$row['bench']][$row['dim1']] = (int) $row['avg_exe_time'];
}
$row_ids = array();
foreach ($table as $bmk) {
foreach ($bmk as $key => $value) {
$row_ids[] = $key;
}
}
$row_ids = array_unique($row_ids);
$tableColor = array();
foreach ($table as $bmk => $values) {
$tableColor[$bmk] = array();
foreach ($row_ids as $rid) {
if (!array_key_exists($rid, $table[$bmk])) {
$table[$bmk][$rid] = 0;
$tableColor[$bmk][$rid] = 0;
} else {
$tableColor[$bmk][$rid] = 1;
}
}
}
$jsonData = '[';
$jsonColor = '[';
foreach ($row_ids as $rid) {
$jsonData = $jsonData . ($jsonData == '[' ? '' : ',') . '[\'' . $rid . '\'';
$jsonColor = $jsonColor . ($jsonColor == '[' ? '' : ',') . '[1';
foreach ($table as $bmk => $values) {
$jsonData = $jsonData . ',' . $table[$bmk][$rid];
$jsonColor = $jsonColor . ',' . $tableColor[$bmk][$rid];
}
$jsonData = $jsonData . ']';
$jsonColor = $jsonColor . ']';
}
$jsonData = $jsonData . ']';
$jsonColor = $jsonColor . ']';
$jsonColumns = '[';
for ($i = 1; $i <= count($table); $i++) {
if ($jsonColumns != '[') {
$jsonColumns = $jsonColumns . ',';
}
$jsonColumns = $jsonColumns . $i;
}
$jsonColumns = $jsonColumns . ']';
} catch (\Exception $e) {
$this->container->getTwig()->addGlobal('message', $e->getMessage() . "\n");
$jsonData = $jsonHeader = $jsonColumns = $jsonColor = '[]';
}
$return_params = array('selected' => 'mldatacollapse', 'jsonEncoded' => $jsonData, 'jsonHeader' => $jsonHeader, 'jsonColumns' => $jsonColumns, 'jsonColor' => $jsonColor, 'instance' => $instance, 'instance' => $instance, 'model_info' => $model_info, 'slice_info' => $slice_info, 'preset' => $preset, 'selPreset' => $selPreset, 'options' => Utils::getFilterOptions($db));
foreach ($param_names as $p) {
$return_params[$p] = $params[$p];
}
foreach ($param_names_additional as $p) {
$return_params[$p] = $params_additional[$p];
}
echo $this->container->getTwig()->render('mltemplate/mldatacollapse.html.twig', $return_params);
}
public function mlcrossvar3dfaAction()
{
$jsonData = $possible_models = $possible_models_id = $other_models = array();
$message = $instance = $possible_models_id = '';
$categories1 = $categories2 = "''";
$maxx = $minx = $maxy = $miny = $maxz = $minz = 0;
$must_wait = 'NO';
try {
$dbml = new \PDO($this->container->get('config')['db_conn_chain'], $this->container->get('config')['mysql_user'], $this->container->get('config')['mysql_pwd']);
$dbml->setAttribute(\PDO::ATTR_ERRMODE, \PDO::ERRMODE_EXCEPTION);
$dbml->setAttribute(\PDO::ATTR_EMULATE_PREPARES, false);
$db = $this->container->getDBUtils();
$this->buildFilters(array('variable2' => array('type' => 'selectOne', 'default' => array('net'), 'table' => 'execs', 'label' => 'Variable 2: ', 'generateChoices' => function () {
return array('bench', 'net', 'disk', 'maps', 'iosf', 'replication', 'iofilebuf', 'comp', 'blk_size', 'id_cluster', 'datanodes', 'bench_type', 'vm_size', 'vm_cores', 'vm_RAM', 'type', 'hadoop_version', 'provider', 'vm_OS', 'exe_time', 'pred_time', 'TOTAL_MAPS');
}, 'beautifier' => function ($value) {
$labels = array('bench' => 'Benchmark', 'net' => 'Network', 'disk' => 'Disk', 'maps' => 'Maps', 'iosf' => 'I/O Sort Factor', 'replication' => 'Replication', 'iofilebuf' => 'I/O File Buffer', 'comp' => 'Compression', 'blk_size' => 'Block size', 'id_cluster' => 'Cluster', 'datanodes' => 'Datanodes', 'bench_type' => 'Benchmark Suite', 'vm_size' => 'VM Size', 'vm_cores' => 'VM cores', 'vm_RAM' => 'VM RAM', 'type' => 'Cluster type', 'hadoop_version' => 'Hadoop Version', 'provider' => 'Provider', 'vm_OS' => 'VM OS', 'exe_time' => 'Exeuction time', 'pred_time' => 'Prediction time', 'TOTAL_MAPS' => 'Total execution maps');
return $labels[$value];
}, 'parseFunction' => function () {
$value = isset($_GET['variable2']) ? $_GET['variable2'] : array('net');
return array('currentChoice' => $value, 'whereClause' => "");
}), 'variable1' => array('type' => 'selectOne', 'default' => array('maps'), 'table' => 'execs', 'label' => 'Variable 1: ', 'generateChoices' => function () {
return array('bench', 'net', 'disk', 'maps', 'iosf', 'replication', 'iofilebuf', 'comp', 'blk_size', 'id_cluster', 'datanodes', 'bench_type', 'vm_size', 'vm_cores', 'vm_RAM', 'type', 'hadoop_version', 'provider', 'vm_OS', 'exe_time', 'pred_time', 'TOTAL_MAPS');
}, 'beautifier' => function ($value) {
$labels = array('bench' => 'Benchmark', 'net' => 'Network', 'disk' => 'Disk', 'maps' => 'Maps', 'iosf' => 'I/O Sort Factor', 'replication' => 'Replication', 'iofilebuf' => 'I/O File Buffer', 'comp' => 'Compression', 'blk_size' => 'Block size', 'id_cluster' => 'Cluster', 'datanodes' => 'Datanodes', 'bench_type' => 'Benchmark Suite', 'vm_size' => 'VM Size', 'vm_cores' => 'VM cores', 'vm_RAM' => 'VM RAM', 'type' => 'Cluster type', 'hadoop_version' => 'Hadoop Version', 'provider' => 'Provider', 'vm_OS' => 'VM OS', 'exe_time' => 'Exeuction time', 'pred_time' => 'Prediction time', 'TOTAL_MAPS' => 'Total execution maps');
return $labels[$value];
}, 'parseFunction' => function () {
$value = isset($_GET['variable1']) ? $_GET['variable1'] : array('maps');
return array('currentChoice' => $value, 'whereClause' => "");
}), 'current_model' => array('type' => 'selectOne', 'default' => null, 'label' => 'Model to use: ', 'generateChoices' => function () {
return array();
}, 'parseFunction' => function () {
$choice = isset($_GET['current_model']) ? $_GET['current_model'] : array("");
return array('whereClause' => '', 'currentChoice' => $choice);
}, 'filterGroup' => 'MLearning'), 'unseen' => array('type' => 'checkbox', 'default' => 1, 'label' => 'Predict with unseen atributes ⚠', 'parseFunction' => function () {
$choice = isset($_GET['unseen']) && !isset($_GET['unseen']) ? 0 : 1;
return array('whereClause' => '', 'currentChoice' => $choice);
}, 'filterGroup' => 'MLearning'), 'minexetime' => array('default' => 0), 'valid' => array('default' => 0), 'filter' => array('default' => 0), 'prepares' => array('default' => 1)));
$this->buildFilterGroups(array('MLearning' => array('label' => 'Machine Learning', 'tabOpenDefault' => true, 'filters' => array('current_model', 'unseen'))));
$where_configs = $this->filters->getWhereClause();
$model_html = '';
$model_info = $db->get_rows("SELECT id_learner, model, algorithm, dataslice FROM aloja_ml.learners");
foreach ($model_info as $row) {
$model_html = $model_html . "<li><b>" . $row['id_learner'] . "</b> => " . $row['algorithm'] . " : " . $row['model'] . " : " . $row['dataslice'] . "</li>";
}
$params = array();
$param_names = array('bench', 'net', 'disk', 'maps', 'iosf', 'replication', 'iofilebuf', 'comp', 'blk_size', 'id_cluster', 'datanodes', 'vm_OS', 'vm_cores', 'vm_RAM', 'provider', 'vm_size', 'type', 'bench_type', 'hadoop_version');
// Order is important
$params = $this->filters->getFiltersSelectedChoices($param_names);
foreach ($param_names as $p) {
if (!is_null($params[$p]) && is_array($params[$p])) {
sort($params[$p]);
}
}
$params_additional = array();
$param_names_additional = array('datefrom', 'dateto', 'minexetime', 'maxexetime', 'valid', 'filter');
// Order is important
$params_additional = $this->filters->getFiltersSelectedChoices($param_names_additional);
$variables = $this->filters->getFiltersSelectedChoices(array('variable1', 'variable2', 'current_model', 'unseen'));
$cross_var1 = $variables['variable1'];
$cross_var2 = $variables['variable2'];
$param_current_model = $variables['current_model'];
$unseen = $variables['unseen'] ? true : false;
$where_configs = str_replace("AND .", "AND ", $where_configs);
$cross_var1 = str_replace("id_cluster", "e.id_cluster", $cross_var1);
$cross_var2 = str_replace("id_cluster", "e.id_cluster", $cross_var2);
// compose instance
$instance = MLUtils::generateSimpleInstance($this->filters, $param_names, $params, true);
$model_info = MLUtils::generateModelInfo($this->filters, $param_names, $params, true);
$slice_info = MLUtils::generateDatasliceInfo($this->filters, $param_names_additional, $params_additional);
// Model for filling
MLUtils::findMatchingModels($model_info, $possible_models, $possible_models_id, $dbml);
$current_model = in_array($param_current_model, $possible_models_id) ? $param_current_model : '';
// Other models for filling
$where_models = '';
if (!empty($possible_models_id)) {
$where_models = " WHERE id_learner NOT IN ('" . implode("','", $possible_models_id) . "')";
}
$result = $dbml->query("SELECT id_learner FROM aloja_ml.learners" . $where_models);
foreach ($result as $row) {
$other_models[] = $row['id_learner'];
}
// Call to MLPrediction, to create a model
if (empty($possible_models_id)) {
$_GET['pass'] = 1;
$mltc1 = new MLPredictionController();
// FIXME - Choose the default modeling algorithm
$mltc1->container = $this->container;
$ret_learn = $mltc1->mlpredictionAction();
$rows = null;
if ($ret_data == 1) {
$must_wait = "YES";
throw new \Exception("WAIT");
} else {
if ($ret_data == -1) {
throw new \Exception("There was an error when creating a model for [" . $instance . "]");
}
}
}
// Call to MLFindAttributes, to generate data
if ($current_model != '') {
$_GET['pass'] = 2;
$mlfa1 = new MLFindAttributesController();
$mlfa1->container = $this->container;
$ret_data = $mlfa1->mlfindattributesAction();
$rows = null;
if ($ret_data == 1) {
$must_wait = "YES";
throw new \Exception("WAIT");
} else {
if ($ret_data == -1) {
throw new \Exception("There was an error when creating predictions for [" . $instance . "]");
}
}
}
// Get stuff from the DB
$query = "SELECT " . $cross_var1 . " AS V1, " . $cross_var2 . " AS V2, AVG(e.pred_time) as V3, e.instance\n\t\t\t\tFROM aloja_ml.predictions as e\n\t\t\t\tWHERE e.id_learner " . ($current_model != '' ? "='" . $current_model . "'" : "IN (SELECT id_learner FROM aloja_ml.trees WHERE model='" . $model_info . "')") . $where_configs . "\n\t\t\t\tGROUP BY e.instance\n\t\t\t\tORDER BY RAND() LIMIT 5000;";
// FIXME - CLUMPSY PATCH FOR BYPASS THE BUG FROM HIGHCHARTS... REMEMBER TO ERASE THIS LINE WHEN THE BUG IS SOLVED
$rows = $db->get_rows($query);
if (empty($rows)) {
if ($current_model == '') {
throw new \Exception('No data matches with your critteria. Try to select a specific model to generate data.');
} else {
throw new \Exception('No data matches with your critteria.');
}
}
// Show the results
$map_var1 = $map_var2 = array();
$count_var1 = $count_var2 = 0;
$categories1 = $categories2 = '';
$var1_categorical = in_array($cross_var1, array("net", "disk", "bench", "vm_OS", "provider", "vm_size", "type", "bench_type"));
$var2_categorical = in_array($cross_var2, array("net", "disk", "bench", "vm_OS", "provider", "vm_size", "type", "bench_type"));
foreach ($rows as $row) {
$entry = array();
if ($var1_categorical) {
if (!array_key_exists($row['V1'], $map_var1)) {
$map_var1[$row['V1']] = $count_var1++;
$categories1 = $categories1 . ($categories1 != '' ? "," : "") . "\"" . $row['V1'] . "\"";
}
$entry['y'] = $map_var1[$row['V1']] * (rand(990, 1010) / 1000);
} else {
$entry['y'] = (int) $row['V1'] * (rand(990, 1010) / 1000);
}
if ($entry['y'] > $maxy) {
$maxy = $entry['y'];
}
if ($entry['y'] < $miny) {
$miny = $entry['y'];
}
if ($var2_categorical) {
if (!array_key_exists($row['V2'], $map_var2)) {
$map_var2[$row['V2']] = $count_var2++;
$categories2 = $categories2 . ($categories2 != '' ? "," : "") . "\"" . $row['V2'] . "\"";
}
$entry['x'] = $map_var2[$row['V2']] * (rand(990, 1010) / 1000);
} else {
$entry['x'] = (int) $row['V2'] * (rand(990, 1010) / 1000);
}
if ($entry['x'] > $maxx) {
$maxx = $entry['x'];
}
if ($entry['x'] < $minx) {
$minx = $entry['x'];
}
$entry['z'] = (int) $row['V3'] * (rand(990, 1010) / 1000);
if ($entry['z'] > $maxz) {
$maxz = $entry['z'];
}
if ($entry['z'] < $minz) {
$minz = $entry['z'];
}
$entry['name'] = $row['instance'];
//$row['V1']." - ".$row['V2']." - ".max(100,(int)$row['V3']);
$jsonData[] = $entry;
}
$jsonData = json_encode($jsonData);
if ($categories1 != '') {
$categories1 = "[" . $categories1 . "]";
} else {
$categories1 = "''";
}
if ($categories2 != '') {
$categories2 = "[" . $categories2 . "]";
} else {
$categories2 = "''";
}
} catch (\Exception $e) {
if ($e->getMessage() != "WAIT") {
$this->container->getTwig()->addGlobal('message', $e->getMessage() . "\n");
}
$jsonData = '[]';
}
$dbml = null;
$return_params = array('jsonData' => $jsonData, 'variable1' => str_replace("e.id_cluster", "id_cluster", $cross_var1), 'variable2' => str_replace("e.id_cluster", "id_cluster", $cross_var2), 'categories1' => $categories1, 'categories2' => $categories2, 'maxx' => $maxx, 'minx' => $minx, 'maxy' => $maxy, 'miny' => $miny, 'maxz' => $maxz, 'minz' => $minz, 'instance' => $instance, 'model_info' => $model_info, 'slice_info' => $slice_info, 'models' => '<li>' . implode('</li><li>', $possible_models) . '</li>', 'must_wait' => $must_wait, 'models' => $model_html, 'current_model' => $current_model);
$this->filters->setCurrentChoices('current_model', array_merge(array("Aggregation of Models"), $possible_models_id, !empty($other_models) ? array('---Other models---') : array(), $other_models));
return $this->render('mltemplate/mlcrossvar3dfa.html.twig', $return_params);
}
public function mloutliersAction()
{
$jsonData = $jsonWarns = $jsonOuts = array();
$message = $instance = $jsonHeader = $jsonTable = '';
$max_x = $max_y = 0;
$must_wait = 'NO';
try {
$dbml = new \PDO($this->container->get('config')['db_conn_chain_ml'], $this->container->get('config')['mysql_user'], $this->container->get('config')['mysql_pwd']);
$dbml->setAttribute(\PDO::ATTR_ERRMODE, \PDO::ERRMODE_EXCEPTION);
$dbml->setAttribute(\PDO::ATTR_EMULATE_PREPARES, false);
$db = $this->container->getDBUtils();
$where_configs = '';
$preset = null;
if (count($_GET) <= 1 || count($_GET) == 2 && array_key_exists('current_model', $_GET) || count($_GET) == 2 && array_key_exists('dump', $_GET) || count($_GET) == 2 && array_key_exists('register', $_GET) || count($_GET) == 3 && array_key_exists('dump', $_GET) && array_key_exists('current_model', $_GET) || count($_GET) == 3 && array_key_exists('register', $_GET) && array_key_exists('current_model', $_GET)) {
$preset = Utils::setDefaultPreset($db, 'mloutliers');
}
$selPreset = isset($_GET['presets']) ? $_GET['presets'] : "none";
$params = array();
$param_names = array('benchs', 'nets', 'disks', 'mapss', 'iosfs', 'replications', 'iofilebufs', 'comps', 'blk_sizes', 'id_clusters', 'datanodess', 'bench_types', 'vm_sizes', 'vm_coress', 'vm_RAMs', 'types');
// Order is important
foreach ($param_names as $p) {
$params[$p] = Utils::read_params($p, $where_configs, FALSE);
sort($params[$p]);
}
$sigma_param = array_key_exists('sigma', $_GET) ? (int) $_GET['sigma'] : 1;
// FIXME PATCH FOR PARAM LIBRARIES WITHOUT LEGACY
$where_configs = str_replace("AND .", "AND ", $where_configs);
// compose instance
$instance = MLUtils::generateSimpleInstance($param_names, $params, true, $db);
// Used only as indicator for WEB
$model_info = MLUtils::generateModelInfo($param_names, $params, true, $db);
// model for filling
MLUtils::findMatchingModels($model_info, $possible_models, $possible_models_id, $dbml);
$current_model = "";
if (array_key_exists('current_model', $_GET) && in_array($_GET['current_model'], $possible_models_id)) {
$current_model = $_GET['current_model'];
}
if (!empty($possible_models_id)) {
if ($current_model == "") {
$query = "SELECT AVG(ABS(exe_time - pred_time)) AS MAE, AVG(ABS(exe_time - pred_time)/exe_time) AS RAE, p.id_learner FROM predictions p, learners l WHERE l.id_learner = p.id_learner AND p.id_learner IN ('" . implode("','", $possible_models_id) . "') AND predict_code > 0 ORDER BY MAE LIMIT 1";
$result = $dbml->query($query);
$row = $result->fetch();
$current_model = $row['id_learner'];
}
$config = $instance . '-' . $current_model . '-' . $sigma_param . '-outliers';
$is_cached_mysql = $dbml->query("SELECT count(*) as total FROM resolutions WHERE id_resolution = '" . md5($config) . "'");
$tmp_result = $is_cached_mysql->fetch();
$is_cached = $tmp_result['total'] > 0;
$cache_ds = getcwd() . '/cache/query/' . md5($config) . '-cache.csv';
$in_process = file_exists(getcwd() . '/cache/query/' . md5($config) . '.lock');
$finished_process = file_exists(getcwd() . '/cache/query/' . md5($config) . '-resolutions.csv');
if (!$is_cached && !$in_process && !$finished_process) {
// get headers for csv
$header_names = array('id_exec' => 'ID', 'bench' => 'Benchmark', 'exe_time' => 'Exe.Time', 'net' => 'Net', 'disk' => 'Disk', 'maps' => 'Maps', 'iosf' => 'IO.SFac', 'replication' => 'Rep', 'iofilebuf' => 'IO.FBuf', 'comp' => 'Comp', 'blk_size' => 'Blk.size', 'e.id_cluster' => 'Cluster', 'name' => 'Cl.Name', 'datanodes' => 'Datanodes', 'headnodes' => 'Headnodes', 'vm_OS' => 'VM.OS', 'vm_cores' => 'VM.Cores', 'vm_RAM' => 'VM.RAM', 'provider' => 'Provider', 'vm_size' => 'VM.Size', 'type' => 'Type', 'bench_type' => 'Bench.Type');
$headers = array_keys($header_names);
$names = array_values($header_names);
// dump the result to csv
$query = "SELECT " . implode(",", $headers) . " FROM execs e LEFT JOIN clusters c ON e.id_cluster = c.id_cluster WHERE e.valid = TRUE AND e.exe_time > 100" . $where_configs . ";";
$rows = $db->get_rows($query);
if (empty($rows)) {
throw new \Exception('No data matches with your critteria.');
}
$fp = fopen($cache_ds, 'w');
fputcsv($fp, $names, ',', '"');
foreach ($rows as $row) {
$row['id_cluster'] = "Cl" . $row['id_cluster'];
// Cluster is numerically codified...
$row['comp'] = "Cmp" . $row['comp'];
// Compression is numerically codified...
fputcsv($fp, array_values($row), ',', '"');
}
// Retrieve file model from DB
$query = "SELECT file FROM model_storage WHERE id_hash='" . $current_model . "' AND type='learner';";
$result = $dbml->query($query);
$row = $result->fetch();
$content = $row['file'];
$filemodel = getcwd() . '/cache/query/' . $current_model . '-object.rds';
$fp = fopen($filemodel, 'w');
fwrite($fp, $content);
fclose($fp);
// launch query
exec('cd ' . getcwd() . '/cache/query ; touch ' . md5($config) . '.lock');
exec(getcwd() . '/resources/queue -c "cd ' . getcwd() . '/cache/query ; ' . getcwd() . '/resources/aloja_cli.r -m aloja_outlier_dataset -d ' . $cache_ds . ' -l ' . $current_model . ' -p sigma=' . $sigma_param . ':hdistance=3:saveall=' . md5($config) . ' > /dev/null 2>&1 ; rm -f ' . md5($config) . '.lock" > /dev/null 2>&1 &');
}
$finished_process = file_exists(getcwd() . '/cache/query/' . md5($config) . '-resolutions.csv');
if ($finished_process && !$is_cached) {
if (($handle = fopen(getcwd() . '/cache/query/' . md5($config) . '-resolutions.csv', 'r')) !== FALSE) {
$header = fgetcsv($handle, 1000, ",");
$token = 0;
$query = "REPLACE INTO resolutions (id_resolution,id_learner,id_exec,instance,model,sigma,outlier_code,predicted,observed) VALUES ";
while (($data = fgetcsv($handle, 1000, ",")) !== FALSE) {
$resolution = $data[0];
$pred_value = (int) $data[1] >= 100 ? (int) $data[1] : 100;
$exec_value = (int) $data[2];
$selected_instance_pre = preg_replace('/\\s+/', '', $data[3]);
$selected_instance_pre = str_replace(':', ',', $selected_instance_pre);
$specific_id = $data[4];
if ($token > 0) {
$query = $query . ",";
}
$token = 1;
$query = $query . "('" . md5($config) . "','" . $current_model . "','" . $specific_id . "','" . $selected_instance_pre . "','" . $model_info . "','" . $sigma_param . "','" . $resolution . "','" . $pred_value . "','" . $exec_value . "') ";
}
if ($dbml->query($query) === FALSE) {
throw new \Exception('Error when saving tree into DB');
}
}
// Store file model to DB
$filemodel = getcwd() . '/cache/query/' . md5($config) . '-object.rds';
$fp = fopen($filemodel, 'r');
$content = fread($fp, filesize($filemodel));
$content = addslashes($content);
fclose($fp);
$query = "INSERT INTO model_storage (id_hash,type,file) VALUES ('" . md5($config) . "','resolution','" . $content . "');";
if ($dbml->query($query) === FALSE) {
throw new \Exception('Error when saving file resolution into DB');
}
// Remove temporary files
$output = shell_exec('rm -f ' . getcwd() . '/cache/query/' . md5($config) . '-*.csv');
$output = shell_exec('rm -f ' . getcwd() . '/cache/query/' . $current_model . '-object.rds');
$output = shell_exec('rm -f ' . getcwd() . '/cache/query/' . md5($config) . '-object.rds');
$is_cached = true;
}
if (!$is_cached) {
$jsonData = $jsonOuts = $jsonWarns = $jsonHeader = $jsonTable = '[]';
$must_wait = 'YES';
if (isset($_GET['dump'])) {
echo "1";
exit(0);
}
} else {
$must_wait = 'NO';
$query = "SELECT predicted, observed, outlier_code, id_exec, instance FROM resolutions WHERE id_resolution = '" . md5($config) . "' LIMIT 5000";
// FIXME - CLUMSY PATCH FOR BYPASS THE BUG FROM HIGHCHARTS... REMEMBER TO ERASE THIS LINE WHEN THE BUG IS SOLVED
$result = $dbml->query($query);
foreach ($result as $row) {
$entry = array('x' => (int) $row['predicted'], 'y' => (int) $row['observed'], 'name' => $row['instance'], 'id' => (int) $row['id_exec']);
if ($row['outlier_code'] == 0) {
$jsonData[] = $entry;
}
if ($row['outlier_code'] == 1) {
$jsonWarns[] = $entry;
}
if ($row['outlier_code'] == 2) {
$jsonOuts[] = $entry;
}
$jsonTable .= ($jsonTable == '' ? '' : ',') . '["' . ($row['outlier_code'] == 0 ? 'Legitimate' : ($row['outlier_code'] == 1 ? 'Warning' : 'Outlier')) . '","' . $row['predicted'] . '","' . $row['observed'] . '","' . str_replace(",", "\",\"", $row['instance']) . '","' . $row['id_exec'] . '"]';
}
$query_var = "SELECT MAX(predicted) as max_x, MAX(observed) as max_y FROM resolutions WHERE id_resolution = '" . md5($config) . "' LIMIT 5000";
$result = $dbml->query($query_var);
$row = $result->fetch();
$max_x = $row['max_x'];
$max_y = $row['max_y'];
$header = array('Prediction', 'Observed', 'Benchmark', 'Net', 'Disk', 'Maps', 'IO.SFS', 'Rep', 'IO.FBuf', 'Comp', 'Blk.Size', 'Cluster', 'Cl.Name', 'Datanodes', 'Headnodes', 'VM.OS', 'VM.Cores', 'VM.RAM', 'Provider', 'VM.Size', 'Type', 'Bench.Type', 'ID');
$jsonHeader = '[{title:""}';
foreach ($header as $title) {
$jsonHeader = $jsonHeader . ',{title:"' . $title . '"}';
}
$jsonHeader = $jsonHeader . ']';
$jsonData = json_encode($jsonData);
$jsonWarns = json_encode($jsonWarns);
$jsonOuts = json_encode($jsonOuts);
$jsonTable = '[' . $jsonTable . ']';
// Dump case
if (isset($_GET['dump'])) {
echo str_replace(array("[", "]", "{title:\"", "\"}"), array('', '', ''), $jsonHeader) . "\n";
echo str_replace(array('],[', '[[', ']]'), array("\n", '', ''), $jsonOuts);
echo str_replace(array('],[', '[[', ']]'), array("\n", '', ''), $jsonWarns);
echo str_replace(array('],[', '[[', ']]'), array("\n", '', ''), $jsonData);
exit(0);
}
// Register case
if (isset($_GET['register'])) {
// Update the predictions table
$query_var = "UPDATE predictions as p, resolutions as r\n\t\t\t\t\t\t\t\tSET p.outlier = r.outlier_code\n\t\t\t\t\t\t\t\tWHERE r.id_exec = p.id_exec\n\t\t\t\t\t\t\t\t\tAND r.id_resolution = '" . md5($config) . "'\n\t\t\t\t\t\t\t\t\tAND p.id_learner = '" . $current_model . "'";
if ($dbml->query($query_var) === FALSE) {
throw new \Exception('Error when updating predictions in DB');
}
}
}
} else {
$message = "There are no prediction models trained for such parameters. Train at least one model in 'ML Prediction' section.";
$must_wait = "NO";
}
$dbml = null;
} catch (\Exception $e) {
$this->container->getTwig()->addGlobal('message', $e->getMessage() . "\n");
$jsonData = $jsonOuts = $jsonWarns = $jsonHeader = $jsonTable = '[]';
$model = '';
$possible_models_id = $possible_models = array();
$dbml = null;
}
echo $this->container->getTwig()->render('mltemplate/mloutliers.html.twig', array('selected' => 'mloutliers', 'jsonData' => $jsonData, 'jsonWarns' => $jsonWarns, 'jsonOuts' => $jsonOuts, 'jsonHeader' => $jsonHeader, 'jsonTable' => $jsonTable, 'max_p' => min(array($max_x, $max_y)), 'benchs' => $params['benchs'], 'nets' => $params['nets'], 'disks' => $params['disks'], 'blk_sizes' => $params['blk_sizes'], 'comps' => $params['comps'], 'id_clusters' => $params['id_clusters'], 'mapss' => $params['mapss'], 'replications' => $params['replications'], 'iosfs' => $params['iosfs'], 'iofilebufs' => $params['iofilebufs'], 'datanodess' => $params['datanodess'], 'bench_types' => $params['bench_types'], 'vm_sizes' => $params['vm_sizes'], 'vm_coress' => $params['vm_coress'], 'vm_RAMs' => $params['vm_RAMs'], 'types' => $params['types'], 'must_wait' => $must_wait, 'models' => '<li>' . implode('</li><li>', $possible_models) . '</li>', 'models_id' => $possible_models_id, 'current_model' => $current_model, 'resolution_id' => md5($config), 'sigma' => $sigma_param, 'message' => $message, 'instance' => $instance, 'preset' => $preset, 'selPreset' => $selPreset, 'options' => Utils::getFilterOptions($db)));
}
