PerSystSqlOperator.cpp 11 KB
Newer Older
1
//================================================================================
2
// Name        : PerSystSqlOperator.cpp
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
// Author      : Carla Guillen
// Contact     : info@dcdb.it
// Copyright   : Leibniz Supercomputing Centre
// Description : Template implementing features to use Units in Operators.
//================================================================================

//================================================================================
// This file is part of DCDB (DataCenter DataBase)
// Copyright (C) 2018-2019 Leibniz Supercomputing Centre
//
// This program is free software; you can redistribute it and/or
// modify it under the terms of the GNU General Public License
// as published by the Free Software Foundation; either version 2
// of the License, or (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
// GNU General Public License for more details.
//
// You should have received a copy of the GNU General Public License
// along with this program; if not, write to the Free Software
// Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
//================================================================================

28
29
#include "PerSystSqlOperator.h"

30
31
32
33
34
#include <boost/log/sources/record_ostream.hpp>
#include <boost/log/trivial.hpp>
#include <boost/log/utility/formatting_ostream.hpp>
#include <boost/parameter/keyword.hpp>
#include <stddef.h>
35
#include <cmath>
36
#include <cstdint>
37
#include <memory>
38
#include <string>
39
#include <numeric>
Carla Guillen's avatar
Carla Guillen committed
40
#include <sstream>
41
42

#include "../../../common/include/logging.h"
43
#include "../../../common/include/sensorbase.h"
44
45
46
#include "../../../common/include/timestamp.h"
#include "../../includes/CommonStatistics.h"
#include "../../includes/QueryEngine.h"
47
#include "../../includes/UnitTemplate.h"
48

Carla Guillen's avatar
Carla Guillen committed
49
int PerSystSqlOperator::_number_of_calls = 0;
Carla Guillen's avatar
Carla Guillen committed
50
PerSystDB PerSystSqlOperator::persystdb;
Carla Guillen's avatar
Carla Guillen committed
51

52
PerSystSqlOperator::PerSystSqlOperator(const std::string& name) :
Carla Guillen's avatar
Carla Guillen committed
53
54
55
		OperatorTemplate(name), JobOperatorTemplate(name), _number_of_even_quantiles(
				0), _severity_formula(NOFORMULA), _severity_threshold(0), _severity_exponent(
				0), _severity_max_memory(0), _go_back_ns(0), _backend(DEFAULT), _scaling_factor(
Carla Guillen's avatar
Carla Guillen committed
56
				1), _property_id(0) {
57
58
}

59
PerSystSqlOperator::~PerSystSqlOperator() {
60
61
}

62
void PerSystSqlOperator::compute(U_Ptr unit, qeJobData& jobData) {
Carla Guillen's avatar
Carla Guillen committed
63
	// Clearing the buffer, if already allocated
64
	_buffer.clear();
Carla Guillen's avatar
Carla Guillen committed
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
	size_t elCtr = 0;
	uint64_t my_timestamp = getTimestamp() - _go_back_ns;
	// Making sure that the aggregation boundaries do not go past the job start/end time
	uint64_t jobEnd =
			jobData.endTime != 0 && my_timestamp > jobData.endTime ?
					jobData.endTime : my_timestamp;
	uint64_t jobStart =
			jobEnd - my_timestamp < jobData.startTime ?
					jobData.startTime : jobEnd - my_timestamp;
	// Job units are hierarchical, and thus we iterate over all sub-units associated to each single node
	for (const auto& subUnit : unit->getSubUnits()) {
		// Getting the most recent values as specified in _window
		// Since we do not clear the internal buffer, all sensor readings will be accumulated in the same vector
		for (const auto& in : subUnit->getInputs()) {
			if (!_queryEngine.querySensor(in->getName(), my_timestamp,
					my_timestamp, _buffer, false)) {
				LOG(debug)<< "PerSystSql Operator " << _name << " cannot read from sensor " << in->getName() << "!";
				return;
			}
		}
	}
	static bool persystdb_initialized = false;
  if(_backend == MARIADB){
	if (!persystdb_initialized) {
		LOG(debug)<< "PerSystSQL Operator Connection information:";
		LOG(debug) << "\tHost=" << _conn.host;
		LOG(debug) << "\tUser=" << _conn.user;
		LOG(debug) << "\tDatabase=" << _conn.database_name;
		LOG(debug) << "\tPort=" << _conn.port;
		LOG(debug) << "\tRotation=" << _conn.rotation;
		LOG(debug) << "\tEvery_X_days=" << _conn.every_x_days;
		bool persystdb_initialized = persystdb.initializeConnection(_conn.host, _conn.user, _conn.password, _conn.database_name, _conn.rotation, _conn.port, _conn.every_x_days);
		if(!persystdb_initialized) {
			LOG(error) << "Unable to establish connection to database";
			return;
		}
	}
  }
	Aggregate_info_t agg_info;
	std::string table_suffix;
  if(_backend == MARIADB){
	std::stringstream jobidBuilder;
	jobidBuilder << jobData.jobId;
108

Carla Guillen's avatar
Carla Guillen committed
109
110
	std::vector<std::string> job_ids;
	job_ids.push_back(jobidBuilder.str());
111

Carla Guillen's avatar
Carla Guillen committed
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
	std::map<std::string, std::string> job_map;
    	if(!persystdb.getTableSuffix(table_suffix)){
        	LOG(error) << "failed to create table!";
        	return;
    	}
    	if(!persystdb.getDBJobIDs(job_ids, job_map)){
       		return;
    	}

    	// handle jobs which are not present
   	 for(auto &job_id_string : job_ids ){
       	 	auto search = job_map.find(job_id_string);
        	if(search == job_map.end()){ //Not found
                	int job_id_db;
                	if(persystdb.insertIntoJob(job_id_string, jobData.userId, job_id_db, table_suffix)){
				agg_info.job_id_db = std::to_string(job_id_db); 
128
                	} else {
Carla Guillen's avatar
Carla Guillen committed
129
                        	continue;
130
                	}
Carla Guillen's avatar
Carla Guillen committed
131
132
133
134
        	}
   	}
	agg_info.timestamp = (my_timestamp/1e9);
  }
135

Carla Guillen's avatar
Carla Guillen committed
136
137
138
139
140
141
142
143
144
145
	compute_internal(unit, _buffer, agg_info);

  if(_backend == MARIADB){
	persystdb.insertInAggregateTable(table_suffix, agg_info);
	if(_number_of_calls % 10 == 0  && persystdb_initialized){
		persystdb.finalizeConnection();
		persystdb_initialized = false;
	}
	_number_of_calls++;
  }
146
147
}

Carla Guillen's avatar
Carla Guillen committed
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
void PerSystSqlOperator::compute_internal(U_Ptr& unit,
		vector<reading_t>& buffer, Aggregate_info_t & agg_info) {
	_quantileSensors.clear();

	reading_t reading;
	AggregatorSensorBase::aggregationOps_t op;
	reading.timestamp = getTimestamp() - _go_back_ns;

	std::vector<double> douBuffer;
	punToDoubles(buffer, douBuffer);
	// Performing the actual aggregation operation
	for (const auto& out : unit->getOutputs()) {
		op = out->getOperation();
		if (op != AggregatorSensorBase::QTL) {
			switch (op) {
			case AggregatorSensorBase::AVG:
				if (_backend == CASSANDRA) {
					reading.value = std::accumulate(douBuffer.begin(),
							douBuffer.end(), 0.0) / douBuffer.size()
							* _scaling_factor;
				} else {
					agg_info.average = std::accumulate(douBuffer.begin(), douBuffer.end(), 0.0) / douBuffer.size();
				}
				break;
			case AggregatorSensorBase::OBS:
				reading.value = computeObs(buffer);
				agg_info.num_of_observations = computeObs(buffer);
				break;
			case AggregatorSensorBase::AVG_SEV:
				if (_backend == CASSANDRA) {
					reading.value = computeSeverityAverage(douBuffer) * _scaling_factor;
				} else {
					agg_info.severity_average = computeSeverityAverage(douBuffer);
				}
				break;
			default:
				LOG(warning)<< _name << ": Operation " << op << " not supported!";
				reading.value = 0;
				break;
			}
			if(_backend == CASSANDRA) {
				out->storeReading(reading);
			}
		} else {
			_quantileSensors.push_back(out);
		}
	}

	if (!_quantileSensors.empty()) {
		vector<double> quantiles;
		computeEvenQuantiles(douBuffer, _number_of_even_quantiles, quantiles);
		if (_backend == CASSANDRA) {
			for (unsigned idx = 0; idx < quantiles.size(); idx++) {
				reading.value = quantiles[idx]*_scaling_factor;
				_quantileSensors[idx]->storeReading(reading);
			}
		} else {
			for(auto q: quantiles){
				agg_info.quantiles.push_back(static_cast<float>(q));
			}
		}
	}
	agg_info.property_type_id = _property_id;

}

void PerSystSqlOperator::compute(U_Ptr unit) {
215
216
//nothing here!
}
Carla Guillen Carias's avatar
Carla Guillen Carias committed
217

Carla Guillen's avatar
Carla Guillen committed
218
double severity_formula1(double metric, double threshold, double exponent) {
219
	double val = metric - threshold;
Carla Guillen Carias's avatar
Carla Guillen Carias committed
220
	if (val > 0) {
221
		double ret = (pow(val, exponent));
Carla Guillen's avatar
Carla Guillen committed
222
		if (ret > 1) {
Carla Guillen Carias's avatar
Carla Guillen Carias committed
223
224
225
226
227
228
229
			return 1;
		}
		return ret;
	}
	return 0;
}

Carla Guillen's avatar
Carla Guillen committed
230
231
double severity_formula2(double metric, double threshold, double exponent) {
	if (!threshold) {
Carla Guillen Carias's avatar
Carla Guillen Carias committed
232
233
		return -1;
	}
234
	double val = metric / threshold - 1;
Carla Guillen Carias's avatar
Carla Guillen Carias committed
235
	if (val > 0) {
Carla Guillen's avatar
Carla Guillen committed
236
237
		double ret = (pow(val, exponent));
		if (ret > 1) {
Carla Guillen Carias's avatar
Carla Guillen Carias committed
238
239
240
241
242
243
244
			return 1;
		}
		return ret;
	}
	return 0;
}

Carla Guillen's avatar
Carla Guillen committed
245
double severity_formula3(double metric, double threshold, double exponent) {
Carla Guillen Carias's avatar
Carla Guillen Carias committed
246
247
248
	if (!threshold) {
		return -1;
	}
249
	double val = metric / threshold;
Carla Guillen Carias's avatar
Carla Guillen Carias committed
250
	if (val > 0) {
Carla Guillen's avatar
Carla Guillen committed
251
252
		double ret = (1 - pow(val, exponent));
		if (ret > 1) {
Carla Guillen Carias's avatar
Carla Guillen Carias committed
253
254
			return 1;
		}
Carla Guillen's avatar
Carla Guillen committed
255
		if (ret < 0) {
Carla Guillen Carias's avatar
Carla Guillen Carias committed
256
257
258
259
260
261
262
			return 0;
		}
		return ret;
	}
	return 0;
}

Carla Guillen's avatar
Carla Guillen committed
263
double severity_memory(double metric, double threshold, double max_memory) {
264
265
	double denominator = max_memory - threshold;
	double severity = -1;
Carla Guillen's avatar
Carla Guillen committed
266
267
268
	if (denominator) {
		severity = metric - threshold / (max_memory - threshold);
		if (severity > 1) {
Carla Guillen Carias's avatar
Carla Guillen Carias committed
269
			severity = 1;
Carla Guillen's avatar
Carla Guillen committed
270
		} else if (severity < 0) {
Carla Guillen Carias's avatar
Carla Guillen Carias committed
271
272
273
274
275
			severity = 0;
		}
	}
	return severity;
}
276

Carla Guillen's avatar
Carla Guillen committed
277
278
double PerSystSqlOperator::computeSeverityAverage(
		std::vector<double> & buffer) {
279
	std::vector<double> severities;
Carla Guillen's avatar
Carla Guillen committed
280
281
282
283
284
285
286
	switch (_severity_formula) {
	case (FORMULA1):
		for (auto val : buffer) {
			auto severity = severity_formula1(val, _severity_threshold,
					_severity_exponent);
			severities.push_back(severity);
		}
287
		break;
Carla Guillen's avatar
Carla Guillen committed
288
289
290
291
292
293
	case (FORMULA2):
		for (auto val : buffer) {
			auto severity = severity_formula2(val, _severity_threshold,
					_severity_exponent);
			severities.push_back(severity);
		}
294
		break;
Carla Guillen's avatar
Carla Guillen committed
295
296
297
298
299
300
	case (FORMULA3):
		for (auto val : buffer) {
			auto severity = severity_formula3(val, _severity_threshold,
					_severity_exponent);
			severities.push_back(severity);
		}
301
		break;
Carla Guillen's avatar
Carla Guillen committed
302
303
304
305
306
307
	case (MEMORY_FORMULA):
		for (auto val : buffer) {
			auto severity = severity_memory(val, _severity_threshold,
					_severity_max_memory);
			severities.push_back(severity);
		}
308
		break;
Carla Guillen's avatar
Carla Guillen committed
309
310
311
312
313
314
315
	case (NOFORMULA):
		for (auto val : buffer) {
			severities.push_back(severity_noformula());
		}
		break;
	default:
		return 0.0;
Carla Guillen Carias's avatar
Carla Guillen Carias committed
316
		break;
317
	}
Carla Guillen's avatar
Carla Guillen committed
318
319
320
	if (severities.size()) {
		return (std::accumulate(severities.begin(), severities.end(), 0.0)
				/ severities.size());
321
322
	}
	return 0.0;
323
}
324

Carla Guillen's avatar
Carla Guillen committed
325
326
327
void punToDoubles(std::vector<reading_t> & buffer,
		std::vector<double> & outDoubleVec) {
	for (auto & reading : buffer) {
328
329
330
331
		outDoubleVec.push_back(punLLToDouble(reading.value));
	}
}

Carla Guillen's avatar
Carla Guillen committed
332
333
334
335
double punLLToDouble(long long value) {
	double * returnval;
	returnval = (double *) (&value);
	return *returnval;
336
337
}

Carla Guillen's avatar
Carla Guillen committed
338
339
340
long long punDoubleToLL(double value) {
	long long * returnval;
	returnval = (long long *) (&value);
341

Carla Guillen's avatar
Carla Guillen committed
342
	return *returnval;
343
344
}

Carla Guillen's avatar
Carla Guillen committed
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
void computeEvenQuantiles(std::vector<double> &data,
		const unsigned int NUMBER_QUANTILES, std::vector<double> &quantiles) {
	if (data.empty() || NUMBER_QUANTILES == 0) {
		return;
	}
	std::sort(data.begin(), data.end());
	int elementNumber = data.size();
	quantiles.resize(NUMBER_QUANTILES + 1); //+min
	double factor = elementNumber / static_cast<double>(NUMBER_QUANTILES);
	quantiles[0] = data[0]; //minimum
	quantiles[NUMBER_QUANTILES] = data[data.size() - 1]; //maximum
	for (unsigned int i = 1; i < NUMBER_QUANTILES; i++) {
		if (elementNumber > 1) {
			int idx = static_cast<int>(std::floor(i * factor));
			if (idx == 0) {
				quantiles[i] = data[0];
			} else {
				double rest = (i * factor) - idx;
				quantiles[i] = data[idx - 1]
						+ rest * (data[idx] - data[idx - 1]); //ToDo scaling factor??
			}
		} else { //optimization, we don't need to calculate all the quantiles
			quantiles[i] = data[0];
		}
	}
370
371
}