1
2
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
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
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
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
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
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
|
#include "BenchmarkUtils.hpp"
#include "LoggingManager.hpp"
#include <iostream>
#include <fstream>
#include <thread>
#include <chrono>
#include <vector>
#include <future>
#include <optional>
#include <iomanip>
#include <filesystem>
struct BenchmarkResult
{
double executionTime;
double throughputEntries;
double logicalThroughputGiB;
double physicalThroughputGiB;
size_t inputDataSizeBytes;
size_t outputDataSizeBytes;
double writeAmplification;
LatencyStats latencyStats;
};
BenchmarkResult runBenchmark(const LoggingConfig &baseConfig, int numWriterThreads, int numProducerThreads,
int entriesPerProducer, int numSpecificFiles, int producerBatchSize, int payloadSize)
{
LoggingConfig config = baseConfig;
config.basePath = "./logs_writers";
config.numWriterThreads = numWriterThreads;
cleanupLogDirectory(config.basePath);
std::cout << "Generating batches with pre-determined destinations for all threads...";
std::vector<BatchWithDestination> batches = generateBatches(entriesPerProducer, numSpecificFiles, producerBatchSize, payloadSize);
std::cout << " Done." << std::endl;
size_t totalDataSizeBytes = calculateTotalDataSize(batches, numProducerThreads);
double totalDataSizeGiB = static_cast<double>(totalDataSizeBytes) / (1024 * 1024 * 1024);
std::cout << "Total data to be written: " << totalDataSizeBytes << " bytes ("
<< totalDataSizeGiB << " GiB)" << std::endl;
LoggingManager loggingManager(config);
loggingManager.start();
auto startTime = std::chrono::high_resolution_clock::now();
// Each future now returns a LatencyCollector with thread-local measurements
std::vector<std::future<LatencyCollector>> futures;
for (int i = 0; i < numProducerThreads; i++)
{
futures.push_back(std::async(
std::launch::async,
appendLogEntries,
std::ref(loggingManager),
std::ref(batches)));
}
// Collect latency measurements from all threads
LatencyCollector masterCollector;
for (auto &future : futures)
{
LatencyCollector threadCollector = future.get();
masterCollector.merge(threadCollector);
}
loggingManager.stop();
auto endTime = std::chrono::high_resolution_clock::now();
std::chrono::duration<double> elapsed = endTime - startTime;
size_t finalStorageSize = calculateDirectorySize(config.basePath);
double writeAmplification = static_cast<double>(finalStorageSize) / totalDataSizeBytes;
double elapsedSeconds = elapsed.count();
const size_t totalEntries = numProducerThreads * entriesPerProducer;
double throughputEntries = totalEntries / elapsedSeconds;
double logicalThroughputGiB = totalDataSizeGiB / elapsedSeconds;
double physicalThroughputGiB = static_cast<double>(finalStorageSize) / (1024.0 * 1024.0 * 1024.0 * elapsedSeconds);
// Calculate latency statistics from merged measurements
LatencyStats latencyStats = calculateLatencyStats(masterCollector);
cleanupLogDirectory(config.basePath);
return BenchmarkResult{
elapsedSeconds,
throughputEntries,
logicalThroughputGiB,
physicalThroughputGiB,
totalDataSizeBytes,
finalStorageSize,
writeAmplification,
latencyStats};
}
// Write CSV header
void writeCSVHeader(std::ofstream &csvFile)
{
csvFile << "writer_threads,execution_time_seconds,throughput_entries_per_sec,logical_throughput_gib_per_sec,"
<< "physical_throughput_gib_per_sec,input_data_size_bytes,output_data_size_bytes,speedup_vs_baseline,"
<< "write_amplification,avg_latency_ms,median_latency_ms,max_latency_ms,latency_count\n";
}
// Write a single result row to CSV
void writeCSVRow(std::ofstream &csvFile, int writerThreads, const BenchmarkResult &result, double speedup)
{
csvFile << writerThreads << ","
<< std::fixed << std::setprecision(6) << result.executionTime << ","
<< std::fixed << std::setprecision(2) << result.throughputEntries << ","
<< std::fixed << std::setprecision(6) << result.logicalThroughputGiB << ","
<< std::fixed << std::setprecision(6) << result.physicalThroughputGiB << ","
<< result.inputDataSizeBytes << ","
<< result.outputDataSizeBytes << ","
<< std::fixed << std::setprecision(6) << speedup << ","
<< std::fixed << std::setprecision(8) << result.writeAmplification << ","
<< std::fixed << std::setprecision(6) << result.latencyStats.avgMs << ","
<< std::fixed << std::setprecision(6) << result.latencyStats.medianMs << ","
<< std::fixed << std::setprecision(6) << result.latencyStats.maxMs << ","
<< result.latencyStats.count << "\n";
}
void runConcurrencyBenchmark(const LoggingConfig &baseConfig, const std::vector<int> &writerThreadCounts,
int numProducerThreads, int entriesPerProducer,
int numSpecificFiles, int producerBatchSize, int payloadSize,
const std::string &csvFilename = "concurrency_benchmark.csv")
{
std::vector<BenchmarkResult> results;
// Open CSV file for writing
std::ofstream csvFile(csvFilename);
if (!csvFile.is_open())
{
std::cerr << "Error: Could not open CSV file " << csvFilename << " for writing." << std::endl;
return;
}
writeCSVHeader(csvFile);
std::cout << "Running concurrency benchmark with " << writerThreadCounts.size() << " data points..." << std::endl;
std::cout << "Results will be saved to: " << csvFilename << std::endl;
for (size_t i = 0; i < writerThreadCounts.size(); i++)
{
int writerCount = writerThreadCounts[i];
std::cout << "\nProgress: " << (i + 1) << "/" << writerThreadCounts.size()
<< " - Running benchmark with " << writerCount << " writer thread(s)..." << std::endl;
BenchmarkResult result = runBenchmark(baseConfig, writerCount, numProducerThreads, entriesPerProducer,
numSpecificFiles, producerBatchSize, payloadSize);
results.push_back(result);
// Calculate speedup (using first result as baseline)
double speedup = results.size() > 1 ? result.throughputEntries / results[0].throughputEntries : 1.0;
// Write result to CSV immediately
writeCSVRow(csvFile, writerCount, result, speedup);
csvFile.flush(); // Ensure data is written in case of early termination
// Print progress summary
std::cout << " Completed: " << std::fixed << std::setprecision(2)
<< result.throughputEntries << " entries/s, "
<< std::fixed << std::setprecision(3) << result.logicalThroughputGiB << " GiB/s, "
<< std::fixed << std::setprecision(2) << speedup << "x speedup" << std::endl;
}
csvFile.close();
std::cout << "\nBenchmark completed! Results saved to " << csvFilename << std::endl;
// Still print summary table to console for immediate review
std::cout << "\n=================== CONCURRENCY BENCHMARK SUMMARY ===================" << std::endl;
std::cout << std::left << std::setw(20) << "Writer Threads"
<< std::setw(15) << "Time (sec)"
<< std::setw(20) << "Throughput (ent/s)"
<< std::setw(15) << "Logical (GiB/s)"
<< std::setw(15) << "Physical (GiB/s)"
<< std::setw(20) << "Input Size (bytes)"
<< std::setw(20) << "Storage Size (bytes)"
<< std::setw(15) << "Write Amp."
<< std::setw(12) << "Speedup vs. 1"
<< std::setw(12) << "Avg Lat(ms)" << std::endl;
std::cout << "--------------------------------------------------------------------------------------------------------------------------------" << std::endl;
double baselineThroughputEntries = results[0].throughputEntries;
for (size_t i = 0; i < writerThreadCounts.size(); i++)
{
double speedup = results[i].throughputEntries / baselineThroughputEntries;
std::cout << std::left << std::setw(20) << writerThreadCounts[i]
<< std::setw(15) << std::fixed << std::setprecision(2) << results[i].executionTime
<< std::setw(20) << std::fixed << std::setprecision(2) << results[i].throughputEntries
<< std::setw(15) << std::fixed << std::setprecision(3) << results[i].logicalThroughputGiB
<< std::setw(15) << std::fixed << std::setprecision(3) << results[i].physicalThroughputGiB
<< std::setw(20) << results[i].inputDataSizeBytes
<< std::setw(20) << results[i].outputDataSizeBytes
<< std::setw(15) << std::fixed << std::setprecision(4) << results[i].writeAmplification
<< std::setw(12) << std::fixed << std::setprecision(2) << speedup
<< std::setw(12) << std::fixed << std::setprecision(3) << results[i].latencyStats.avgMs << std::endl;
}
std::cout << "================================================================================================================================" << std::endl;
}
int main()
{
// system parameters
LoggingConfig baseConfig;
baseConfig.baseFilename = "default";
baseConfig.maxSegmentSize = 50 * 1024 * 1024; // 50 MB
baseConfig.maxAttempts = 5;
baseConfig.baseRetryDelay = std::chrono::milliseconds(1);
baseConfig.queueCapacity = 3000000;
baseConfig.maxExplicitProducers = 16;
baseConfig.batchSize = 8192;
baseConfig.appendTimeout = std::chrono::minutes(5);
baseConfig.useEncryption = true;
baseConfig.compressionLevel = 9;
// benchmark parameters
const int numSpecificFiles = 256;
const int producerBatchSize = 512;
const int numProducers = 16;
const int entriesPerProducer = 2000000;
const int payloadSize = 2048;
std::vector<int> writerThreadCounts = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16};
runConcurrencyBenchmark(baseConfig,
writerThreadCounts,
numProducers,
entriesPerProducer,
numSpecificFiles,
producerBatchSize,
payloadSize,
"concurrency_benchmark_results.csv");
return 0;
}
|