1 | #region License Information
|
---|
2 | /* HeuristicLab
|
---|
3 | * Copyright (C) 2002-2011 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
4 | *
|
---|
5 | * This file is part of HeuristicLab.
|
---|
6 | *
|
---|
7 | * HeuristicLab is free software: you can redistribute it and/or modify
|
---|
8 | * it under the terms of the GNU General Public License as published by
|
---|
9 | * the Free Software Foundation, either version 3 of the License, or
|
---|
10 | * (at your option) any later version.
|
---|
11 | *
|
---|
12 | * HeuristicLab is distributed in the hope that it will be useful,
|
---|
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
15 | * GNU General Public License for more details.
|
---|
16 | *
|
---|
17 | * You should have received a copy of the GNU General Public License
|
---|
18 | * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
|
---|
19 | */
|
---|
20 | #endregion
|
---|
21 |
|
---|
22 | using System;
|
---|
23 | using System.Collections.Generic;
|
---|
24 | using System.Linq;
|
---|
25 | using System.Threading;
|
---|
26 | using System.Threading.Tasks;
|
---|
27 | using HeuristicLab.Clients.Hive;
|
---|
28 | using HeuristicLab.Common;
|
---|
29 | using HeuristicLab.Core;
|
---|
30 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
31 |
|
---|
32 | namespace HeuristicLab.HiveEngine {
|
---|
33 | /// <summary>
|
---|
34 | /// Represents an engine that executes operations which can be executed in parallel on the hive
|
---|
35 | /// </summary>
|
---|
36 | [StorableClass]
|
---|
37 | [Item("Hive Engine", "Engine for parallel execution on the hive. You need enable `Parallel` for at least one operator in your operator graph to have all childoperations parallelized. Also those childoperations must not have sideeffects on a higher scope.")]
|
---|
38 | public class HiveEngine : Engine {
|
---|
39 | private static object locker = new object();
|
---|
40 | private static object logLocker = new object();
|
---|
41 | private CancellationToken cancellationToken;
|
---|
42 | private bool firstRun = true;
|
---|
43 |
|
---|
44 | [Storable]
|
---|
45 | private IOperator currentOperator;
|
---|
46 |
|
---|
47 | [Storable]
|
---|
48 | public string ResourceNames { get; set; }
|
---|
49 |
|
---|
50 | [Storable]
|
---|
51 | private int priority;
|
---|
52 | public int Priority {
|
---|
53 | get { return priority; }
|
---|
54 | set { priority = value; }
|
---|
55 | }
|
---|
56 |
|
---|
57 | [Storable]
|
---|
58 | private TimeSpan executionTimeOnHive;
|
---|
59 | public TimeSpan ExecutionTimeOnHive {
|
---|
60 | get { return executionTimeOnHive; }
|
---|
61 | set {
|
---|
62 | if (value != executionTimeOnHive) {
|
---|
63 | executionTimeOnHive = value;
|
---|
64 | OnExecutionTimeOnHiveChanged();
|
---|
65 | }
|
---|
66 | }
|
---|
67 | }
|
---|
68 |
|
---|
69 | [Storable]
|
---|
70 | private bool useLocalPlugins;
|
---|
71 | public bool UseLocalPlugins {
|
---|
72 | get { return useLocalPlugins; }
|
---|
73 | set { useLocalPlugins = value; }
|
---|
74 | }
|
---|
75 |
|
---|
76 | // [Storable] -> HiveExperiment can't be storable, so RefreshableHiveExperiment can't be stored
|
---|
77 | private ItemCollection<RefreshableHiveExperiment> hiveExperiments = new ItemCollection<RefreshableHiveExperiment>();
|
---|
78 | public ItemCollection<RefreshableHiveExperiment> HiveExperiments {
|
---|
79 | get { return hiveExperiments; }
|
---|
80 | set { hiveExperiments = value; }
|
---|
81 | }
|
---|
82 |
|
---|
83 | private List<Plugin> onlinePlugins;
|
---|
84 | public List<Plugin> OnlinePlugins {
|
---|
85 | get { return onlinePlugins; }
|
---|
86 | set { onlinePlugins = value; }
|
---|
87 | }
|
---|
88 |
|
---|
89 | private List<Plugin> alreadyUploadedPlugins;
|
---|
90 | public List<Plugin> AlreadyUploadedPlugins {
|
---|
91 | get { return alreadyUploadedPlugins; }
|
---|
92 | set { alreadyUploadedPlugins = value; }
|
---|
93 | }
|
---|
94 |
|
---|
95 | #region constructors and cloning
|
---|
96 | public HiveEngine() {
|
---|
97 | ResourceNames = "HEAL";
|
---|
98 | Priority = 0;
|
---|
99 | }
|
---|
100 |
|
---|
101 | [StorableConstructor]
|
---|
102 | protected HiveEngine(bool deserializing) : base(deserializing) { }
|
---|
103 | protected HiveEngine(HiveEngine original, Cloner cloner)
|
---|
104 | : base(original, cloner) {
|
---|
105 | this.ResourceNames = original.ResourceNames;
|
---|
106 | this.currentOperator = cloner.Clone(original.currentOperator);
|
---|
107 | this.priority = original.priority;
|
---|
108 | this.executionTimeOnHive = original.executionTimeOnHive;
|
---|
109 | this.useLocalPlugins = original.useLocalPlugins;
|
---|
110 | // this.hiveExperiments = cloner.Clone(original.hiveExperiments); do not clone hiveExperiments - otherwise they would be sent with every job
|
---|
111 | }
|
---|
112 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
113 | return new HiveEngine(this, cloner);
|
---|
114 | }
|
---|
115 | #endregion
|
---|
116 |
|
---|
117 | #region Events
|
---|
118 | protected override void OnPrepared() {
|
---|
119 | base.OnPrepared();
|
---|
120 | this.ExecutionTimeOnHive = TimeSpan.Zero;
|
---|
121 | }
|
---|
122 |
|
---|
123 | public event EventHandler ExecutionTimeOnHiveChanged;
|
---|
124 | protected virtual void OnExecutionTimeOnHiveChanged() {
|
---|
125 | var handler = ExecutionTimeOnHiveChanged;
|
---|
126 | if (handler != null) handler(this, EventArgs.Empty);
|
---|
127 | }
|
---|
128 | #endregion
|
---|
129 |
|
---|
130 | protected override void Run(CancellationToken cancellationToken) {
|
---|
131 | this.cancellationToken = cancellationToken;
|
---|
132 | Run(ExecutionStack);
|
---|
133 | }
|
---|
134 |
|
---|
135 | private void Run(object state) {
|
---|
136 | Stack<IOperation> executionStack = (Stack<IOperation>)state;
|
---|
137 | IOperation next;
|
---|
138 | OperationCollection coll;
|
---|
139 | IAtomicOperation operation;
|
---|
140 |
|
---|
141 | if (firstRun) {
|
---|
142 | TaskScheduler.UnobservedTaskException += new EventHandler<UnobservedTaskExceptionEventArgs>(TaskScheduler_UnobservedTaskException);
|
---|
143 | this.OnlinePlugins = ServiceLocator.Instance.CallHiveService(s => s.GetPlugins()).Where(x => x.IsLocal == false).ToList();
|
---|
144 | this.AlreadyUploadedPlugins = new List<Plugin>();
|
---|
145 | firstRun = false;
|
---|
146 | }
|
---|
147 |
|
---|
148 | while (executionStack.Count > 0) {
|
---|
149 | cancellationToken.ThrowIfCancellationRequested();
|
---|
150 |
|
---|
151 | next = executionStack.Pop();
|
---|
152 | bool isOpCollection = next is OperationCollection;
|
---|
153 | int collCount = isOpCollection ? ((OperationCollection)next).Count : 0;
|
---|
154 | string opName = !isOpCollection ? ((IAtomicOperation)next).Operator.Name : "OpCollection";
|
---|
155 |
|
---|
156 | if (next is OperationCollection) {
|
---|
157 | coll = (OperationCollection)next;
|
---|
158 |
|
---|
159 | bool isPMOEvaluator = coll.Count > 0 && coll.First() is HeuristicLab.Core.ExecutionContext && ((HeuristicLab.Core.ExecutionContext)coll.First()).Operator.GetType().Name == "PMOEvaluator";
|
---|
160 | bool isAlgorithmEvaluator = coll.Count > 0 && coll.First() is HeuristicLab.Core.ExecutionContext && ((HeuristicLab.Core.ExecutionContext)coll.First()).Operator.GetType().Name == "AlgorithmEvaluator";
|
---|
161 |
|
---|
162 | if (coll.Parallel && isPMOEvaluator) {
|
---|
163 | Task[] tasks = new Task[coll.Count];
|
---|
164 | Stack<IOperation>[] stacks = new Stack<IOperation>[coll.Count];
|
---|
165 | for (int i = 0; i < coll.Count; i++) {
|
---|
166 | stacks[i] = new Stack<IOperation>();
|
---|
167 | stacks[i].Push(coll[i]);
|
---|
168 | tasks[i] = Task.Factory.StartNew(Run, stacks[i], cancellationToken);
|
---|
169 | }
|
---|
170 | try {
|
---|
171 | Task.WaitAll(tasks);
|
---|
172 | }
|
---|
173 | catch (AggregateException ex) {
|
---|
174 | OperationCollection remaining = new OperationCollection() { Parallel = true };
|
---|
175 | for (int i = 0; i < stacks.Length; i++) {
|
---|
176 | if (stacks[i].Count == 1)
|
---|
177 | remaining.Add(stacks[i].Pop());
|
---|
178 | if (stacks[i].Count > 1) {
|
---|
179 | OperationCollection ops = new OperationCollection();
|
---|
180 | while (stacks[i].Count > 0)
|
---|
181 | ops.Add(stacks[i].Pop());
|
---|
182 | remaining.Add(ops);
|
---|
183 | }
|
---|
184 | }
|
---|
185 | if (remaining.Count > 0) executionStack.Push(remaining);
|
---|
186 | throw ex;
|
---|
187 | }
|
---|
188 | } else if (coll.Parallel) {
|
---|
189 | // clone the parent scope here and reuse it for each operation. otherwise for each job the whole scope-tree first needs to be copied and then cleaned, which causes a lot of work for the Garbage Collector
|
---|
190 | IScope parentScopeClone = (IScope)((IAtomicOperation)coll.First()).Scope.Parent.Clone();
|
---|
191 | parentScopeClone.SubScopes.Clear();
|
---|
192 | parentScopeClone.ClearParentScopes();
|
---|
193 |
|
---|
194 | EngineJob[] jobs = new EngineJob[coll.Count];
|
---|
195 | for (int i = 0; i < coll.Count; i++) {
|
---|
196 | jobs[i] = new EngineJob(coll[i], new SequentialEngine.SequentialEngine());
|
---|
197 | }
|
---|
198 |
|
---|
199 | var experiment = CreateHiveExperiment();
|
---|
200 | IScope[] scopes = ExecuteOnHive(experiment, jobs, parentScopeClone, cancellationToken);
|
---|
201 | DisposeHiveExperiment(experiment);
|
---|
202 |
|
---|
203 | for (int i = 0; i < coll.Count; i++) {
|
---|
204 | if (coll[i] is IAtomicOperation) {
|
---|
205 | ExchangeScope(scopes[i], ((IAtomicOperation)coll[i]).Scope);
|
---|
206 | } else if (coll[i] is OperationCollection) {
|
---|
207 | // todo ??
|
---|
208 | }
|
---|
209 | }
|
---|
210 | } else {
|
---|
211 | for (int i = coll.Count - 1; i >= 0; i--)
|
---|
212 | if (coll[i] != null) executionStack.Push(coll[i]);
|
---|
213 | }
|
---|
214 | } else if (next is IAtomicOperation) {
|
---|
215 | operation = (IAtomicOperation)next;
|
---|
216 | try {
|
---|
217 | next = operation.Operator.Execute((IExecutionContext)operation, cancellationToken);
|
---|
218 | }
|
---|
219 | catch (Exception ex) {
|
---|
220 | executionStack.Push(operation);
|
---|
221 | if (ex is OperationCanceledException) throw ex;
|
---|
222 | else throw new OperatorExecutionException(operation.Operator, ex);
|
---|
223 | }
|
---|
224 | if (next != null) executionStack.Push(next);
|
---|
225 |
|
---|
226 | if (operation.Operator.Breakpoint) {
|
---|
227 | LogMessage(string.Format("Breakpoint: {0}", operation.Operator.Name != string.Empty ? operation.Operator.Name : operation.Operator.ItemName));
|
---|
228 | Pause();
|
---|
229 | }
|
---|
230 | }
|
---|
231 | }
|
---|
232 | }
|
---|
233 |
|
---|
234 | private void TaskScheduler_UnobservedTaskException(object sender, UnobservedTaskExceptionEventArgs e) {
|
---|
235 | e.SetObserved(); // avoid crash of process
|
---|
236 | }
|
---|
237 |
|
---|
238 | private IRandom FindRandomParameter(IExecutionContext ec) {
|
---|
239 | try {
|
---|
240 | if (ec == null)
|
---|
241 | return null;
|
---|
242 |
|
---|
243 | foreach (var p in ec.Parameters) {
|
---|
244 | if (p.Name == "Random" && p is IValueParameter)
|
---|
245 | return ((IValueParameter)p).Value as IRandom;
|
---|
246 | }
|
---|
247 | return FindRandomParameter(ec.Parent);
|
---|
248 | }
|
---|
249 | catch { return null; }
|
---|
250 | }
|
---|
251 |
|
---|
252 | private static void ReIntegrateScope(IAtomicOperation source, IAtomicOperation target) {
|
---|
253 | ExchangeScope(source.Scope, target.Scope);
|
---|
254 | }
|
---|
255 |
|
---|
256 | private static void ExchangeScope(IScope source, IScope target) {
|
---|
257 | target.Variables.Clear();
|
---|
258 | target.Variables.AddRange(source.Variables);
|
---|
259 | target.SubScopes.Clear();
|
---|
260 | target.SubScopes.AddRange(source.SubScopes);
|
---|
261 | // TODO: validate if parent scopes match - otherwise source is invalid
|
---|
262 | }
|
---|
263 |
|
---|
264 | /// <summary>
|
---|
265 | /// This method blocks until all jobs are finished
|
---|
266 | /// TODO: Cancelation needs to be refined; all tasks currently stay in Semaphore.WaitOne after cancelation
|
---|
267 | /// </summary>
|
---|
268 | /// <param name="jobs"></param>
|
---|
269 | private IScope[] ExecuteOnHive(RefreshableHiveExperiment refreshableHiveExperiment, EngineJob[] jobs, IScope parentScopeClone, CancellationToken cancellationToken) {
|
---|
270 | LogMessage(string.Format("Executing {0} operations on the hive.", jobs.Length));
|
---|
271 | IScope[] scopes = new Scope[jobs.Length];
|
---|
272 | object locker = new object();
|
---|
273 | IDictionary<Guid, int> jobIndices = new Dictionary<Guid, int>();
|
---|
274 | var hiveExperiment = refreshableHiveExperiment.HiveExperiment;
|
---|
275 |
|
---|
276 | try {
|
---|
277 | List<Guid> remainingJobIds = new List<Guid>();
|
---|
278 |
|
---|
279 | // create upload-tasks
|
---|
280 | var uploadTasks = new List<Task<Job>>();
|
---|
281 | for (int i = 0; i < jobs.Length; i++) {
|
---|
282 | hiveExperiment.HiveJobs.Add(new EngineHiveJob(jobs[i], parentScopeClone));
|
---|
283 |
|
---|
284 | // shuffle random variable to avoid the same random sequence in each operation; todo: does not yet work (it cannot find the random variable)
|
---|
285 | IRandom random = FindRandomParameter(jobs[i].InitialOperation as IExecutionContext);
|
---|
286 | if (random != null)
|
---|
287 | random.Reset(random.Next());
|
---|
288 | }
|
---|
289 | ExperimentManagerClient.StartExperiment((e) => {
|
---|
290 | LogException(e);
|
---|
291 | }, refreshableHiveExperiment);
|
---|
292 |
|
---|
293 | // do polling until experiment is finished and all jobs are downloaded
|
---|
294 | while (!refreshableHiveExperiment.AllJobsFinished()) {
|
---|
295 | Thread.Sleep(500);
|
---|
296 | this.ExecutionTimeOnHive = TimeSpan.FromMilliseconds(hiveExperiments.Sum(x => x.HiveExperiment.ExecutionTime.TotalMilliseconds));
|
---|
297 | cancellationToken.ThrowIfCancellationRequested();
|
---|
298 | }
|
---|
299 | LogMessage(string.Format("{0} finished (TotalExecutionTime: {1}).", refreshableHiveExperiment.ToString(), refreshableHiveExperiment.HiveExperiment.ExecutionTime));
|
---|
300 |
|
---|
301 | // get scopes
|
---|
302 | int j = 0;
|
---|
303 | foreach (var hiveJob in hiveExperiment.HiveJobs) {
|
---|
304 | if (hiveJob.Job.State != JobState.Finished)
|
---|
305 | throw new HiveEngineException("Job failed: " + hiveJob.Job.StateLog.Last().Exception);
|
---|
306 |
|
---|
307 | var scope = ((IAtomicOperation)((EngineJob)hiveJob.ItemJob).InitialOperation).Scope;
|
---|
308 | scopes[j++] = scope;
|
---|
309 | }
|
---|
310 | return scopes;
|
---|
311 | }
|
---|
312 | catch (OperationCanceledException e) {
|
---|
313 | lock (locker) {
|
---|
314 | if (jobIndices != null) DeleteHiveExperiment(hiveExperiment.Id);
|
---|
315 | }
|
---|
316 | throw e;
|
---|
317 | }
|
---|
318 | catch (Exception e) {
|
---|
319 | lock (locker) {
|
---|
320 | if (jobIndices != null) DeleteHiveExperiment(hiveExperiment.Id);
|
---|
321 | }
|
---|
322 | LogException(e);
|
---|
323 | throw e;
|
---|
324 | }
|
---|
325 | }
|
---|
326 |
|
---|
327 | private RefreshableHiveExperiment CreateHiveExperiment() {
|
---|
328 | lock (locker) {
|
---|
329 | var hiveExperiment = new HiveExperiment();
|
---|
330 | hiveExperiment.Name = "HiveEngine Run " + hiveExperiments.Count;
|
---|
331 | hiveExperiment.DateCreated = DateTime.Now;
|
---|
332 | hiveExperiment.UseLocalPlugins = this.UseLocalPlugins;
|
---|
333 | hiveExperiment.ResourceNames = this.ResourceNames;
|
---|
334 | var refreshableHiveExperiment = new RefreshableHiveExperiment(hiveExperiment);
|
---|
335 | refreshableHiveExperiment.IsControllable = false;
|
---|
336 | hiveExperiments.Add(refreshableHiveExperiment);
|
---|
337 | return refreshableHiveExperiment;
|
---|
338 | }
|
---|
339 | }
|
---|
340 |
|
---|
341 | private void DisposeHiveExperiment(RefreshableHiveExperiment refreshableHiveExperiment) {
|
---|
342 | refreshableHiveExperiment.RefreshAutomatically = false;
|
---|
343 | DeleteHiveExperiment(refreshableHiveExperiment.HiveExperiment.Id);
|
---|
344 | ClearData(refreshableHiveExperiment);
|
---|
345 | }
|
---|
346 |
|
---|
347 | private void ClearData(RefreshableHiveExperiment refreshableHiveExperiment) {
|
---|
348 | var jobs = refreshableHiveExperiment.HiveExperiment.GetAllHiveJobs();
|
---|
349 | foreach (var job in jobs) {
|
---|
350 | job.ClearData();
|
---|
351 | }
|
---|
352 | }
|
---|
353 |
|
---|
354 | private void DeleteHiveExperiment(Guid hiveExperimentId) {
|
---|
355 | ExperimentManagerClient.TryAndRepeat(() => {
|
---|
356 | ServiceLocator.Instance.CallHiveService(s => s.DeleteHiveExperiment(hiveExperimentId));
|
---|
357 | }, 5, string.Format("Could not delete jobs"));
|
---|
358 | }
|
---|
359 |
|
---|
360 | private List<Guid> GetResourceIds() {
|
---|
361 | return ServiceLocator.Instance.CallHiveService(service => {
|
---|
362 | var resourceNames = ResourceNames.Split(';');
|
---|
363 | var resourceIds = new List<Guid>();
|
---|
364 | foreach (var resourceName in resourceNames) {
|
---|
365 | Guid resourceId = service.GetResourceId(resourceName);
|
---|
366 | if (resourceId == Guid.Empty) {
|
---|
367 | throw new ResourceNotFoundException(string.Format("Could not find the resource '{0}'", resourceName));
|
---|
368 | }
|
---|
369 | resourceIds.Add(resourceId);
|
---|
370 | }
|
---|
371 | return resourceIds;
|
---|
372 | });
|
---|
373 | }
|
---|
374 |
|
---|
375 | /// <summary>
|
---|
376 | /// Threadsafe message logging
|
---|
377 | /// </summary>
|
---|
378 | private void LogMessage(string message) {
|
---|
379 | lock (logLocker) {
|
---|
380 | Log.LogMessage(message);
|
---|
381 | }
|
---|
382 | }
|
---|
383 |
|
---|
384 | /// <summary>
|
---|
385 | /// Threadsafe exception logging
|
---|
386 | /// </summary>
|
---|
387 | private void LogException(Exception exception) {
|
---|
388 | lock (logLocker) {
|
---|
389 | Log.LogException(exception);
|
---|
390 | }
|
---|
391 | }
|
---|
392 |
|
---|
393 | // testfunction:
|
---|
394 | //private IScope[] ExecuteLocally(EngineJob[] jobs, IScope parentScopeClone, CancellationToken cancellationToken) {
|
---|
395 | // IScope[] scopes = new Scope[jobs.Length];
|
---|
396 | // for (int i = 0; i < jobs.Length; i++) {
|
---|
397 | // var serialized = PersistenceUtil.Serialize(jobs[i]);
|
---|
398 | // var deserialized = PersistenceUtil.Deserialize<IJob>(serialized);
|
---|
399 | // deserialized.Start();
|
---|
400 | // while (deserialized.ExecutionState != ExecutionState.Stopped) {
|
---|
401 | // Thread.Sleep(100);
|
---|
402 | // }
|
---|
403 | // var serialized2 = PersistenceUtil.Serialize(deserialized);
|
---|
404 | // var deserialized2 = PersistenceUtil.Deserialize<EngineJob>(serialized2);
|
---|
405 | // var newScope = ((IAtomicOperation)deserialized2.InitialOperation).Scope;
|
---|
406 | // scopes[i] = newScope;
|
---|
407 | // }
|
---|
408 | // return scopes;
|
---|
409 | //}
|
---|
410 | }
|
---|
411 | }
|
---|