| 1 | /* |
| 2 | * DLB.hpp |
| 3 | * |
| 4 | * Created on: Nov 20, 2015 |
| 5 | * Author: Antonio Leo |
| 6 | */ |
| 7 | |
| 8 | #ifndef SRC_DECOMPOSITION_DLB_HPP_ |
| 9 | #define SRC_DECOMPOSITION_DLB_HPP_ |
| 10 | |
| 11 | //! Time structure for statistical purposes |
| 12 | struct Times |
| 13 | { |
| 14 | //! starting time of the simulation (0) |
| 15 | size_t simulationStartTime = 0; |
| 16 | |
| 17 | //! End iteration of the simulation |
| 18 | size_t simulationEndTime; |
| 19 | |
| 20 | //! integration time |
| 21 | double timeStep = 0.1; |
| 22 | |
| 23 | //! Interval between teo rebalance |
| 24 | |
| 25 | //! Start time |
| 26 | size_t iterationStartTime; |
| 27 | |
| 28 | //! End time |
| 29 | size_t iterationEndTime; |
| 30 | }; |
| 31 | |
| 32 | /*! Class that implements the two heuristics to determine when a re-balance of the distribution is needed. |
| 33 | * |
| 34 | * Used heuristics are: SAR and Un-balance Threshold (Default)\n |
| 35 | * |
| 36 | * To chose the heuristic use the method setHeuristic(Heuristic) |
| 37 | * |
| 38 | * In the SAR heuristic the following formula is applied:\n |
| 39 | * \f$W_{n} = \frac{\sum_{j=1}^{n} (T_{max}(j) - T_{avg}(j)) + C} {n}\f$ |
| 40 | * |
| 41 | * \f$T_{max}(j)\f$ – wall-clock time of bottleneck process in time step j\n |
| 42 | * \f$T_{avg}(j)\f$ – average wall-clock time for time step j over all processes\n |
| 43 | * \f$C\f$ – cost of re-decomposing the problem\n |
| 44 | * \f$n\f$ – number of time steps since last re-decomposition\n |
| 45 | * \n |
| 46 | * For small n, load balance is good and W decreases since C is amortized over an increasing number of time steps. |
| 47 | * As the accumulated idle time starts to dominate, W starts to rise. At this point, C has been fully amortized. |
| 48 | * Re-decompose when \f$W_{n} > W_{n-1}\f$\n |
| 49 | * |
| 50 | * In the Un-balance Threshold heuristic the re-balance is triggered when the un-balance level exceeds a certain level. |
| 51 | * Levels can be chosen in the ThresholdLevel type. |
| 52 | */ |
| 53 | class DLB |
| 54 | { |
| 55 | public: |
| 56 | |
| 57 | //! Type of DLB heuristics |
| 58 | enum Heuristic |
| 59 | { |
| 60 | SAR_HEURISTIC, UNBALANCE_THRLD |
| 61 | }; |
| 62 | |
| 63 | //! Level of un-balance needed to trigger the re-balance |
| 64 | enum ThresholdLevel |
| 65 | { |
| 66 | THRLD_LOW = 5, THRLD_MEDIUM = 7, THRLD_HIGH = 10 |
| 67 | }; |
| 68 | |
| 69 | private: |
| 70 | |
| 71 | //! Runtime virtual cluster machine |
| 72 | Vcluster<> & v_cl; |
| 73 | |
| 74 | //! Structure that will contain all the timings |
| 75 | Times timeInfo; |
| 76 | |
| 77 | //! Wn for SAR heuristic |
| 78 | float w_n = -1; |
| 79 | |
| 80 | //! Computation cost for SAR heuristic |
| 81 | float c_c = 5; |
| 82 | |
| 83 | //! Number of time-steps since the previous DLB |
| 84 | size_t n_ts = 1; |
| 85 | |
| 86 | //! Idle time accumulated so far, needed for SAR heuristic |
| 87 | float i_time = 0; |
| 88 | |
| 89 | //! Vector to collect all timings |
| 90 | openfpm::vector<long> times; |
| 91 | |
| 92 | //! Type of the heuristic to use |
| 93 | Heuristic heuristic = UNBALANCE_THRLD; |
| 94 | |
| 95 | //! Un-balance value |
| 96 | float unbalance = -1; |
| 97 | |
| 98 | //! Threshold value |
| 99 | ThresholdLevel thl = THRLD_MEDIUM; |
| 100 | |
| 101 | /*! \brief Function that gather times informations and decides if a rebalance is needed it uses the SAR heuristic |
| 102 | * |
| 103 | * \return true if re-balance is needed |
| 104 | * |
| 105 | */ |
| 106 | inline bool SAR() |
| 107 | { |
| 108 | long t = timeInfo.iterationEndTime - timeInfo.iterationStartTime; |
| 109 | float t_max = t, t_avg = t; |
| 110 | |
| 111 | // Exchange time informations through processors |
| 112 | v_cl.max(t_max); |
| 113 | v_cl.sum(t_avg); |
| 114 | v_cl.execute(); |
| 115 | |
| 116 | t_avg /= v_cl.getProcessingUnits(); |
| 117 | |
| 118 | // add idle time to vector |
| 119 | i_time += t_max - t_avg; |
| 120 | |
| 121 | // Compute Wn |
| 122 | float nw_n = (i_time + c_c) / n_ts; |
| 123 | |
| 124 | if (w_n == -1) |
| 125 | w_n = nw_n; |
| 126 | |
| 127 | if (nw_n > w_n) |
| 128 | { |
| 129 | i_time = 0; |
| 130 | n_ts = 1; |
| 131 | w_n = nw_n; |
| 132 | return true; |
| 133 | } |
| 134 | else |
| 135 | { |
| 136 | ++n_ts; |
| 137 | w_n = nw_n; |
| 138 | return false; |
| 139 | } |
| 140 | } |
| 141 | |
| 142 | /*! \brief Check if the un-balance has exceeded the threshold |
| 143 | * |
| 144 | * \return true if re-balance is needed, false otherwise |
| 145 | */ |
| 146 | bool unbalanceThreshold() |
| 147 | { |
| 148 | if (unbalance == -1) |
| 149 | { |
| 150 | std::cerr << "Error: Un-balance value must be set before checking DLB." ; |
| 151 | return false; |
| 152 | } |
| 153 | |
| 154 | if (unbalance > thl) |
| 155 | { |
| 156 | return true; |
| 157 | } |
| 158 | |
| 159 | return false; |
| 160 | } |
| 161 | |
| 162 | public: |
| 163 | |
| 164 | /*! \brief Constructor for DLB class |
| 165 | * |
| 166 | * \param v_cl virtual cluster object |
| 167 | */ |
| 168 | DLB(Vcluster<> & v_cl) : |
| 169 | v_cl(v_cl) |
| 170 | { |
| 171 | } |
| 172 | |
| 173 | /*! \brief Set the heuristic to use (default: un-balance threshold) |
| 174 | * |
| 175 | * \param h |
| 176 | */ |
| 177 | void setHeurisitc(Heuristic h) |
| 178 | { |
| 179 | heuristic = h; |
| 180 | } |
| 181 | |
| 182 | /*! \brief Get the heuristic |
| 183 | * |
| 184 | * Indicate which heuristic model is used to calculate when a rebalance |
| 185 | * is needed |
| 186 | * |
| 187 | * \return the Heuristic used by DLB |
| 188 | * |
| 189 | */ |
| 190 | Heuristic getHeurisitc() |
| 191 | { |
| 192 | return heuristic; |
| 193 | } |
| 194 | |
| 195 | /*! \brief check if a re-balance is needed using the selected heuristic |
| 196 | * |
| 197 | * \return true if the rebalance is needed |
| 198 | * |
| 199 | */ |
| 200 | bool rebalanceNeeded() |
| 201 | { |
| 202 | if (heuristic == SAR_HEURISTIC) |
| 203 | { |
| 204 | return SAR(); |
| 205 | } |
| 206 | else |
| 207 | { |
| 208 | return unbalanceThreshold(); |
| 209 | } |
| 210 | } |
| 211 | |
| 212 | /*! \brief Set start time for the simulation |
| 213 | * |
| 214 | * \param t time when the whole simulation starts |
| 215 | */ |
| 216 | void setSimulationStartTime(size_t t) |
| 217 | { |
| 218 | timeInfo.simulationStartTime = t; |
| 219 | } |
| 220 | |
| 221 | /*! \brief Get start time for the simulation |
| 222 | * |
| 223 | * \return the start point of the simulation |
| 224 | * |
| 225 | */ |
| 226 | size_t getSimulationStartTime() |
| 227 | { |
| 228 | return timeInfo.simulationStartTime; |
| 229 | } |
| 230 | |
| 231 | /*! \brief Set end time for the simulation |
| 232 | * |
| 233 | * \param t time when the whole simulation ends |
| 234 | */ |
| 235 | void setSimulationEndTime(size_t t) |
| 236 | { |
| 237 | timeInfo.simulationEndTime = t; |
| 238 | } |
| 239 | |
| 240 | /*! \brief Get end time for the simulation |
| 241 | * |
| 242 | * \return the end time of the simulation |
| 243 | * |
| 244 | */ |
| 245 | size_t getSimulationEndTime() |
| 246 | { |
| 247 | return timeInfo.simulationEndTime; |
| 248 | } |
| 249 | |
| 250 | /*! \brief Set start time for the single iteration |
| 251 | * |
| 252 | */ |
| 253 | void startIteration() |
| 254 | { |
| 255 | timeInfo.iterationStartTime = clock(); |
| 256 | } |
| 257 | |
| 258 | /*! \brief Set start time for the single iteration |
| 259 | * |
| 260 | * \param t time when the one iteration starts |
| 261 | */ |
| 262 | void startIteration(size_t t) |
| 263 | { |
| 264 | timeInfo.iterationStartTime = t; |
| 265 | } |
| 266 | |
| 267 | /*! \brief Set end time for the single iteration |
| 268 | * |
| 269 | * \param time when one iteration is completed |
| 270 | * |
| 271 | */ |
| 272 | void endIteration() |
| 273 | { |
| 274 | timeInfo.iterationEndTime = clock(); |
| 275 | } |
| 276 | |
| 277 | /*! \brief Set the end time when the previous rebalance has been performed |
| 278 | * |
| 279 | * \param t time when one iteration ends |
| 280 | */ |
| 281 | void endIteration(size_t t) |
| 282 | { |
| 283 | timeInfo.iterationEndTime = t; |
| 284 | } |
| 285 | |
| 286 | /*! \brief Set delta time step for one iteration (Computation time) |
| 287 | * |
| 288 | * \param t timestep |
| 289 | */ |
| 290 | void setTimeStep(double t) |
| 291 | { |
| 292 | timeInfo.timeStep = t; |
| 293 | } |
| 294 | |
| 295 | /*! \brief Set time step for the single iteration |
| 296 | * |
| 297 | * \param computation value of the computation cost (default: 5) |
| 298 | */ |
| 299 | void setComputationCost(size_t computation) |
| 300 | { |
| 301 | c_c = computation; |
| 302 | } |
| 303 | |
| 304 | /*! \brief Get how many time-steps have passed since the last re-balancing |
| 305 | * |
| 306 | * \return number of timesteos |
| 307 | * |
| 308 | */ |
| 309 | size_t getNTimeStepSinceDLB() |
| 310 | { |
| 311 | return n_ts; |
| 312 | } |
| 313 | |
| 314 | /*! \brief Set un-balance value |
| 315 | * |
| 316 | * \param u unbalance |
| 317 | */ |
| 318 | void setUnbalance(float u) |
| 319 | { |
| 320 | unbalance = u; |
| 321 | } |
| 322 | |
| 323 | /*! \brief threshold of umbalance to start a rebalance |
| 324 | * |
| 325 | * \param t threshold level |
| 326 | */ |
| 327 | void setThresholdLevel(ThresholdLevel t) |
| 328 | { |
| 329 | thl = t; |
| 330 | } |
| 331 | |
| 332 | }; |
| 333 | |
| 334 | #endif /* SRC_DECOMPOSITION_DLB_HPP_ */ |
| 335 | |