scheduler: add logic for check for migration opportunities method

Author: csegarragonz

include/faabric/scheduler/Scheduler.h

@@ -22,6 +22,10 @@
 
 namespace faabric::scheduler {
 
+typedef std::pair<std::shared_ptr<BatchExecuteRequest>,
+                  std::shared_ptr<faabric::util::SchedulingDecision>>
+  InFlightPair;
+
 class Scheduler;
 
 Scheduler& getScheduler();
@@ -223,7 +227,12 @@ class Scheduler
     // ----------------------------------
     // Function Migration
     // ----------------------------------
-    void checkForMigrationOpportunities();
+    void checkForMigrationOpportunities(
+      faabric::util::MigrationStrategy =
+        faabric::util::MigrationStrategy::BIN_PACK);
+
+    std::shared_ptr<faabric::PendingMigrations> canAppBeMigrated(
+      uint32_t appId);
 
   private:
     std::string thisHost;
@@ -295,6 +304,11 @@ class Scheduler
 
     // ---- Point-to-point ----
     faabric::transport::PointToPointBroker& broker;
+
+    // ---- Function migration ----
+    std::unordered_map<uint32_t, InFlightPair> inFlightRequests;
+    std::unordered_map<uint32_t, std::shared_ptr<faabric::PendingMigrations>>
+      pendingMigrations;
 };
 
 }

include/faabric/util/scheduling.h

@@ -54,4 +54,16 @@ enum SchedulingTopologyHint
     FORCE_LOCAL,
     NEVER_ALONE
 };
+
+// Migration strategies help the scheduler decide wether the scheduling decision
+// for a batch request could be changed with the new set of available resources.
+// - BIN_PACK: sort hosts by the number of functions from the batch they are
+//             running. Bin-pack batches in increasing order to hosts in
+//             decreasing order.
+// - EMPTY_HOSTS: pack batches in increasing order to empty hosts.
+enum MigrationStrategy
+{
+    BIN_PACK,
+    EMPTY_HOSTS
+};
 }

src/proto/faabric.proto

@@ -245,3 +245,20 @@ message PointToPointMappings {
 
     repeated PointToPointMapping mappings = 3;
 }
+
+// ---------------------------------------------
+// FUNCTION MIGRATIONS
+// ---------------------------------------------
+
+message PendingMigrations {
+    int32 appId = 1;
+    int32 groupId = 2;
+
+    message PendingMigration {
+        int32 messageId = 1;
+        string srcHost = 2;
+        string dstHost = 3;
+    }
+
+    repeated PendingMigration migrations = 3;
+}

src/scheduler/Scheduler.cpp

@@ -137,6 +137,10 @@ void Scheduler::reset()
     threadResults.clear();
     pushedSnapshotsMap.clear();
 
+    // Reset function migration tracking
+    inFlightRequests.clear();
+    pendingMigrations.clear();
+
     // Records
     recordedMessagesAll.clear();
     recordedMessagesLocal.clear();
@@ -452,6 +456,25 @@ faabric::util::SchedulingDecision Scheduler::doCallFunctions(
         throw std::runtime_error("Invalid scheduler hint for messages");
     }
 
+    // Record in-flight request if function desires to be migrated
+    if (firstMsg.migrationcheckperiod() > 0) {
+        auto decisionPtr =
+          std::make_shared<faabric::util::SchedulingDecision>(decision);
+        inFlightRequests[decision.appId] = std::make_pair(req, decisionPtr);
+        /*
+        if (inFlightRequests.size() == 1) {
+            functionMigrationThread.start(firstMsg.migrationcheckperiod());
+        } else if (firstMsg.migrationcheckperiod() !=
+                   functionMigrationThread.wakeUpPeriodSeconds) {
+            SPDLOG_WARN("Ignoring migration check period as the migration"
+                        "thread was initialised with a different one."
+                        "(provided: {}, current: {})",
+                        firstMsg.migrationcheckperiod(),
+                        functionMigrationThread.wakeUpPeriodSeconds);
+        }
+        */
+    }
+
     // NOTE: we want to schedule things on this host _last_, otherwise functions
     // may start executing before all messages have been dispatched, thus
     // slowing the remaining scheduling.
@@ -889,6 +912,20 @@ void Scheduler::setFunctionResult(faabric::Message& msg)
     // Write the successful result to the result queue
     std::vector<uint8_t> inputData = faabric::util::messageToBytes(msg);
     redis.publishSchedulerResult(key, msg.statuskey(), inputData);
+
+    // Remove the app from in-flight map if still there, and this host is the
+    // master host for the message
+    if (msg.masterhost() == thisHost) {
+        faabric::util::FullLock lock(mx);
+
+        inFlightRequests.erase(msg.appid());
+        pendingMigrations.erase(msg.appid());
+        // If there are no more apps to track, stop the thread checking for
+        // migration opportunities
+        if (inFlightRequests.size() == 0) {
+            // functionMigrationThread.stop();
+        }
+    }
 }
 
 void Scheduler::registerThread(uint32_t msgId)
@@ -1130,8 +1167,136 @@ ExecGraphNode Scheduler::getFunctionExecGraphNode(unsigned int messageId)
     return node;
 }
 
-void Scheduler::checkForMigrationOpportunities()
+void Scheduler::checkForMigrationOpportunities(
+  faabric::util::MigrationStrategy migrationStrategy)
 {
-    SPDLOG_INFO("Not implemented");
+    // Vector to cache all migrations we have to do, and update the shared map
+    // at the very end just once. This is because we need a unique lock to write
+    // to the shared map, but the rest of this method can do with a shared lock.
+    std::vector<std::shared_ptr<faabric::PendingMigrations>>
+      tmpPendingMigrations;
+
+    {
+        faabric::util::SharedLock lock(mx);
+
+        // For each in-flight request, check if there is an opportunity to
+        // migrate
+        for (const auto& app : inFlightRequests) {
+            auto req = app.second.first;
+            auto originalDecision = *app.second.second;
+
+            // If we have already recorded a pending migration for this req,
+            // skip
+            if (canAppBeMigrated(originalDecision.appId) != nullptr) {
+                continue;
+            }
+
+            faabric::PendingMigrations msg;
+            msg.set_appid(originalDecision.appId);
+            // TODO - generate a new groupId here for processes to wait on
+            // during the migration? msg.set_groupid();
+
+            if (migrationStrategy ==
+                faabric::util::MigrationStrategy::BIN_PACK) {
+                // We assume the batch was originally scheduled using
+                // bin-packing, thus the scheduling decision has at the begining
+                // (left) the hosts with the most allocated requests, and at the
+                // end (right) the hosts with the fewest. To check for migration
+                // oportunities, we compare a pointer to the possible
+                // destination of the migration (left), with one to the possible
+                // source of the migration (right). NOTE - this is a slight
+                // simplification, but makes the code simpler.
+                auto left = originalDecision.hosts.begin();
+                auto right = originalDecision.hosts.end() - 1;
+                faabric::HostResources r = (*left == thisHost)
+                                             ? getThisHostResources()
+                                             : getHostResources(*left);
+                auto nAvailable = [&r]() -> int {
+                    return r.slots() - r.usedslots();
+                };
+                auto claimSlot = [&r]() {
+                    int currentUsedSlots = r.usedslots();
+                    SPDLOG_INFO("Old slots: {} - New slots: {}",
+                                currentUsedSlots,
+                                currentUsedSlots + 1);
+                    r.set_usedslots(currentUsedSlots + 1);
+                };
+                while (left < right) {
+                    // If both pointers point to the same host, no migration
+                    // opportunity, and must check another possible source of
+                    // the migration
+                    if (*left == *right) {
+                        --right;
+                        continue;
+                    }
+
+                    // If the left pointer (possible destination of the
+                    // migration) is out of available resources, no migration
+                    // opportunity, and must check another possible destination
+                    // of migration
+                    if (nAvailable() == 0) {
+                        auto oldHost = *left;
+                        ++left;
+                        if (*left != oldHost) {
+                            r = (*left == thisHost) ? getThisHostResources()
+                                                    : getHostResources(*left);
+                        }
+                        continue;
+                    }
+
+                    // If each pointer points to a request scheduled in a
+                    // different host, and the possible destination has slots,
+                    // there is a migration opportunity
+                    auto* migration = msg.add_migrations();
+                    auto msgIdPtr =
+                      originalDecision.messageIds.begin() +
+                      std::distance(originalDecision.hosts.begin(), right);
+                    migration->set_messageid(*msgIdPtr);
+                    migration->set_srchost(*right);
+                    migration->set_dsthost(*left);
+                    // Decrement by one the availability, and check for more
+                    // possible sources of migration
+                    claimSlot();
+                    --right;
+                }
+            } else {
+                SPDLOG_ERROR("Unrecognised migration strategy: {}",
+                             migrationStrategy);
+                throw std::runtime_error("Unrecognised migration strategy.");
+            }
+
+            if (msg.migrations_size() > 0) {
+                tmpPendingMigrations.emplace_back(
+                  std::make_shared<faabric::PendingMigrations>(msg));
+                SPDLOG_DEBUG("Detected migration opportunity for app: {}",
+                             msg.appid());
+            } else {
+                SPDLOG_DEBUG("No migration opportunity detected for app: {}",
+                             msg.appid());
+            }
+        }
+    }
+
+    // Finally, store all the pending migrations in the shared map acquiring
+    // a unique lock.
+    if (tmpPendingMigrations.size() > 0) {
+        faabric::util::FullLock lock(mx);
+        for (auto msgPtr : tmpPendingMigrations) {
+            SPDLOG_INFO("Adding app: {}", msgPtr->appid());
+            pendingMigrations[msgPtr->appid()] = std::move(msgPtr);
+        }
+    }
+}
+
+std::shared_ptr<faabric::PendingMigrations> Scheduler::canAppBeMigrated(
+  uint32_t appId)
+{
+    faabric::util::SharedLock lock(mx);
+
+    if (pendingMigrations.find(appId) == pendingMigrations.end()) {
+        return nullptr;
+    }
+
+    return pendingMigrations[appId];
 }
 }

tests/test/scheduler/test_executor.cpp

@@ -201,6 +201,18 @@ int32_t TestExecutor::executeTask(
         throw std::runtime_error("This is a test error");
     }
 
+    if (msg.function() == "migration") {
+        // Sleep for sufficiently more than the check period
+        SPDLOG_DEBUG("Migration test function going to sleep");
+        SLEEP_MS(SHORT_TEST_TIMEOUT_MS);
+        SPDLOG_DEBUG("Migration test function waking up");
+
+        msg.set_outputdata(
+          fmt::format("Migration test function {} executed", msg.id()));
+
+        return 0;
+    }
+
     if (reqOrig->type() == faabric::BatchExecuteRequest::THREADS) {
         SPDLOG_DEBUG("TestExecutor executing simple thread {}", msg.id());
         return msg.id() / 100;