diff --git a/engine/resources.go b/engine/resources.go
index ac8d05e0e..0c3baeae1 100755
--- a/engine/resources.go
+++ b/engine/resources.go
@@ -19,11 +19,13 @@ package engine
 
 import (
 	"fmt"
+	"math/rand"
 	"reflect"
 	"sort"
 	"sync"
 	"time"
 
+	"github.com/cgrates/cgrates/cache"
 	"github.com/cgrates/cgrates/config"
 	"github.com/cgrates/cgrates/utils"
 	"github.com/cgrates/rpcclient"
@@ -61,9 +63,9 @@ type Resource struct {
 	ID     string
 	Usages map[string]*ResourceUsage
 	TTLIdx []string     // holds ordered list of ResourceIDs based on their TTL, empty if feature is disabled
-	rCfg   *ResourceCfg // optional configuration attached
 	tUsage *float64     // sum of all usages
 	dirty  *bool        // the usages were modified, needs save, *bool so we only save if enabled in config
+	rCfg   *ResourceCfg // for ordering purposes
 }
 
 // removeExpiredUnits removes units which are expired from the resource
@@ -218,22 +220,93 @@ func NewResourceService(cfg *config.CGRConfig, dataDB DataDB, statS rpcclient.Rp
 
 // ResourceService is the service handling resources
 type ResourceService struct {
-	dataDB         DataDB // So we can load the data in cache and index it
-	statS          rpcclient.RpcClientConnection
-	eventResources map[string][]string // map[ruID][]string{rID} for faster queries
-	erMux          sync.RWMutex
+	dataDB          DataDB // So we can load the data in cache and index it
+	statS           rpcclient.RpcClientConnection
+	eventResources  map[string][]string // map[ruID][]string{rID} for faster queries
+	erMux           sync.RWMutex
+	storedResources utils.StringMap // keep a record of resources which need saving, map[resID]bool
+	srMux           sync.RWMutex
+	stopBackup      chan struct{} // control storing process
+	backupInterval  time.Duration
 }
 
 // Called to start the service
-func (rS *ResourceService) ListenAndServe() error {
+func (rS *ResourceService) ListenAndServe(exitChan chan bool) error {
+	go rS.runBackup() // start backup loop
+	e := <-exitChan
+	exitChan <- e // put back for the others listening for shutdown request
 	return nil
 }
 
 // Called to shutdown the service
 func (rS *ResourceService) ServiceShutdown() error {
+	utils.Logger.Info("<ResourceS> service shutdown initialized")
+	close(rS.stopBackup)
+	rS.storeResources()
+	utils.Logger.Info("<ResourceS> service shutdown complete")
 	return nil
 }
 
+// storeResource stores the necessary storedMetrics to dataDB
+func (rS *ResourceService) StoreResource(r *Resource) (err error) {
+	if r.dirty == nil || !*r.dirty {
+		return
+	}
+	if err = rS.dataDB.SetResource(r); err != nil {
+		utils.Logger.Warning(
+			fmt.Sprintf("<ResourceS> failed saving Resource with ID: %s, error: %s",
+				r.ID, err.Error()))
+	} else {
+		*r.dirty = false
+	}
+	return
+}
+
+// storeResources executes one task of complete backup
+func (rS *ResourceService) storeResources() {
+	var failedRIDs []string
+	for {
+		rS.srMux.Lock()
+		rID := rS.storedResources.GetOne()
+		if rID != "" {
+			delete(rS.storedResources, rID)
+		}
+		rS.srMux.Unlock()
+		if rID == "" {
+			break // no more keys, backup completed
+		}
+		if rIf, ok := cache.Get(utils.ResourcesPrefix + rID); !ok || rIf == nil {
+			utils.Logger.Warning(fmt.Sprintf("<ResourceS> failed retrieving from cache resource with ID: %s"))
+		} else if err := rS.StoreResource(rIf.(*Resource)); err != nil {
+			failedRIDs = append(failedRIDs, rID) // record failure so we can schedule it for next backup
+		}
+		// randomize the CPU load and give up thread control
+		time.Sleep(time.Duration(rand.Intn(1000)) * time.Nanosecond)
+	}
+	if len(failedRIDs) != 0 { // there were errors on save, schedule the keys for next backup
+		rS.srMux.Lock()
+		for _, rID := range failedRIDs {
+			rS.storedResources[rID] = true
+		}
+		rS.srMux.Unlock()
+	}
+}
+
+// backup will regularly store resources changed to dataDB
+func (rS *ResourceService) runBackup() {
+	if rS.backupInterval <= 0 {
+		return
+	}
+	for {
+		select {
+		case <-rS.stopBackup:
+			return
+		}
+		rS.storeResources()
+	}
+	time.Sleep(rS.backupInterval)
+}
+
 // cachedResourcesForEvent attempts to retrieve cached resources for an event
 // returns nil if event not cached or errors occur
 func (rS *ResourceService) cachedResourcesForEvent(evUUID string) (rs Resources) {
@@ -295,6 +368,9 @@ func (rS *ResourceService) matchingResourcesForEvent(ev map[string]interface{})
 		if err != nil {
 			return nil, err
 		}
+		if rCfg.Stored {
+			r.dirty = utils.BoolPointer(false)
+		}
 		r.rCfg = rCfg
 		matchingResources[rCfg.ID] = r // Cannot save it here since we could have errors after and resource will remain unused
 	}
@@ -362,9 +438,21 @@ func (rS *ResourceService) V1AllocateResource(args utils.AttrRLsResourceUsage, r
 	if err != nil {
 		return err
 	}
+	// index it for matching out of cache
 	rS.erMux.Lock()
 	rS.eventResources[args.UsageID] = mtcRLs.ids()
 	rS.erMux.Unlock()
+	// index it for storing
+	rS.srMux.Lock()
+	for _, r := range mtcRLs {
+		if rS.backupInterval == -1 {
+			rS.StoreResource(r)
+		} else if r.dirty != nil {
+			*r.dirty = true // mark it to be saved
+		}
+		rS.storedResources[r.ID] = true
+	}
+	rS.srMux.Unlock()
 	*reply = alcMsg
 	return
 }
@@ -381,6 +469,22 @@ func (rS *ResourceService) V1ReleaseResource(args utils.AttrRLsResourceUsage, re
 	rS.erMux.Lock()
 	delete(rS.eventResources, args.UsageID)
 	rS.erMux.Unlock()
+	if rS.backupInterval != -1 {
+		rS.srMux.Lock()
+	}
+	for _, r := range mtcRLs {
+		if r.dirty != nil {
+			if rS.backupInterval == -1 {
+				rS.StoreResource(r)
+			} else {
+				*r.dirty = true // mark it to be saved
+				rS.storedResources[r.ID] = true
+			}
+		}
+	}
+	if rS.backupInterval != -1 {
+		rS.srMux.Unlock()
+	}
 	*reply = utils.OK
 	return nil
 }