diff --git a/command/init.go b/command/init.go
index 0b9be934b..8827f5e9d 100644
--- a/command/init.go
+++ b/command/init.go
@@ -104,6 +104,17 @@ job "example" {
 		# Defaults to 1
 		# count = 1
 
+		# Restart Policy - This block defines the restart policy for TaskGroups
+		# attempts defines the number of restarts Nomad will do if Tasks
+		# in this TaskGroup fails in a rolling window of interval duration
+		# The delay value makes Nomad wait for that duration to restart after a Task
+		# fails or crashes.
+		restart {
+			interval = 5m
+			attempts = 10
+			delay = 25s
+		}
+
 		# Define a task to run
 		task "redis" {
 			# Use Docker to run the task.
diff --git a/jobspec/parse.go b/jobspec/parse.go
index f63ac5294..c3c71ac9c 100644
--- a/jobspec/parse.go
+++ b/jobspec/parse.go
@@ -30,6 +30,7 @@ func Parse(r io.Reader) (*structs.Job, error) {
 
 	// Parse the buffer
 	obj, err := hcl.Parse(buf.String())
+
 	if err != nil {
 		return nil, fmt.Errorf("error parsing: %s", err)
 	}
@@ -124,7 +125,7 @@ func parseJob(result *structs.Job, obj *hclobj.Object) error {
 		}
 	}
 
-	// If we have tasks outside, do those
+	// If we have tasks outside, create TaskGroups for them
 	if o := obj.Get("task", false); o != nil {
 		var tasks []*structs.Task
 		if err := parseTasks(&tasks, o); err != nil {
@@ -134,9 +135,10 @@ func parseJob(result *structs.Job, obj *hclobj.Object) error {
 		result.TaskGroups = make([]*structs.TaskGroup, len(tasks), len(tasks)*2)
 		for i, t := range tasks {
 			result.TaskGroups[i] = &structs.TaskGroup{
-				Name:  t.Name,
-				Count: 1,
-				Tasks: []*structs.Task{t},
+				Name:          t.Name,
+				Count:         1,
+				Tasks:         []*structs.Task{t},
+				RestartPolicy: structs.NewRestartPolicy(result.Type),
 			}
 		}
 	}
@@ -180,6 +182,7 @@ func parseGroups(result *structs.Job, obj *hclobj.Object) error {
 		delete(m, "constraint")
 		delete(m, "meta")
 		delete(m, "task")
+		delete(m, "restart")
 
 		// Default count to 1 if not specified
 		if _, ok := m["count"]; !ok {
@@ -200,6 +203,10 @@ func parseGroups(result *structs.Job, obj *hclobj.Object) error {
 			}
 		}
 
+		if err := parseRestartPolicy(structs.NewRestartPolicy(result.Type), o); err != nil {
+			return err
+		}
+
 		// Parse out meta fields. These are in HCL as a list so we need
 		// to iterate over them and merge them.
 		if metaO := o.Get("meta", false); metaO != nil {
@@ -228,6 +235,42 @@ func parseGroups(result *structs.Job, obj *hclobj.Object) error {
 	return nil
 }
 
+func parseRestartPolicy(result *structs.RestartPolicy, obj *hclobj.Object) error {
+	var restartHclObj *hclobj.Object
+	var m map[string]interface{}
+	if restartHclObj = obj.Get("restart", false); restartHclObj == nil {
+		return nil
+	}
+	if err := hcl.DecodeObject(&m, restartHclObj); err != nil {
+		return err
+	}
+
+	if delay, ok := m["delay"]; ok {
+		d, err := toDuration(delay)
+		if err != nil {
+			return fmt.Errorf("Invalid Delay time in restart policy: %v", err)
+		}
+		result.Delay = d
+	}
+
+	if interval, ok := m["interval"]; ok {
+		i, err := toDuration(interval)
+		if err != nil {
+			return fmt.Errorf("Invalid Interval time in restart policy: %v", err)
+		}
+		result.Interval = i
+	}
+
+	if attempts, ok := m["attempts"]; ok {
+		a, err := toInteger(attempts)
+		if err != nil {
+			return fmt.Errorf("Invalid value in attempts: %v", err)
+		}
+		result.Attempts = a
+	}
+	return nil
+}
+
 func parseConstraints(result *[]*structs.Constraint, obj *hclobj.Object) error {
 	for _, o := range obj.Elem(false) {
 		var m map[string]interface{}
@@ -477,3 +520,35 @@ func parseUpdate(result *structs.UpdateStrategy, obj *hclobj.Object) error {
 	}
 	return nil
 }
+
+func toDuration(value interface{}) (time.Duration, error) {
+	var dur time.Duration
+	var err error
+	switch v := value.(type) {
+	case string:
+		dur, err = time.ParseDuration(v)
+	case int:
+		dur = time.Duration(v) * time.Second
+	default:
+		err = fmt.Errorf("Invalid time %s", value)
+	}
+
+	return dur, err
+}
+
+func toInteger(value interface{}) (int, error) {
+	var integer int
+	var err error
+	switch v := value.(type) {
+	case string:
+		var i int64
+		i, err = strconv.ParseInt(v, 10, 32)
+		integer = int(i)
+	case int:
+		integer = v
+	default:
+		err = fmt.Errorf("Value: %v can't be parsed into int", value)
+	}
+
+	return integer, err
+}
diff --git a/jobspec/parse_test.go b/jobspec/parse_test.go
index f91789ddb..c3b91e785 100644
--- a/jobspec/parse_test.go
+++ b/jobspec/parse_test.go
@@ -48,6 +48,11 @@ func TestParse(t *testing.T) {
 					&structs.TaskGroup{
 						Name:  "outside",
 						Count: 1,
+						RestartPolicy: &structs.RestartPolicy{
+							Attempts: 2,
+							Interval: 1 * time.Minute,
+							Delay:    15 * time.Second,
+						},
 						Tasks: []*structs.Task{
 							&structs.Task{
 								Name:   "outside",
diff --git a/jobspec/test-fixtures/basic.hcl b/jobspec/test-fixtures/basic.hcl
index 941272b2d..bf81a6ae7 100644
--- a/jobspec/test-fixtures/basic.hcl
+++ b/jobspec/test-fixtures/basic.hcl
@@ -31,6 +31,11 @@ job "binstore-storagelocker" {
 
     group "binsl" {
         count = 5
+        restart {
+            attempts = 5
+            interval = "10m"
+            delay = "15s"
+        }
         task "binstore" {
             driver = "docker"
             config {
diff --git a/nomad/structs/structs.go b/nomad/structs/structs.go
index f5d20552a..8afe1c452 100644
--- a/nomad/structs/structs.go
+++ b/nomad/structs/structs.go
@@ -898,6 +898,37 @@ func (u *UpdateStrategy) Rolling() bool {
 	return u.Stagger > 0 && u.MaxParallel > 0
 }
 
+// RestartPolicy influences how Nomad restarts Tasks when they
+// crash or fail.
+type RestartPolicy struct {
+	Attempts int
+	Interval time.Duration
+	Delay    time.Duration
+}
+
+func (r *RestartPolicy) Validate() error {
+	if time.Duration(r.Attempts)*r.Delay > r.Interval {
+		return fmt.Errorf("Nomad can't restart the TaskGroup %v times in an interval of %v with a delay of %v", r.Attempts, r.Interval, r.Delay)
+	}
+	return nil
+}
+
+func NewRestartPolicy(jobType string) *RestartPolicy {
+	defaultDelayBetweenRestarts := 15 * time.Second
+	defaultAttempts := 15
+	var defaultRestartInterval time.Duration
+
+	if jobType == "service" {
+		defaultRestartInterval = 1 * time.Minute
+		defaultAttempts = 2
+	}
+	return &RestartPolicy{
+		Attempts: defaultAttempts,
+		Interval: defaultRestartInterval,
+		Delay:    defaultDelayBetweenRestarts,
+	}
+}
+
 // TaskGroup is an atomic unit of placement. Each task group belongs to
 // a job and may contain any number of tasks. A task group support running
 // in many replicas using the same configuration..
@@ -913,6 +944,9 @@ type TaskGroup struct {
 	// all the tasks contained.
 	Constraints []*Constraint
 
+	//RestartPolicy of a TaskGroup
+	RestartPolicy *RestartPolicy
+
 	// Tasks are the collection of tasks that this task group needs to run
 	Tasks []*Task
 
@@ -940,6 +974,10 @@ func (tg *TaskGroup) Validate() error {
 		}
 	}
 
+	if err := tg.RestartPolicy.Validate(); err != nil {
+		mErr.Errors = append(mErr.Errors, err)
+	}
+
 	// Check for duplicate tasks
 	tasks := make(map[string]int)
 	for idx, task := range tg.Tasks {
@@ -954,6 +992,7 @@ func (tg *TaskGroup) Validate() error {
 
 	// Validate the tasks
 	for idx, task := range tg.Tasks {
+
 		if err := task.Validate(); err != nil {
 			outer := fmt.Errorf("Task %d validation failed: %s", idx+1, err)
 			mErr.Errors = append(mErr.Errors, outer)