diff --git a/elasticsearch/helpers/__init__.py b/elasticsearch/helpers/__init__.py
index 5f0f240f..7b86da40 100644
--- a/elasticsearch/helpers/__init__.py
+++ b/elasticsearch/helpers/__init__.py
@@ -47,12 +47,13 @@ def expand_action(data):
 
     return action, data.get('_source', data)
 
-def _chunk_actions(actions, chunk_size, max_chunk_bytes, serializer):
+def _chunk_actions(actions, chunk_size, max_chunk_bytes, serializer, include_data=False):
     """
     Split actions into chunks by number or size, serialize them into strings in
     the process.
     """
     bulk_actions = []
+    bulk_data = []
     size, action_count = 0, 0
     for action, data in actions:
         action = serializer.dumps(action)
@@ -64,18 +65,32 @@ def _chunk_actions(actions, chunk_size, max_chunk_bytes, serializer):
 
         # full chunk, send it and start a new one
         if bulk_actions and (size + cur_size > max_chunk_bytes or action_count == chunk_size):
-            yield bulk_actions
+            if include_data:
+                yield bulk_data, bulk_actions
+            else:
+                yield bulk_actions
             bulk_actions = []
+            bulk_data = []
             size, action_count = 0, 0
 
         bulk_actions.append(action)
         if data is not None:
             bulk_actions.append(data)
+
+        if include_data:
+            if data is not None:
+                bulk_data.append((action, data))
+            else:
+                bulk_data.append((action, ))
+
         size += cur_size
         action_count += 1
 
     if bulk_actions:
-        yield bulk_actions
+        if include_data:
+            yield bulk_data, bulk_actions
+        else:
+            yield bulk_actions
 
 def _process_bulk_chunk(client, bulk_actions, raise_on_exception=True, raise_on_error=True, **kwargs):
     """
diff --git a/elasticsearch/helpers/backoff.py b/elasticsearch/helpers/backoff.py
new file mode 100644
index 00000000..950441a5
--- /dev/null
+++ b/elasticsearch/helpers/backoff.py
@@ -0,0 +1,63 @@
+import time
+
+from . import expand_action, _chunk_actions, _process_bulk_chunk
+from ..exceptions import TransportError
+
+
+def backoff_bulk(client, actions, chunk_size=500, max_chunk_bytes=100 * 1024 * 1024,
+                 expand_action_callback=expand_action, max_retries=-1,
+                 initial_backoff=2, max_backoff=600, **kwargs):
+    """
+    Bulk helper that implements proper retry strategy when dealing with
+    rejecttions - when the target cluster is overloaded, returning ``429``
+    responses, the bulk actions will be retried after a backoff period to allow
+    the cluster to recover.
+
+    This function is a generator yielding all documents that failed to index.
+
+    :arg client: instance of :class:`~elasticsearch.Elasticsearch` to use
+    :arg actions: iterable containing the actions to be executed
+    :arg chunk_size: number of docs in one chunk sent to es (default: 500)
+    :arg max_chunk_bytes: the maximum size of the request in bytes (default: 100MB)
+    :arg expand_action_callback: callback executed on each action passed in,
+        should return a tuple containing the action line and the data line
+        (`None` if data line should be omitted).
+    :arg max_retries: maximum number of times a document will be retried
+    :arg initial_backoff: number of seconds we should wait before the first
+        retry. Any subsequent retries will be powers of ``inittial_backoff *
+        2**retry_number``
+    :arg max_backoff: maximum number of seconds a retry will wait
+    """
+    actions = map(expand_action_callback, actions)
+
+    for bulk_data, bulk_actions in _chunk_actions(actions, chunk_size, max_chunk_bytes, client.transport.serializer, include_data=True):
+        retry = 0
+        while max_retries == -1 or retry <= max_retries:
+            to_retry = []
+            to_retry_data = []
+            if retry:
+                time.sleep(min(max_backoff, initial_backoff * 2**(retry-1)))
+
+            try:
+                for data, (ok, info) in zip(bulk_data, _process_bulk_chunk(client, bulk_actions, raise_on_exception=True, raise_on_error=False, **kwargs)):
+                    if not ok:
+                        action, info = info.popitem()
+                        if info['status'] == 429 and (retry+1) <= max_retries:
+                            to_retry.extend(data)
+                            to_retry_data.append(data)
+                        else:
+                            info['data'] = data
+                            yield {action: info}
+
+            except TransportError as e:
+                if e.status_code != 429:
+                    raise
+
+                retry += 1
+            else:
+                if not to_retry:
+                    break
+                bulk_actions = to_retry
+                bulk_data = to_retry_data
+
+                retry += 1