diff --git a/data/database.py b/data/database.py
index c8fbc27ea..ffea9aadd 100644
--- a/data/database.py
+++ b/data/database.py
@@ -853,9 +853,13 @@ class TorrentInfo(BaseModel):
   piece_length = IntegerField()
   pieces = Base64BinaryField()
 
-  indexes = (
-    (('storage', 'piece_length'), True),
-  )
+  class Meta:
+    database = db
+    read_slaves = (read_slave,)
+    indexes = (
+      # we may want to compute the piece hashes multiple times with different piece lengths
+      (('storage', 'piece_length'), True),
+    )
 
 
 all_models = [User, Repository, Image, AccessToken, Role, RepositoryPermission, Visibility,
diff --git a/data/migrations/versions/23ca04d0bc8e_add_the_torrentinfo_table_and_torrent_.py b/data/migrations/versions/23ca04d0bc8e_add_the_torrentinfo_table_and_torrent_.py
new file mode 100644
index 000000000..1e16f9838
--- /dev/null
+++ b/data/migrations/versions/23ca04d0bc8e_add_the_torrentinfo_table_and_torrent_.py
@@ -0,0 +1,38 @@
+"""Add the torrentinfo table and torrent fields on blobuploads.
+
+Revision ID: 23ca04d0bc8e
+Revises: 471caec2cb66
+Create Date: 2016-01-06 13:25:24.597037
+
+"""
+
+# revision identifiers, used by Alembic.
+revision = '23ca04d0bc8e'
+down_revision = '471caec2cb66'
+
+from alembic import op
+import sqlalchemy as sa
+
+def upgrade(tables):
+    ### commands auto generated by Alembic - please adjust! ###
+    op.create_table('torrentinfo',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('storage_id', sa.Integer(), nullable=False),
+    sa.Column('piece_length', sa.Integer(), nullable=False),
+    sa.Column('pieces', sa.Text(), nullable=False),
+    sa.ForeignKeyConstraint(['storage_id'], ['imagestorage.id'], name=op.f('fk_torrentinfo_storage_id_imagestorage')),
+    sa.PrimaryKeyConstraint('id', name=op.f('pk_torrentinfo'))
+    )
+    op.create_index('torrentinfo_storage_id', 'torrentinfo', ['storage_id'], unique=False)
+    op.create_index('torrentinfo_storage_id_piece_length', 'torrentinfo', ['storage_id', 'piece_length'], unique=True)
+    op.add_column(u'blobupload', sa.Column('piece_hashes', sa.Text(), nullable=False))
+    op.add_column(u'blobupload', sa.Column('piece_sha_state', sa.Text(), nullable=True))
+    ### end Alembic commands ###
+
+
+def downgrade(tables):
+    ### commands auto generated by Alembic - please adjust! ###
+    op.drop_column(u'blobupload', 'piece_sha_state')
+    op.drop_column(u'blobupload', 'piece_hashes')
+    op.drop_table('torrentinfo')
+    ### end Alembic commands ###
diff --git a/endpoints/v1/registry.py b/endpoints/v1/registry.py
index d4d33c630..a378f6c5d 100644
--- a/endpoints/v1/registry.py
+++ b/endpoints/v1/registry.py
@@ -239,7 +239,7 @@ def put_image_layer(namespace, repository, image_id):
   updated_storage = model.storage.set_image_storage_metadata(image_id, namespace, repository,
                                                              size_info.compressed_size,
                                                              size_info.uncompressed_size)
-  pieces_bytes = piece_hasher.piece_hashes + piece_hasher.hash_fragment.digest()
+  pieces_bytes = piece_hasher.final_piece_hashes()
   model.storage.save_torrent_info(updated_storage, app.config['TORRENT_PIECE_SIZE'], pieces_bytes)
 
   # Append the computed checksum.
diff --git a/endpoints/verbs.py b/endpoints/verbs.py
index f91a73565..4a0c6edf5 100644
--- a/endpoints/verbs.py
+++ b/endpoints/verbs.py
@@ -15,7 +15,10 @@ from storage import Storage
 
 from util.registry.queuefile import QueueFile
 from util.registry.queueprocess import QueueProcess
-from util.registry.torrent import make_torrent, per_user_torrent_filename, public_torrent_filename
+from util.registry.torrent import (make_torrent, per_user_torrent_filename, public_torrent_filename,
+                                   PieceHasher)
+from util.registry.filelike import wrap_with_handler
+from util.registry.gzipstream import SizeInfo
 from formats.squashed import SquashedDockerImage
 from formats.aci import ACIImage
 from endpoints.v2.blob import BLOB_DIGEST_ROUTE
@@ -25,7 +28,8 @@ verbs = Blueprint('verbs', __name__)
 logger = logging.getLogger(__name__)
 
 
-def _open_stream(formatter, namespace, repository, tag, synthetic_image_id, image_json, repo_image):
+def _open_stream(formatter, namespace, repository, tag, synthetic_image_id, image_json, repo_image,
+                 handlers):
   store = Storage(app)
 
   # For performance reasons, we load the full image list here, cache it, then disconnect from
@@ -55,10 +59,13 @@ def _open_stream(formatter, namespace, repository, tag, synthetic_image_id, imag
   stream = formatter.build_stream(namespace, repository, tag, synthetic_image_id, image_json,
                                   get_next_image, get_next_layer, get_image_json)
 
+  for handler_fn in handlers:
+    stream = wrap_with_handler(stream, handler_fn)
+
   return stream.read
 
 
-def _sign_sythentic_image(verb, linked_storage_uuid, queue_file):
+def _sign_synthetic_image(verb, linked_storage_uuid, queue_file):
   signature = None
   try:
     signature = signer.detached_sign(queue_file)
@@ -200,12 +207,20 @@ def _repo_verb(namespace, repository, tag, verb, formatter, sign=False, checker=
     # Close any existing DB connection once the process has exited.
     database.close_db_filter(None)
 
+  hasher = PieceHasher(app.config['TORRENT_PIECE_SIZE'])
+  def _store_metadata_and_cleanup():
+    with database.UseThenDisconnect(app.config):
+      model.storage.save_torrent_info(derived, app.config['TORRENT_PIECE_SIZE'],
+                                      hasher.final_piece_hashes())
+
   # Create a queue process to generate the data. The queue files will read from the process
   # and send the results to the client and storage.
-  args = (formatter, namespace, repository, tag, synthetic_image_id, image_json, repo_image)
+  handlers = [hasher.update]
+  args = (formatter, namespace, repository, tag, synthetic_image_id, image_json, repo_image,
+          handlers)
   queue_process = QueueProcess(_open_stream,
                                8 * 1024, 10 * 1024 * 1024, # 8K/10M chunk/max
-                               args, finished=_cleanup)
+                               args, finished=_store_metadata_and_cleanup)
 
   client_queue_file = QueueFile(queue_process.create_queue(), 'client')
   storage_queue_file = QueueFile(queue_process.create_queue(), 'storage')
@@ -224,7 +239,7 @@ def _repo_verb(namespace, repository, tag, verb, formatter, sign=False, checker=
 
   if sign and signer.name:
     signing_args = (verb, derived.uuid, signing_queue_file)
-    QueueProcess.run_process(_sign_sythentic_image, signing_args, finished=_cleanup)
+    QueueProcess.run_process(_sign_synthetic_image, signing_args, finished=_cleanup)
 
   # Close the database handle here for this process before we send the long download.
   database.close_db_filter(None)
diff --git a/requirements.txt b/requirements.txt
index cc98dad31..d6323dc2b 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -4,7 +4,7 @@ APScheduler==3.0.3
 autobahn==0.9.3-3
 Babel==1.3
 beautifulsoup4==4.4.0
-bintrees==2.0.2
+bintrees==2.0.3
 blinker==1.3
 boto==2.38.0
 cachetools==1.0.3
diff --git a/test/data/test.db b/test/data/test.db
index 16a8c3581..e26e28a9b 100644
Binary files a/test/data/test.db and b/test/data/test.db differ
diff --git a/util/registry/torrent.py b/util/registry/torrent.py
index 0cc9cb349..18ff45a9b 100644
--- a/util/registry/torrent.py
+++ b/util/registry/torrent.py
@@ -80,3 +80,6 @@ class PieceHasher(object):
   @property
   def hash_fragment(self):
     return self._hash_fragment
+
+  def final_piece_hashes(self):
+    return self._piece_hashes + self._hash_fragment.digest()