diff --git a/cephsum/actions.py b/cephsum/actions.py
index 184f291..405875f 100644
--- a/cephsum/actions.py
+++ b/cephsum/actions.py
@@ -13,17 +13,17 @@
 
 
 
-def get_from_metatdata(ioctx, path, xattr_name = "XrdCks.adler32"):
+def get_from_metatdata(ioctx, path, xattr_name = "XrdCks.adler32", use_multithreading=False, mt_workers=1):
     """Try to get checksum info from metadata only.
     """
     xrdcks = cephtools.cks_from_metadata(ioctx,path,xattr_name)
     logging.info(xrdcks)
     return xrdcks  # returns None if not existing
 
-def get_from_file(ioctx, path, readsize):
+def get_from_file(ioctx, path, readsize,use_multithreading=False, mt_workers=1):
     """Try to get checksum info from file only.
     """
-    xrdcks = cephtools.cks_from_file(ioctx,path,readsize)
+    xrdcks = cephtools.cks_from_file(ioctx,path,readsize, use_multithreading=use_multithreading, max_workers=mt_workers)
     logging.info(xrdcks)
     return xrdcks  # returns None if not existing
 
@@ -41,7 +41,7 @@ def get_checksum(ioctx, path, readsize, xattr_name = "XrdCks.adler32"):
 
 
 
-def inget(ioctx, path, readsize, xattr_name = "XrdCks.adler32",rewriteto_littleendian=True):
+def inget(ioctx, path, readsize, xattr_name = "XrdCks.adler32",rewriteto_littleendian=True, use_multithreading=False, mt_workers=1):
     """Return a checksum; if in metadata, just return that. If no metadata, obtain from file and store metadata.
     If rewriteto_littleendian and metadata was stored in big endian; write it back as little endian
     """
@@ -57,7 +57,7 @@ def inget(ioctx, path, readsize, xattr_name = "XrdCks.adler32",rewriteto_littlee
 
     if xrdcks is None:
         source = 'file'
-        xrdcks = cephtools.cks_from_file(ioctx, path,readsize)
+        xrdcks = cephtools.cks_from_file(ioctx, path,readsize, use_multithreading=use_multithreading, max_workers=mt_workers)
         if xrdcks is None:
             logging.warning(f"No checksum possible for {path} from file")
             return None
@@ -73,7 +73,7 @@ def inget(ioctx, path, readsize, xattr_name = "XrdCks.adler32",rewriteto_littlee
     return xrdcks 
 
 
-def verify(ioctx, path, readsize, xattr_name = "XrdCks.adler32", force_fileread=False):
+def verify(ioctx, path, readsize, xattr_name = "XrdCks.adler32", force_fileread=False, use_multithreading=False, mt_workers=1):
     """compare the stored checksum against the file-computed value.
     If no stored metadata, still compute file (if requested), but compare as false.
     """
@@ -85,7 +85,7 @@ def verify(ioctx, path, readsize, xattr_name = "XrdCks.adler32", force_fileread=
     if xrdcks_stored is None and not force_fileread:
         xrdcks_file = None
     else:
-        xrdcks_file = cephtools.cks_from_file(ioctx, path,readsize)
+        xrdcks_file = cephtools.cks_from_file(ioctx, path,readsize, use_multithreading=use_multithreading, max_workers=mt_workers)
 
     if xrdcks_stored is None:
         matching = False
diff --git a/cephsum/adler32.py b/cephsum/adler32.py
index c6a95b5..a5683fb 100644
--- a/cephsum/adler32.py
+++ b/cephsum/adler32.py
@@ -44,14 +44,18 @@ def calc_checksum(self,buffer):
         value  = 1 # initilising value
         bytes_read = 0
         counter = 0
-        for buf in buffer:
-            # need to consider intra-file chunks
-            value = zlib.adler32( buf, value)
-            bytes_read += len(buf)
-            counter += 1
-            if self.log_each_step:
-                logging.debug('%s: %s %s %s' % (self.name, self.adler32_inttohex(value), len(buf), bytes_read) )
-        
+        try:
+            for buf in buffer:
+                # need to consider intra-file chunks
+                value = zlib.adler32( buf, value)
+                bytes_read += len(buf)
+                counter += 1
+                if self.log_each_step:
+                    logging.debug('%s: %s %s %s' % (self.name, self.adler32_inttohex(value), len(buf), bytes_read) )
+        except StopIteration:
+            logging.debug(f"Got Stopiteration after: {bytes_read} bytes")
+
+
         self.value      = self.adler32_inttohex(value)
         self.bytes_read = bytes_read
         self.number_buffers = counter
diff --git a/cephsum/cephsum.py b/cephsum/cephsum.py
index 6e6fcd6..2358907 100644
--- a/cephsum/cephsum.py
+++ b/cephsum/cephsum.py
@@ -88,6 +88,10 @@ def convert_path(path, xmlfile=None):
     parser.add_argument('--cephuser',default='client.xrootd', dest='ceph_user', 
                         help='ceph user name for the client keyring')
 
+    parser.add_argument('--mt',default=False,help='Use multithreading',action='store_true')
+    parser.add_argument('--workers',default=1,type=int,help='If multithreading, specify the max number of workers')
+    
+
 
     # actual path to use, as a positional argument; only one allowed
     parser.add_argument('path', nargs=1)
@@ -136,6 +140,10 @@ def convert_path(path, xmlfile=None):
     if args.action == 'check' and source_checksum is None:
         raise ValueError("Need --type|-C in form adler32:<checksum> for 'check' action with source checksum value")
 
+
+    use_multithreading = args.mt
+    mt_workers = args.workers
+
     xrdcks,adler = None,None
     timestart = datetime.now()
 
@@ -146,15 +154,15 @@ def convert_path(path, xmlfile=None):
     try:
         with cluster.open_ioctx(pool) as ioctx:
             if args.action in ['inget','check']:
-                xrdcks = actions.inget(ioctx,path,readsize,xattr_name)
+                xrdcks = actions.inget(ioctx,path,readsize,xattr_name, use_multithreading=use_multithreading, mt_workers=mt_workers)
             elif args.action == 'verify':
-                xrdcks = actions.verify(ioctx,path,readsize,xattr_name)
+                xrdcks = actions.verify(ioctx,path,readsize,xattr_name, use_multithreading=use_multithreading, mt_workers=mt_workers)
             elif args.action == 'get':
-                xrdcks = actions.get_checksum(ioctx,path,readsize, xattr_name)
+                xrdcks = actions.get_checksum(ioctx,path,readsize, xattr_name, use_multithreading=use_multithreading, mt_workers=mt_workers)
             elif args.action == 'metaonly':
                 xrdcks = actions.get_from_metatdata(ioctx,path,xattr_name)
             elif args.action == 'fileonly':
-                xrdcks = actions.get_from_file(ioctx,path, readsize)    
+                xrdcks = actions.get_from_file(ioctx,path, readsize, use_multithreading=use_multithreading, mt_workers=mt_workers)    
             else:
                 logging.warning(f'Action {args.action} is not implemented')
                 raise NotImplementedError(f'Action {args.action} is not implemented')
diff --git a/cephsum/cephtools.py b/cephsum/cephtools.py
index 155b364..d0b0e31 100644
--- a/cephsum/cephtools.py
+++ b/cephsum/cephtools.py
@@ -5,6 +5,10 @@
 import XrdCks,adler32
 import rados
 
+import queue
+from concurrent.futures import ThreadPoolExecutor, CancelledError, TimeoutError
+
+
 chunk0=f'.{0:016x}' # Chunks are hex valued
 nZeros=16
 
@@ -53,9 +57,9 @@ def get_chunks(ioctx,path, stripe_count=None):
     while True:
         try:
             oid = path+f'.{counter:016x}'  # chunks are hex encoded
-            ioctx.stat(oid)
+            size, ts = ioctx.stat(oid)
             #logging.debug(oid)
-            yield oid
+            yield oid, size
         except rados.ObjectNotFound:
             raise StopIteration
         counter += 1
@@ -105,12 +109,74 @@ def read_file_btyes(ioctx, path, stripe_size_bytes=None, number_of_stripes=None,
     if stripe_size_bytes is None, will use READSIZE and read each stripe for all data.
     if stripe_size_bytes is given, will assume each chunk is the given size.
     """
-    for oid in get_chunks(ioctx, path, number_of_stripes):
+    for oid, _ in get_chunks(ioctx, path, number_of_stripes):
         for buffer in read_oid_bytes(ioctx, oid, stripe_size_bytes, readsize=readsize):
             yield buffer
     # Sanity stop statement at end.
     raise StopIteration
 
+stop_all_workers = False
+
+def readio_mt(ioctx, oid: str, extent: slice) -> bytearray:
+    global stop_all_workers
+    if stop_all_workers: # cancellation signal
+        logging.debug(f'Got cancellation for thread: {oid}: {extent}')
+        return oid, extent, None
+    buf = ioctx.read(oid, length = extent.stop - extent.start, offset = extent.start)
+    return oid,extent, buf
+
+def read_file_btyes_multithreaded(ioctx, path, stripe_size_bytes=None, number_of_stripes=None,readsize=64*1024*1024, max_workers=1):
+    """Yield all bytes in a file, looping over chunks, and then bytes with the file using a multi-threaded threadpool
+
+    if stripe_size_bytes is None, will use READSIZE and read each stripe for all data.
+    if stripe_size_bytes is given, will assume each chunk is the given size.
+    """
+    assert stripe_size_bytes is not None # must know the number of bytes per chunk here
+    assert number_of_stripes is not None # must know the number of chunks here
+    assert readsize <= stripe_size_bytes # only read at most the number of bytes in a chunk
+    assert stripe_size_bytes % readsize == 0 # 
+    
+    futures = queue.deque()
+    with ThreadPoolExecutor(max_workers=max_workers) as executor:
+        # load up the deque will all the slices
+        # get_chunks does a stat on each file to check for existence (what if the file is deleted mid read?)
+        for oid, chunk_size in get_chunks(ioctx, path, number_of_stripes):
+            # create a iterable of slices corresponding to the reads within each chunk
+            # only allow to read to the length of the chunk
+            for extent in map(lambda offset: slice(offset,min(chunk_size, offset+readsize),None),  range(0, min(chunk_size, stripe_size_bytes), readsize)):
+                futures.append(executor.submit(readio_mt, ioctx, oid, extent)  )
+
+        # wait for the ordered futures to completed:
+        global stop_all_workers
+        while True:
+            try:
+                f = futures.popleft()
+            except IndexError: # no more entries
+                break
+            try:
+                oid, extent, buf =  f.result(timeout=120)  # return the buffer of bytes
+                logging.debug(f"Result for {oid} {extent} {'None' if buf is None else len(buf)}")
+            except TimeoutError:
+                logging.error("Reached read timeout; aborting")
+                stop_all_workers = True
+                raise TimeoutError()
+
+
+            except CancelledError:
+                logging.debug(f'Got Cancelled error; aborted the remaining tasks')
+
+            if buf is None or len(buf) == 0:
+                # a problem or were at the end; cancel the other 
+                # global stop_all_workers
+                stop_all_workers = True
+                logging.debug(f"Got 0 len array; stop all workers")
+
+                continue # don't yield below, but make sure we pop all the futures
+            yield buf
+
+    #raise StopIteration
+    ## Sanity stop statement at end.
+
 
 
 
@@ -234,7 +300,7 @@ def cks_write_metadata(ioctx, path, xattr_name, xattr_value, force_overwrite=Fal
 
 
 
-def cks_from_file(ioctx, path, readsize):
+def cks_from_file(ioctx, path, readsize, use_multithreading=False, max_workers=1):
     """Calculate checksum from path. Returns None or checksum object
     Raise error if not existing"""
 
@@ -257,7 +323,10 @@ def cks_from_file(ioctx, path, readsize):
 
     try:
         cks_alg = adler32.adler32('adler32')
-        cks_hex = cks_alg.calc_checksum( read_file_btyes(ioctx, path, rados_object_size, num_stripes,readsize) )
+        if use_multithreading:
+            cks_hex = cks_alg.calc_checksum( read_file_btyes_multithreaded(ioctx, path, rados_object_size, num_stripes,readsize, max_workers) )
+        else:
+            cks_hex = cks_alg.calc_checksum( read_file_btyes(ioctx, path, rados_object_size, num_stripes,readsize) )
         bytes_read = cks_alg.bytes_read
     except Exception as e:
         raise e