345.7 GB Transfer ANL->NERSC

1
345.7 GB Transfer ANL-gtNERSC

• Transfer between ANL(dg0n1.mcs.anl.gov) NERSC
  (swift.nersc.gov)
• Total 1260 files, B04.16.atm collection
  (1200x287.86MB) 60 files from uvcloud_csm
  collection (3.68MB each). Started 01/20/02 at
  1014 am
• There were 5 files transferring simultaneously
  all the time
• Since NERSCs server did not support restart
  markers and DCAU, we could not obtain performance
  data, modify the TCP buffer size, or change the
  number of TCP streams.
• 25 minutes after transfer was started NFS went
  down )- Everything got killed since its been
  started from the remote machine
• The next day RFT was restarted (23 hours later,
  at 931am 1/21/02) and the transfer continued
  (restarted where it left off)
• One more NFS outage for 20 minutes (1/21/02,
  320pm). This time both the server and the client
  survived the crash and resumed the transfer on
  its own after NFS came back.
• Some problems with dg0n1.mcs.anl.gov the
  finished ftpd processes were still running and
  taking most of the swap space (256MB!). After
  they were killed explicitly by root the
  transfers didnt resume. They couldve been
  resumed only one-by-one by hand (from the GUI)
• 1257 were transferred total. 3 never started, 7
  partially transferred (were cancelled by me)

2
RFT Transfer ANL-gtISI

• Total number of files transferred 70
• Each file was 287.86 MB (B04.16.atm collection)
• Took about 4 hours to transfer (- couple of
  minutes)
• 2 files transferred concurrently all the time
• At various times I modified the number of TCP
  streams (at runtime) and saw increased
  performance
• No network or other problems affected the
  transfer