Page Comparison

...

Once you have an AWS account, you need to create an IAM user with the permissions to use S3 and Glacier. Please refer to the amazon documentation on how to do that.

Using other cloud providers

See the list of rclone-supported cloud providers for notes on how to set up providers other than AWS. In general, this is similar to the AWS setup shown below - just follow the instructions of the configuration tool.

Configure rclone to work with AWS

Rclone is available on the SCU nodes. The following example is on pascal.med.cornell.edu.

...

$ rclone copy -P amazon_store:rclone-tutorial ~/test-download
Transferred:   	      192M / 192 MBytes, 100%, 77.531 MBytes/s, ETA 0s
Errors:                 0
Checks:                 0 / 0, -
Transferred:            2 / 2, 100%
Elapsed time:        2.4s
$ ls ~/test-download/
test1.file  test2.file

Tuning rclone performance

By default, rclone is not optimized for our infrastructure. Increasing the maximum number of parallel transfers and the chunk size can increase transfer speed. This will however take more bandwidth and RAM, so depending on which node this is run on, the results will vary. The following flags should be used:

--bwlimit=0          # Do not limit bandwidth
--buffer-size=128M   # Buffer for each transfer
--checkers=32        # Run 32 checksum checkers in parallel
--transfers=32       # Run 32 transfers in parallel

Please be advised that the actual performance gain depends on both the source and destination system, as well as the current usage of those systems. Also, depending on the type of data transferred (many small files, or few large files?), results will vary. Use these parameters as a starting point for your individual fine tuning.

Use these parameters as follows:

$ rclone --bwlimit=0 --buffer-size=128M --checkers=32 --transfers=32 copy -P ~/local/source amazon_store:bucket-name

rclone browser - Graphical user interface

...