BCE is currently available in two forms, both documented below.
Before you begin, the download and install process requires at least 14GB free (only 10 Gb if you access the .ova file from a USB key or separate hard drive) on your hard drive or you are likely to encounter problems installing and running BCE due to out of disk space errors. If you do not have 14GB free, then visit the help page to inquire about other options.
The current recommended release is 2015 Fall.
1) Download and install Oracle VirtualBox for:
Mac OS X
2) Download the BCE Virtual Machine (VM) Appliance file (.ova):
Fall 2015 (BCE-2015-fall.ova) (3.7GB)
Older versions of BCE are available from the BCE Downloads page.
3) Open the VirtualBox application you installed in Step 1.
4) From the menu select File > Import appliance and locate the BCE .ova file you downloaded in Step 2.
5) Wait a few minutes while the BCE .ova file is imported as a VirtualBox VM.
6) Once importing finishes, it will appear in the Powered Off state in the Oracle VM VirtualBox Manager. Select it, then click the Start button.
7) This will start a virtual Linux computer within your own machine. After a few seconds you should see black screen and then the VM’s desktop.
Installation trouble? See our help page.
We provide screencasts demonstrating installation based on these instructions.
We provide both BCE-2015-fall and BCE-2015-spring versions for EC2. In general we recommend the newer BCE-2015-fall version, but for anyone wishing to use BCE with StarCluster, you should use BCE-2015-spring as BCE-2015-fall is incompatible with StarCluster.
ssh -i ~/.ssh/your_ssh_key email@example.com the IP address after the ‘@’ will be given in the instructions.
CfnCluster allows you to easily start up multiple EC2 instances to form your own Linux cluster. You can then login to the master node of the cluster and submit jobs via a scheduler that will manage those jobs. One of the nice things about CfnCluster is that if you request more cores or nodes than are available, additional EC2 instances will be added to your cluster dynamically. And unused instances will be terminated to save you money.
easy_install. You could do this in your VirtualBox BCE VM if you like.
cfncluster create myclusterto start your virtual cluster. Note that the AMI is only available in the us-west-2 region.
MasterPublicIPprinted to the screen at the end of the CfnCluster startup process and with the flag
-i ~/.ssh/ec2.pemwhere you should use the name of your private key in place of
ec2.pem. Now you’re ready to run your parallel computation.
squeuegives an error, then run
sudo /etc/init.d/slurm start.
srun --nodes=2 --ntasks=4 --pty /bin/bash(for an interactive job) or
sbatch --nodes=2 --ntasks=4 myjob.shfor a batch job.
We’ve created a version of BCE with software for GPU computation, in particular CUDA, PyCUDA, RCUDA, and MAGMA. Simply choose the “BCE-2015-fall-gpu” public image (ami-5154b331) in the us-west-2 (Oregon) region.
This use case is deprecated; we recommend use of BCE with CfnCluster (see above).
This will only work with BCE-2015-spring as StarCluster is not compatible with the newer version of Ubuntu used in BCE-2015-fall.
starcluster sshmaster -u ubuntu mycluster. Now you’re ready to run your parallel computation.
Keep your eyes out here for future documentation for starting EC2 instances and clusters via Vagrant and boto.