Computing at BCB

USEFUL THINGS TO REMEMBER
Your email address is: [email protected]
You can also use [email protected] as an easier way to
remember your email address.
To access BCB email: https://roaming.dfci.harvard.edu
To access Partners email : https://www.partners.org/email
Web site of the department: http://bcb.dfci.harvard.edu
(for staff login use biostats/B10stat$)
Computing site of the department: http://bcbcomputing.dfci.harvard.edu
To manage your Partners password: https://myprofile.partners.org
Institute’s web site: https://dfcionline.org
Wireless network: phswifi3 (Use Partners userid/password to authenticate)
For any computing questions/concerns/issues: [email protected]
Off-hour support: (only critical issues) call 617-632-3060
Email
Settings for all mail clients (PC, Mac, Linux, Android, iOS)
Reading email (IMAP Server): imap.dfci.harvard.edu port 993 using SSL
Sending email (SMTP server): pascal.dfci.harvard.edu port 465 using SSL
(For all the above use BCB userid/password to authenticate)
Useful thunderbird add-ons
Accessing computing resources from home
There are 3 main options, depending on what kind of resources you need to
connect to.
GoToMyPC If you mainly use your desktop computer for work without
having the need to use the departmental servers, and your desktop is either a
PC or a Mac then you need to apply for a GoToMyPC account. When you
connect to your Mac or PC, what you see is a real-time image of your
computer’s screen; you can work with your files, and programs from anywhere
just as if you were at your desk. To apply call helpdesk at 2-3399 and open a
ticket to request a GoToMyPC account. You will receive an email with
additional instructions.
VPN If you only have one device (laptop) to work with both at work and at
home then you need a VPN account. A Virtual Private Network (VPN) is a
network technology that creates a secure network connection over a public
network such as the Internet extending connectivity between your laptop and
the DFCI networks; your laptop “feels” that is inside Partners space and you
do not have to change anything to connect to DFCI resources. To apply for a
VPN account, go to https://vpnrequests.partners.org/
ssh ssh is the most flexible method, but requires knowledge of how it works
and it’s only recommended for advanced users!
To be able to access a login machine from outside Partners you need to use ssh
version 2. For example, ssh ada.dfci.harvard.edu should log you in to ada
To be able to access graphical applications, you need to enable “X tunneling”
in your ssh client and also have an Xserver started in passive mode on your
home .
Free PC ssh client at http://www.putty.org/
Free PC ssh client/server and other UNIX utilities (find, grep,X,…) can be
found at https://www.cygwin.com/
Download and run the setup.exe, which will be used to choose all the utilities
to download. Make sure that you choose all the defaults and everything under
openssh. For an X server make sure you choose from the X11 packages the
xorg-x11-base along with any dependencies.
All Macs have ssh. Recent versions of the OS do not have an X server installed.
Using the Finder, go to Applications - Utilities - X11 and follow the
instructions to download and install Xquartz.
GoToMyPC and VPN are mutually exclusive: You can only get one of the two!
Logging in to a UNIX host
Currently we have the following login machines
ada
noah
leo
orion
santiam
sphinx
mys
monet
jaws
crunch
LINUX 16Gb
LINUX 384Gb
LINUX 384Gb
LINUX 36Gb
LINUX 24Gb
SOLARIS
SOLARIS
SOLARIS
SOLARIS
SOLARIS
8cores
32cores
32cores
16cores
16cores
(Xeon
(Xeon
(Xeon
(Xeon
(Xeon
E5-2603 1.8GHz)
E5-2650 2.5MHz)
E5-2650 2.0MHz) account
E5620 2.4Ghz)
E5620 2.4Ghz) IB
If you have a desktop PC, you need to configure your Xserver software
(exceed) to be able to run graphical applications. Make sure that exceed runs in
“Passive mode”, the setting can be found under “communication settings.”
UNIX The core operating system consisting of the kernel (computer
resources like cpu, memory, io devices) and system calls
Shell: command line interpreter (sh, csh, tcsh, bash, ksh)
Special characters in UNIX : <RETURN>, <ESCAPE>, <SPACE>,
<TAB>, #, “, ‘, `, /, \, &, *
Wildcards: * (0 or more), ? (one), [abe], [1-4]
Shell invocation: .login, .cshrc, /usr/skel/cshrc
Shell variables: PATH, MANPATH, HOME, USER, PAGER, DISPLAY (do
not ever set)
Use printenv, or ‘echo $VAR to display the value of a variable (echo $PATH)
Useful commands: man, ls, which, cp, mv, rm
Other useful commands: df, top, nice, kill, {ctrl-z,fg/bg, jobs, kill%},
redirection (<,>,|), gzip (zcat), tar, find, grep, cat, sort (uniq), cut, paste, awk
Controlling file access
file attributes : Read,Write,eXecute (rwx, r=4,w=2,x=1)
file agents: owner, group, world
By setting the attributes for every agent we completely control the access to a
file (dir)
“chmod N1N2N3 file” where N1 controls the attributes of the owner, N2 that of
the group, and N3 of everyone else.
High Performance Cluster
We recently merged the department cluster with the Institute’s research
computing cluster to form one cluster
To apply for a cluster account :
http://bcb.dfci.harvard.edu/cluster/newuserrequest.html
To open a support case: http://bcb.dfci.harvard.edu/hpc/
To access the cluster, ssh from within Partners network to one of the 2 submit
hosts using your Partners credentials.
Submit hosts : rcsgc-s1,2
rcsgc11-32 LINUX 57Gb 10 cores each
brsgx1-24 LINUX 48Gb 6 cores (Inter Xeon X5650 2.67GHz) each
rcapps2 LINUX 528Gb 32 cores (AMD Opteron 6328) each
Storage
Please store your files in your /bcb/groupname/username folder
Do not use your home directory for storing files other than login files. Once
the 200MB is full you will not be able to log in.
The cluster is using SGE as the cluster queuing software.
There are 3 different queues for program submission:
alllong: Jobs take more than 4 hours to complete
allmedium: Jobs take between 2 and 4 hours to complete
allshort: Jobs take less that 2 hours to complete
Sample Script:
#!/bin/bash
#$ -cwd
#$ -N ProgramName
#$ -o /bcb/groupfoldername/username/myfile.out
#$ -e /bcb/groupfoldername/username/myfile.error
/bin/date
How to submit a job
Copy the simple script example to a file
Submit the file with qsub myprog
How to check all job status
qstat to check all of your jobs
qstat -u "*" to check the status of the cluster
States: qw(waiting in the queue), t(transferring to a node, about to start),
r(running), h(held back by user), E(error)
How to get detailed information on a job during a run
qstat -j jobid, where jobid is showed when you check on your job status above
How to find out why your job failed
Examine /bcb/groupfoldername/username/myfile.out and
/bcb/groupfoldername/username/myfile.error to get more information
How to delete a job
qdel jobid
Other Productivity Tools






Dropbox
slack
skype
Keepass
Evernote
Xmarks (browser addon)
Other Services
Out-of-office calendar
Scan to pdf
ftp server (ftp.dfci.harvard.edu, supports anonymous ftp)
Rstudio server: http://rstudio.dfci.harvard.edu:8787
(login with BCB credentials)
RStudio is a free and open source integrated development environment for R.
It includes a console, syntax-highlighting editor that supports direct code
execution, as well as tools for plotting, history, debugging and workspace
management.
Terminal Server: arcem.dfci.harvard.edu
Windows server that currently has :
Office 2010
Adobe Acrobat Pro 9
Adobe Illustrator
Adobe Photoshop
MiKTEX 2.9
Stata 13
nQuery 3.0
EAST 6 and 5.4
Cytel Studio 11
logXact, statXact
GraphPAD Prism 6
Multc Lean Desktop 2.1
To access it please follow instructions from
http://bcbcomputing.dfci.harvard.edu/index.php/desktopcomputing/windows-remote-desktop
GPU Computing
We recently acquired our first GPU server.
CPUs and GPUs have significantly different architectures that make them
better suited to different tasks. A GPU can handle large amounts of data in
many streams, performing relatively simple operations on them, but is ill-suited
to heavy or complex processing on a single or few streams of data. A CPU is
much faster on a per-core basis (in terms of instructions per second) and can
perform complex operations on a single or few streams of data more easily, but
cannot efficiently handle many streams simultaneously.
As a result, GPUs are not suited to handle tasks that do not significantly benefit
from or cannot be parallelized, including many common consumer applications
such as word processors. Furthermore, GPUs use a fundamentally different
architecture; one would have to program an application specifically for a GPU
for it to work, and significantly different techniques are required to program
GPUs. For more information on the techniques needed, look up NVIDIA's
parallel computing language CUDA which has hooks for C, C++, and fortran.
Login to the server is by invitation only  (at least in the beginning !)