An Introduction to Linux and HPC

• Specialist part of IT Services.

Contact Details

• Based in Devonshire House (University Precinct, next to RBS)
  • Informal visitors welcome (we have good coffee machine), but
  • swipecard access. . .
• Email: rcs@manchester.ac.uk
• http://www.rcs.manchester.ac.uk
• http://wiki.rcs.manchester.ac.uk/community

What is Research Computing?

• Computing to support research! Examples:
  • running complex simulations;
  • performing vast parameter searches.

High Throughput Computing (HTC)

• Large amounts of comp. power over a "long" time:
  • Running long jobs!
  • Running the same experiment many (1000s) times, with different inputs.

High Performance Computing (HPC)

• Large amounts of comp. power over a "short" time:
  • many CPUs simultaneously to run complex models quicker.
• Many compute-nodes' RAM simultaneously to handle very big jobs.

Data Analysis and Visualization

• Getting the information out of the vast quantities of data.

Free Stuff!

Provision of resources:

• Horace, Man1, Man2, Mace01, Redqueen. . .
• Condor pools; NW-Grid, NGS.

Administration of HPC/HTC Clusters:

• Administer and support University, NW-Grid, NGS and some school and research group HPC clusters.

Support and Training:

• Documentation — Web and Wiki.
• Courses!
• Usage of HPC/HTC (Inc. Condor) clusters,
• application support,

In-depth support and collaborations

Free dedicated short-term help

• Advice on parallelisation of code, or
• advanced use of HTC (inc. Condor).

More in-depth help and collaborations

• Optimising code/models: scoping, estimate, coding — dedicated resources may require funding.
• Example: one year's dedicated effort extracting maximumum performance. Named resource/researchers on RCUK/EU etc. grants.

Introduction to Condor:

• CPU-cycle scavenging and HPC cluster backfill;
• Web pages; details and on-line booking.

Introduction to OpenMP:

• details and on-line booking.

Introduction to MPI:

• details and on-line booking.

Introduction to LaTeX:

• document preparation system;
• Web pages; details and on-line booking.

Basic Linux Stuff

• Necessary so you can get your research done!
• GUI stuff; command-line power; using Linux and MSW simultaneously on the same machine; more. . .

How-Tos

Give context — divert to required Linux topics as required:

• How do I run a job on an HPC cluster?
• How can I get my results faster?
• How can I work away from my office?
• How can I backup my stuff?

Flexible and Future

• Something particular you want covered? Ask. . .

Linux?

What is Linux?

Linux and its accomplices are:

• an operating system;
• graphical user interfaces (GUIs);
• applications software;

More — Linux "newbie" sites:

• Ultimate Linux Newbie Guide
• Linux Newbie Administrator Guide
• What is Linux?, from Linux Format magazine.
• Linux Advocacy from get GNU/Linux: What is Linux? and Switching to Linux.

Why on Earth should I use Linux?

• Unix/Linux developed over many years by researchers for researchers — ideal environment for computational research!
• Almost all HPC facilities have Linux as the OS.
• Easier to collaborate and share resources with Linux:
  • SSH, X11;
  • multiple simultaneous users — desktop MS Win crippled. . .
• Completely free (in both senses) — plus masses of free apps.
• Linux is a very flexible system:
  • choice of distros and. . .
  • . . .of user-interface (GUI): GNOME vs KDE vs XFCE vs LXDE. . .
  • GUI and very powerful commandline;
  • open-source.

In this module we describe:

• user-interfaces (GUIs);
• the commandline — BASH — and scripts;
• file-handling, permissions, editors;
• process-management and stdout/in/err;
• using MS Windows and Linux at the same time;
• distros;
• getting Linux-related help at UoM.

The Edit, Compile, Link Cycle

• Important for anyone who handles Fortran or C code.
• Turning Fortran 77/90/95 or C/Obj-C/C++ code into executable programmes.

• Can't always use a commercial application or someone else's programme — may need to write your own.
• Understanding this. . .

      error while loading shared libraries: libgfortran.so.1: 
      cannot open shared object file: No such file or directory

  . . .and what to do about it!

What:

In this module we look at:

• Basic, necessary steps — editing and compiling.
• The hidden step — linking.
• Things that can and do go wrong!
• [[Static vs. dynamic linking.]]

Not Matlab, Perl, Python, etc.

HPC

What is HPC?

• High-Performance Computing.
• The application of "supercomputers" to computational problems that are
  • too big for desktops/laptops, or
  • would take too long.
• A "supercomputer" has many CPUs/cores (hundreds, thousands. . .); a desktop computer has only 1, 2 or 4.

I just run things on my desktop machine, can't I?

Yes

• The CPUs in a desktop are similar to those in a "supercomputer"

But

• Do you want to run 100%-CPU jobs on your desktop machine?
• Would you like to:
  • Run a dozen or more jobs at once?
  • Run a (parallel) job on 64 or 128 CPUs/cores at once?
  • Use 32 GB RAM?
• What happens if the cleaner unplugs your desktop 90% through a month-long computation?

How?

In this module we describe the nature of HPC clusters, and how to:

• remote, secure access to HPC machine (using SSH);
• copy files to/from remote HPC machine (using SCP/SFTP);
• understand nature and structure of HPC machine;
• run programs/jobs and share resources with other users — submit job to batch/queue system.

HPC Facilities and Services Available at UoM

In this module we outline:

• Man2 and NW-Grid
• Mace01
• Bennu
• Condor pools
• Horace
• National Grid Service (NGS)

How can I get results faster?

Optimisation

• make your program more efficient

High Throughput Computing (HTC)

• run lots of copies of your program

High Performance Computing (HPC)/Parallelisation

• run (one instance of) your program on many CPUs at the same time

Optimisation, Distribution, Parallelisation — How?

In this module we survey specifics:

• Optimise your code:
  • compiler options
  • use available numerical libraries
  • optimise memory use
  • profiling
• High throughput computing:
  • SGE Arrays
  • Condor
• Parallelisation/HPC:
  • OpenMP
  • MPI

Remote Working

In this section we look at:

• connecting to remote machines and file-transfer;
• using GUIs remotely;
• address the questions, "How can I work away from my office? Or using my laptop?"

Remote Working Tools

These the the tools of the trade:

• SSH (and SCP/SFTP)
• X-Windows (aka X11)
• GNU Screen
• VNC.

Working From Home

I want to:

• work from home;
• connect to sessions I was running at work, from home;
• do all my work on my laptop, at home, work and elsewhere.

In this module we outline how to do this (using the tools listed above).

Backups!

Accidental Deletion

• Everyone makes mistakes;
• very difficult to get files back once deleted.

Catastrophic Hardware/Software Failure

• It happens! ("It'll never happen to me. . .")
• RAID helps — but is not a panacea.

Three Catastrophes

Usto-Oran

• RAID redundancy of disks — but two disks failed!
• Used as archive, but no backups!

---

Research Group at UMIST (Some years ago!)

• All data and source code on one Solaris server.
• All backed up nightly — but no one checked the backups!

---

Thesis MS Word File Corruption

• On a more personal level. . .
• File corrupted; hard copy only; type it all in again!

In this module we describe:

• Why you should do backups!
• Different types of backups.
• Some practical methods of making backups of personal desktop and laptop machines.

Help and Community

• Where can I get help?
  • Two sources. . .

Very different from (mainstream) IT Services. . .

• Mainstream IT Services deals with "commodity" services:
  • word-processing, email. . .
  • 1000 customers ~ 1 IT staff
• Research computing is fundamentally different:
  • postgrads must learn a lot in a short time;
  • much wider and deeper support required;
  • 30 customers ~ 1 IT staff?

In this module we describe:

Formal:

• School and faculty IS teams;
• IT Services virtual helpdesk;
• RCS!

Informal:

• Wiki
• Email list

The End (Almost)

• rcs@manchester.ac.uk
• Research Computing Website
  • Service information and support for Research Computing at The University of Manchester
  • http://www.rcs.manchester.ac.uk
• Community Wiki
  • For everyboyd to share experience and tips
  • http://wiki.rcs.manchester.ac.uk/community

More on Grid Stuff

Parallel Computation

• OpenMP
• MPI (MPICH, OpenMPI)
• Other RCS courses