Other languages:

Expected availability: April 2018
Login node: niagara.computecanada.ca
Globus endpoint: TBA
System Status Page: https://wiki.scinet.utoronto.ca/wiki/index.php/System_Alerts

Niagara is a homogeneous cluster, owned by the University of Toronto and operated by SciNet, intended to enable large parallel jobs of 1040 cores and more. It was designed to optimize throughput of a range of scientific codes running at scale, energy efficiency, and network and storage performance and capacity.

The user experience on Niagara will be similar to that on Graham and Cedar, but slightly different. Specific instructions on how to use the Niagara cluster will be available in April 2018.

Niagara is an allocatable resource in the 2018 Resource Allocation Competition (RAC 2018), which comes into effect on April 4, 2018.

Niagara installation update at the SciNet User Group Meeting on February 14th, 2018

Niagara installation time-lag video

Niagara hardware specifications

1500 nodes, each with 40 Intel Skylake cores at 2.4GHz, for a total of 60,000 cores.
202 GB (188 GiB) of RAM per node.
EDR Infiniband network in a so-called 'Dragonfly+' topology.
6PB of scratch, 3PB of project space (parallel filesystem: IBM Spectrum Scale, formerly known as GPFS).
256 TB burst buffer (Excelero + IBM Spectrum Scale).
No local disks.
No GPUs.
Rpeak of 4.61 PF.
Rmax of 3.0 PF.
685 kW power consumption.

Attached storage systems

Home space 600TB total volume Parallel high-performance filesystem (IBM Spectrum Scale)	Location of home directories. Available as the `$HOME` environment variable. Each home directory has a small, fixed quota of 100GB. Not allocated, standard amount for each user. For larger storage requirements, use scratch or project. Has daily backup.
Scratch space 6PB total volume Parallel high-performance filesystem (IBM Spectrum Scale)	For active or temporary (`/scratch`) storage (~ 80 GB/s). Available as the `$SCRATCH` environment variable. Not allocated. Large fixed quota per user and per group path. Inactive data will be purged.
Burst buffer 256TB total volume Parallel extra high-performance filesystem (Excelero+IBM Spectrum Scale)	For active fast storage during a job (160GB/s, and very high IOPS). Not fully configured yet, but data will be purged very frequently (i.e. soon after a job has ended) and space on this storage tier will not be RAC allocatable.
Project space 3PB total volume. Parallel high-performance filesystem (IBM Spectrum Scale	Allocated via RAC. For active but low data turnover storage and relatively fixed datasets Available as the `$PROJECT` environment variable. quota set per project path. Backed up.
Archive Space 10PB total volume High Performance Storage System (IBM HPSS)	Allocated via RAC. Nearline accessible space intended for large datasets requiring offload from active filesystems. Available as the `$ARCHIVE` environment variable. Large fixed quota per group. Tape based backend (dual copy).

High-performance interconnect

The Niagara cluster has an EDR Infiniband network in a so-called 'Dragonfly+' topology, with four wings. Each wing of maximually 432 nodes (i.e., 17280) has 1-to-1 connections. Network traffic between wings is done through adaptive routing, which alleviates network congestion and yields an effective blocking of 2:1 between nodes of different wings.

Node characteristics

CPU: 2 sockets with 20 Intel Skylake cores (2.4GHz, AVX512), for a total of 40 cores per node
Computational performance: 3 TFlops theoretical peak.
Network connection: 100Gb/s EDR Dragonfly+
Memory: 202 GB (188 GiB) of RAM, i.e., a bit over 4GiB per core.
Local disk: none. GPUs/Accelerators: none.
Operating system: Linux CentOS 7

Scheduling

The Niagara cluster will use the Slurm scheduler to run jobs. The basic scheduling commands will therefore be similar to those for Cedar and Graham, with a few differences:

Scheduling will be by node only. This means jobs will always need to use multiples of 40 cores per job.
Asking for specific amounts of memory will not be necessary and is discouraged; all nodes have the same amount of memory (202GB/188GiB minus some operating system overhead).

Details, such as how to request burst buffer usage in jobs, are still being worked out.

Software

Module-based software stack.
Both the standard Compute Canada software stack as well as cluster-specific software tuned for Niagara will be available.
In contrast with Cedar and Graham, no modules will be loaded by default to prevent accidental conflicts in versions. There will be a simple mechanism to load the software stack that a user would see on Graham and Cedar.

Migration to Niagara

Migration for Existing Users of the GPC

Accounts, $HOME & $PROJECT of active GPC users transferred to Niagara (except dot-files in ~).
Data stored in $SCRATCH will not be transfered automatically.
Users are to clean up $SCRATCH on the GPC as much as possible (remember it's temporary data!). Then they can transfer what they need using datamover nodes. Let us know if you need help.
To enable this transfer, there will be a short period during which you can have access to Niagara as well as to the GPC storage resources. This period will end no later than May 9, 2018.

For Non-GPC Users

Those of you new to SciNet, but with 2018 RAC allocations on Niagara, will have your accounts created and ready for you to login.
New, non-RAC users: we are still working out the procedure to get access. If you can't wait, for now, you can follow the old route of requesting a SciNet Consortium Account on the CCDB site.

Using Niagara (Quickstart)

Logging in

As with all SciNet and CC (Compute Canada) compute systems, access to Niagara is via ssh (secure shell) only.

To access SciNet systems, first open a terminal window (e.g. MobaXTerm on Windows).

Then ssh into the Niagara login nodes with your CC credentials:

$ ssh -Y MYCCUSERNAME@niagara.scinet.utoronto.ca

or

$ ssh -Y MYCCUSERNAME@niagara.computecanada.ca

The Niagara login nodes are where you develop, edit, compile, prepare and submit jobs.
These login nodes are not part of the Niagara compute cluster, but have the same architecture, operating system, and software stack.
The optional -Y is needed to open windows from the Niagara command-line onto your local X server.
To run on Niagara's compute nodes, you must submit a batch job.

Storage Systems and Locations

Home and scratch

You have a home and scratch directory on the system, whose locations will be given by

$HOME=/home/g/groupname/myccusername

$SCRATCH=/scratch/g/groupname/myccusername

nia-login07:~$ pwd
/home/s/scinet/rzon

nia-login07:~$ cd $SCRATCH

nia-login07:rzon$ pwd
/scratch/s/scinet/rzon

Project location

Users from groups with a RAC allocation will also have a project directory.

$PROJECT=/project/g/groupname/myccusername

IMPORTANT: Future-proof your scripts

Use the environment variables (HOME, SCRATCH, PROJECT) instead of the actual paths! The paths may change in the future.

Storage Limits on Niagara

location	quota	block size	expiration time	backed up	on login	on compute
$HOME	100 GB	1 MB		yes	yes	read-only
$SCRATCH	25 TB	16 MB	2 months	no	yes	yes
$PROJECT	by group allocation	16 MB		yes	yes	yes
$ARCHIVE	by group allocation			dual-copy	no	no
$BBUFFER	?	1 MB	very short	no	?	?

Compute nodes do not have local storage.
Archive space is on HPSS.
Backup means a recent snapshot, not an achive of all data that ever was.
$BBUFFER stands for the Burst Buffer, a functionality that is still being set up. This will be a faster parallel storage tier for temporary data.

Moving data

Move amounts less than 10GB through the login nodes.

Only Niagara login nodes visible from outside SciNet.
Use scp or rsync to niagara.scinet.utoronto.ca or niagara.computecanada.ca (no difference).
This will time out for amounts larger than about 10GB.

Move amounts larger than 10GB through the datamover node.

From a Niagara login node, ssh to nia-datamover1.
Transfers must originate from this datamover.
The other side (e.g. your machine) must be reachable from the outside.
If you do this often, consider using Globus, a web-based tool for data transfer.

Moving data to HPSS/Archive/Nearline using the scheduler.

HPSS is a tape-based storage solution, and is SciNet's nearline a.k.a. archive facility.
Storage space on HPSS is allocated through the annual Compute Canada RAC allocation.

Software and Libraries

Modules

Once you are on one of the login nodes, what software is already installed?

Other than essentials, all installed software is made available using module commands.
These set environment variables (PATH, etc.)
Allows multiple, conflicting versions of a given package to be available.
module spider shows the available software.

nia-login07:~$ module spider
---------------------------------------------------
The following is a list of the modules currently av
---------------------------------------------------
  CCEnv: CCEnv

  NiaEnv: NiaEnv/2018a

  anaconda2: anaconda2/5.1.0

  anaconda3: anaconda3/5.1.0

  autotools: autotools/2017
    autoconf, automake, and libtool 

  boost: boost/1.66.0

  cfitsio: cfitsio/3.430

  cmake: cmake/3.10.2 cmake/3.10.3

  ...

module load <module-name>

use particular software
module purge

remove currently loaded modules
module spider

(or module spider <module-name>)

list available software packages
module avail

list loadable software packages
module list

list loaded modules

On Niagara, there are really two software stacks:

A Niagara software stack tuned and compiled for this machine. This stack is available by default, but if not, can be reloaded with
```
module load NiaEnv
```
The same software stack available on Compute Canada's General Purpose clusters Graham and Cedar, compiled (for now) for a previous generation of CPUs:
```
module load CCEnv
```
If you want the same default modules loaded as on Cedar and Graham, then afterwards also module load StdEnv.

Note: the *Env modules are sticky; remove them by --force.

Tips for loading software

We advise against loading modules in your .bashrc.

This could lead to very confusing behaviour under certain circumstances.
Instead, load modules by hand when needed, or by sourcing a separate script.
Load run-specific modules inside your job submission script.
Short names give default versions; e.g. intel → intel/2018.2.

It is usually better to be explicit about the versions, for future reproducibility.
Handy abbreviations:

        ml → module list
        ml NAME → module load NAME  # if NAME is an existing module
        ml X → module X

Modules sometimes require other modules to be loaded first.

Solve these dependencies by using module spider.

Module spider

Oddly named, the module subcommand spider is the search-and-advice facility for modules.

nia-login07:~$ module load openmpi
Lmod has detected the error:  These module(s) exist but cannot be loaded as requested: "openmpi"
   Try: "module spider openmpi" to see how to load the module(s).

nia-login07:~$ module spider openmpi
------------------------------------------------------------------------------------------------------
  openmpi:
------------------------------------------------------------------------------------------------------
     Versions:
        openmpi/2.1.3
        openmpi/3.0.1
        openmpi/3.1.0rc3

------------------------------------------------------------------------------------------------------
  For detailed information about a specific "openmpi" module (including how to load the modules) use
  the module s full name.
  For example:

     $ module spider openmpi/3.1.0rc3
------------------------------------------------------------------------------------------------------

nia-login07:~$ module spider openmpi/3.1.0rc3
------------------------------------------------------------------------------------------------------
  openmpi: openmpi/3.1.0rc3
------------------------------------------------------------------------------------------------------
    You will need to load all module(s) on any one of the lines below before the "openmpi/3.1.0rc3"
    module is available to load.

      NiaEnv/2018a  gcc/7.3.0
      NiaEnv/2018a  intel/2018.2

nia-login07:~$ module load NiaEnv/2018a  intel/2018.2   # note: NiaEnv is usually already loaded
nia-login07:~$ module load openmpi/3.1.0rc3

nia-login07:~$ module list
Currently Loaded Modules:
  1) NiaEnv/2018a (S)   2) intel/2018.2   3) openmpi/3.1.0.rc3

  Where:
   S:  Module is Sticky, requires --force to unload or purge

Can I Run Commercial Software?

Possibly, but you have to bring your own license for it.
SciNet and Compute Canada have an extremely large and broad user base of thousands of users, so we cannot provide licenses for everyone's favorite software.
Thus, the only commercial software installed on Niagara is software that can benefit everyone: Compilers, math libraries and debuggers.
That means no Matlab, Gaussian, IDL,
Open source alternatives like Octave, Python, R are available.
We are happy to help you to install commercial software for which you have a license.
In some cases, if you have a license, you can use software in the Compute Canada stack.

Compiling on Niagara: Example

nia-login07:~$ module list
Currently Loaded Modules:
  1) NiaEnv/2018a (S)
  Where:
   S:  Module is Sticky, requires --force to unload or purge

nia-login07:~$ module load intel/2018.2 gsl/2.4

nia-login07:~$ ls
main.c module.c

nia-login07:~$ icc -c -O3 -xHost -o main.o main.c
nia-login07:~$ icc -c -O3 -xHost -o module.o module.c
nia-login07:~$ icc  -o main module.o main.o -lgsl -mkl

nia-login07:~$ ./main

Testing

You really should test your code before you submit it to the cluster to know if your code is correct and what kind of resources you need.

Small test jobs can be run on the login nodes.

Rule of thumb: couple of minutes, taking at most about 1-2GB of memory, couple of cores.
You can run the the ddt debugger on the login nodes after module load ddt.
Short tests that do not fit on a login node, or for which you need a dedicated node, request an
interactive debug job with the salloc command
```
nia-login07:~$ salloc -pdebug --nodes N --time=1:00:00
```
where N is the number of nodes. The duration of your interactive debug session can be at most one hour, can use at most 4 nodes, and each user can only have one such session at a time.

Submitting jobs

Niagara uses SLURM as its job scheduler.
You submit jobs from a login node by passing a script to the sbatch command:
```
nia-login07:~$ sbatch jobscript.sh
```
This puts the job in the queue. It will run on the compute nodes in due course.
Jobs will run under their group's RRG allocation, or, if the group has none, under a RAS allocation (previously called `default' allocation).

Keep in mind:

Scheduling is by node, so in multiples of 40-cores.
Maximum walltime is 24 hours.
Jobs must write to your scratch or project directory (home is read-only on compute nodes).
Compute nodes have no internet access.

Download data you need beforehand on a login node.

Example submission script (OpenMP)

#!/bin/bash
#SBATCH --nodes=1
#SBATCH --cpus-per-task=40
#SBATCH --time=1:00:00
#SBATCH --job-name openmp_job
#SBATCH --output=openmp_output_%j.txt

cd $SLURM_SUBMIT_DIR

module load intel/2018.2

export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK

./openmp_example
# or "srun ./openmp_example".

nia-login07:~$ sbatch openmp_job.sh

First line indicates that this is a bash script.
Lines starting with #SBATCH go to SLURM.
sbatch reads these lines as a job request (which it gives the name openmp_job) .
In this case, SLURM looks for one node with 40 cores to be run inside one task, for 1 hour.
Once it found such a node, it runs the script:
- Change to the submission directory;
- Loads modules;
- Sets an environment variable;
- Runs the openmp_example application.

Example submission script (MPI)

#!/bin/bash 
#SBATCH --nodes=8
#SBATCH --ntasks=320
#SBATCH --time=1:00:00
#SBATCH --job-name mpi_job
#SBATCH --output=mpi_output_%j.txt

cd $SLURM_SUBMIT_DIR

module load intel/2018.2
module load openmpi/3.1.0rc3

mpirun ./mpi_example
# or "srun ./mpi_example"

nia-login07:~$ sbatch mpi_job.sh

First line indicates that this is a bash script.
Lines starting with #SBATCH go to SLURM.
sbatch reads these lines as a job request (which it gives the name mpi_job)
In this case, SLURM looks for 8 nodes with 40 cores on which to run 320 tasks, for 1 hour.
Once it found such a node, it runs the script:
- Change to the submission directory;
- Loads modules;
- Runs the mpi_example application.

Monitoring queued jobs

Once the job is incorporated into the queue, there are some command you can use to monitor its progress.

squeue to show the job queue (squeue -u $USER for just your jobs);
squeue -j JOBID to get information on a specific job

(alternatively, scontrol show job JOBID, which is more verbose).
squeue -j JOBID -o "%.9i %.9P %.8j %.8u %.2t %.10M %.6D %S" to get an estimate for when a job will run.
scancel -i JOBID to cancel the job.
sinfo -pcompute to look at available nodes.
More utilities like those that were available on the GPC are under development.

Data Management and I/O Tips

$HOME, $SCRATCH, and $PROJECT all use the parallel file system called GPFS.
Your files can be seen on all Niagara login and compute nodes.
GPFS is a high-performance file system which provides rapid reads and writes to large data sets in parallel from many nodes.
But accessing data sets which consist of many, small files leads to poor performance.
Avoid reading and writing lots of small amounts of data to disk.

Many small files on the system would waste space and would be slower to access, read and write.
Write data out in binary. Faster and takes less space.
Burst buffer (to come) is better for i/o heavy jobs and to speed up checkpoints.