NewRiver is a 173-node system capable of tackling the full spectrum of computational workloads, from problems requiring hundreds of compute cores to data-intensive problems requiring large amount of memory and storage resources. NewRiver contains six compute engines designed for distinct workloads.
- General - Distributed, scalable workloads. With two of Intel’s latest-generation Haswell processors, 24 cores, and 128 GB memory on each node, this 100-node compute engine is suitable for traditional HPC jobs and large codes using MPI.
- Big Data - High performance data analytics. With 43.2 TB of direct-attached storage for each node, this system enables processing and analysis of massive datasets. The 16 nodes in this compute engine also have 512 gigabytes (GB) of memory, making them suitable for jobs requiring large memory.
- Visualization (K80 GPU) - Data visualization. Nvidia K80 GPUs in this system can be used for remote rendering, allowing large datasets to be viewed in place without lengthy data transfers to desktop PCs. The 8 nodes in this compute engine also have 512 GB of memory, making them suitable for jobs requiring large memory.
- Interactive - Rapid code development and interactive usage. Each of the eight nodes in this compute engine have 256 GB memory and a K1200 graphics card. A browser-based, client-server architecture allows for a responsive, desktop-like experience when interacting with graphics-intensive programs.
- Very Large Memory - Graph analytics and very large datasets. With 3TB (3072 gigabytes) of memory, four 15-core processors and 6 direct attached hard drives, each of the two servers in this system will enable analysis of large highly-connected datasets, in-memory database applications, and speedier solution of other large problems.
The table below lists the specifications for NewRiver's compute engines. View a detailed schematic of NewRiver and the ARC Core Network here.
|Compute Engine||#||Hosts||CPU||Cores||Memory||Local Disk||Other Features|
|General||100||nr027-nr126||2 x E5-2680v3 2.5GHz (Haswell)||24||128 GB, 2133 MHz||1.8TB 10K RPM SAS|
|Big Data||16||nr003-nr018||2 x E5-2680v3 2.5GHz (Haswell)||24||512 GB||43.2 TB (24 x 1.8 TB) 10K RPM SAS||2x 200 GB SSD|
|Visualization (K80 GPU)||8||nr019-nr026||2 x E5-2680v3 2.5GHz (Haswell)||24||512 GB||3.6 TB (2 x 1.8 TB) 10K RPM SAS||NVIDIA K80 GPU|
|Interactive||8||newriver1-newriver8||2 x E5-2680v3 2.5GHz (Haswell)||24||256 GB||NVIDIA K1200 GPU|
|Very Large Memory||2||nr001-nr002||4 x E7-4890v2 2.8GHz (Ivy Bridge)||60||3 TB||10.8 TB (6 x 1.8 TB) 10K RPM SAS|
- Theoretical peak (CPU-only) performance of a single NewRiver Haswell node is 960 GFlops/s, compared with 330 for a BlueRidge node. In testing, threaded matrix multiplies (dgemm) peaked at ~800 Gflops/s for gcc/ATLAS, ~780 Gflops/s for Intel/MKL, and ~730 Gflops/s for gcc/OpenBLAS.
- 1.8 TB NVMe drives on BigData nodes will be made available soon.
- K80 GPU Notes: There are 2 CUDA Devices Although the K80s are a single physical device in 1 PCIe slot, there are 2 separate GPU chips inside. They will be shown as 2 separate devices to CUDA code. nvidia-smi will show this.
- 100 Gbps EDR-Infiniband provides low latency communication between compute nodes for MPI traffic.
- Compute nodes are connected to storage (GPFS) via 100 Gbps EDR-IB. They can also be connected via dual 10 Gbps Ethernet.
- Interactive nodes are connected to the public internet by 10 Gbps Ethernet.
Note: NewRiver is governed by an allocation manager, meaning that in order to run most jobs on it, you must be an authorized user of an allocation that has been submitted and approved. The open_q queue is available to jobs that are not charged to an allocation, but it has tight usage restrictions (see below for details) limits and so is best used for initial testing in preparing allocation requests. For more on allocations, click here.
NewRiver has seven queues:
- normal_q for production (research) runs.
- largemem_q for production (research) runs on the large memory nodes.
- dev_q for short testing, debugging, and interactive sessions. dev_q provides slightly elevated job priority to facilitate code development and job testing prior to production runs.
- vis_q for interactive visualization on the GPU nodes.
- open_q provides access for small jobs and evaluating system features. open_q does not require an allocation; it can be used by new users or researchers evaluating system performance for an allocation request.
The settings for the queues are:
|Access to||nr004-nr018, nr029-nr126||nr001-nr002||nr003-nr018, nr027-nr126||nr019-nr026||nr004-nr018, nr029-nr126|
|Max Jobs||6 per user
12 per allocation
|1 per user||1 per user||1 per user||1 per user|
|Max Cores*||768 per user
1,152 per allocation
|60 per user||768 per user||48 per user||96 per user|
|Max Memory*||4 TB per user
6 TB per allocation
|3 TB per user||4 TB per user
6 TB per allocation
|384 GB per user||1 TB per user|
|Max Walltime||144 hr||144 hr||2 hr||4 hr||4 hr|
|Max Core-Hours*||27,648 per user
41,472 per allocation
|8,640 per user||192 per user||192 per user||192 per user|
*Entries with two values (e.g., XX/YY) represent soft and hard limits, respectively. Hard limits apply when the queue is not busy (there are idle resources); soft limits apply when the queue is full. The two different limits allow users to use more computational resources when those resources are not in demand.
- Shared node access: more than one job can run on a node (Note: This is different from other ARC systems)
For list of software available on NewRiver, as well as a comparison of software available on all ARC systems, click here.
Note that a user will have to load the appropriate module(s) in order to use a given software package on the cluster. The module avail and module spider commands can also be used to find software packages available on a given system.
NewRiver provides two access methods: traditional terminal access and broswer-based access for improved graphical experiences.
The cluster is accessed via ssh to one of the login nodes below. Log in using your username (usually Virginia Tech PID) and password. You will need an SSH Client to log in; see here for information on how to obtain and use an SSH Client.
NewRiver offers web browser-based access to interactive nodes via ETX. This interface provides faster, more interactive access to graphical user interfaces than standard X11 forwarding (e.g. ssh -X). ETX will automatically load balance users between the eight interactive nodes.
To access it, use a web browser (e.g. Firefox or Safari, not Chrome) to go to
For more information, see the ETX page.
Access to all compute engines (aside from interactive nodes) is controlled via the job scheduler. See the Job Submission page here. The basic flags are:
#PBS -l walltime=dd:hh:mm:ss #PBS -l [resource request, see below] #PBS -q normal_q (or other queue, see Policies) #PBS -A <yourAllocation> (see Policies) #PBS -W group_list=newriver
Compute nodes are not dedicated to a single job (as is done on BlueRidge). NewRiver has more options for requesting resources to ensure the scheduler can optimally place jobs. Resources can be requested by specifying the number of nodes:ppn (like on BlueRidge), but also cores, memory, GPUs, etc. See example resource requests below:
Request 2 nodes with 24 cores each #PBS -l nodes=2:ppn=24 Request 4 cores (on any number of nodes) #PBS -l procs=4 Request 12 cores with 20gb memory per core #PBS -l procs=12,pmem=20gb Request 2 nodes with 24 cores each and 20gb memory per core (will give two 512gb nodes) #PBS -l nodes=2:ppn=24,pmem=20gb Request 2 nodes with 24 cores per node and 1 gpu per node #PBS -l nodes=2:ppn=24:gpus=1 Request 2 cores with 1 gpu each #PBS -l procs=2,gpus=1
This shell script provides a template for submission of jobs on NewRiver. The comments in the script include notes about how to request resources, load modules, submit MPI jobs, etc.
To utilize this script template, create your own copy and edit as described here.