Hardware Information

The ACTnowHPC cluster consists of the latest hardware to make sure that your compute jobs run quickly and efficiently.  Advanced Clustering knows how to build and configure the best of breed system.  After all, we've been building, installing and deploying HPC cluster for our customers for more than 15 years.

Unlike other cloud services, all of the systems that will be running your jobs are running on bare metal hardware, with no virtualization system to eat resources, and no extra software or services taking away compute cycles from your HPC work.

 

Compute nodes

Based on our Pinnacle Flex 1FX3601 hardware, our compute nodes have the following specifications:

Processor 2x Intel Xeon E5-2650 v3 "Haswell" 10 core processors (20 cores per node)
RAM 64GB of DDR4 2133MHz (over 3GB per core)
Interconnect Mellanox FDR 56Gb/s low-latency InfiniBand network

High Performance Networks

The ACTnowHPC system uses multiple networks:

Network Features Purpose
FDR InfiniBand 56Gb/s with sub 1µs latencies Message passing (MPI) inter-node communications, and high speed storage access on compute nodes.
40Gb Ethernet 40Gb/s standards-based Ethernet Used for login node VM hosts, and filesystem access.
1Gb Ethernet Standard Ethernet Management network

Filesystems

Each customer on the ACTnowHPC service has their own dedicated storage volumes.  Volumes are placed on our high-performance storage servers and connected via both FDR InfiniBand and 40Gb Ethernet.

Login Nodes

Login nodes are the only virtualized component in the ACTnowHPC system. These login nodes are not for running jobs, but for you to compile code, submit jobs, edit files, etc. Each login node is unique and dedicated to each customer.

An ACTnowHPC customer has their own dedicated login node that is a virtual machine running on our Pinnacle 2FX3602 hardware.  We closely monitor the number of virtual machines per system and maintain the optimal balance of performance.

You can find out more about our login nodes here.