The Nimbix Cloud – True HPC in the Cloud

June 28, 2018

By: Tom McNeill

VIEW ALL

When most people think of cloud computing they think about hosting a web server or maybe spinning up a huge Hadoop cluster composed of commodity hardware to tackle a MapReduce problem.  This usually involves renting a virtual machine, installing all of your software, dealing with cryptic error messages, and getting eye-watering bills that charge you for every bit you move and round up your time to the nearest hour (use the machine for 15 minutes, get charged for an hour).  

Well, at Nimbix things are a little different.  We’re PaaS and SaaS, rather than IaaS.

Your software is probably already installed, so click on the software icon, choose your machine and run.  If not, containerize it with Docker or Singularity. Our platform happily handles both. Then bring it in with our PushToCompute™ system.  You can be running in under ten minutes. That takes care of putting software on the platform.  

One of the benefits of containerization is that Nimbix containers are on bare metal.  This allows you to run your applications faster because you don’t have the computational overhead of a hypervisor that mediates between a virtual machine and the bare metal hardware.  This gives you something in the neighborhood of a 10-20% performance increase simply by avoiding the hypervisor. That is the point of high performance computing; speed and capacity, so removing the hypervisor bottleneck increases your performance.  

Another performance increase comes from the network fabric.  All Nimbix compute nodes are connected with a 100-gigabit spine and 56-gigabit branches allowing the fastest possible data throughput.  We also have high-speed storage capable of over ten thousand IOPS.  You need that kind of throughput because when you have GPU and FPGA accelerated machines running your hardware, you need to keep your accelerators full.  

We are on the cutting edge of high-speed processing hardware as well.  We have the NVIDIA DGX with V100 GPUs, IBM P8 and IBM P9 systems running NVIDIA P100 and V100 GPUs respectively. We also have Intel machines running the newest Intel chipsets.  If that weren’t enough, we have a full catalog of Xilinx Field programmable FPGAs. We’ve got an accelerated solution for every use case.

Give one of our sales executives a call and let’s see how we can accelerate your computing needs.

TwitterLinkedInFacebookPrintGoogle+

Other Articles to Read