Nimbix Blog

Super musing about all things supercomputing

3 Signs You Need Cloud HPC

Written by: Leo Reiter on July 22, 2014

Cloud HPCAs Cloud HPC continues to grow in options and popularity, how do you know when it’s time to take advantage of it?  Here are 3 signs that indicate it’s time for Cloud HPC…

Your Cluster is Out of Capacity… Again!

Despite having a supercomputer or HPC cluster in house, it’s fully booked as usual.  If you had all the time in the world, you could wait for your turn patiently… but let’s face it, you don’t.  Your project (or assignment) is due at the end of the week, and it’s Tuesday already.  Sound familiar?  Get used to it – unless all current trends suddenly reverse, demand for high performance compute time will continue to skyrocket.  Even if your organization is planning a major upgrade in capacity, chances are it won’t be enough to satisfy future demand once it’s online.  This is why Cloud HPC is so powerful – it grows as fast as you need it to, and it’s available right now.

You’re Constantly Simplifying your Models or Breaking up your Workflows

That high end laptop or desktop PC may sound powerful, but with a consumer grade GPU and limited processor and memory capacity, it’s no HPC cluster.  If you find yourself trimming your models or workflows just to get them to simulate or process quickly, you’re wasting precious time and accepting lower quality results than you deserve.  Let the PC do its productivity/collaboration thing, and let Cloud HPC actually run your HPC workloads as they were intended to run.  With professional grade accelerators such as passively cooled GPUs and Infiniband interconnects, you’ll get results much faster, and it will take a lot of compute time to add up to the cost of running and managing a high end PC.

You Need Cloud HPC, not Commodity Web Services

If you’ve already looked to the cloud only to find slow, oversubscribed commodity “web services” that can’t get the job done, it’s time to look at Cloud HPC.  Rather than spinning up countless instances and hacking them together with complex configurations, scripts, and API calls, just look for a service that allows you to provide your data, select your application, and get your results back.  You don’t need to manage your own cloud-hosted cluster in order to do this.  A higher level Cloud HPC Platform-as-a-Service or Software-as-a-Service offering spares you the hassle so you can get straight to what’s important – your results!  This is why you looked to the cloud in the first place.  If your project or assignment is due at the end of the week, you can’t be figuring out how to install and configure TORQUE on Tuesday.  And don’t be fooled by seemingly low per-core-hour pricing.  Commodity instances are much slower and less reliable than real Cloud HPC, which means you need much more scale to approach the performance you’d get from purpose built solutions.  Add in the time you spend configuring the virtual cluster, and you end up paying more for the job than you would with a higher value service such as JARVICE!

Whether it’s capacity, performance, or complexity constraints, it’s time you looked at Cloud HPC to get results faster and easier than ever before.  Let us show you how today.