THE BRIGHT BLOG

I recently completed my third year as the CEO of Bright Computing, and it seems that a day doesn’t go by without me discovering some new (to me) capability within our product, Bright Cluster Manager.  With more than a decade of development and customer implementations under its belt, coupled with a product team of 40 people working on the product every day, it’s obviously hard to keep up with everything this product can do.  But while the product has grown in capability, it has remained true to its reputation for being easy to use. 

At the 2019 GPU Technology Conference in March, NVIDIA Founder & CEO Jensen Huang positioned NVIDIA DGX servers at the intersection of scale-up and scale-out architectures, sitting squarely in the sweet spot of data science driven by the combination of increased concurrency of data science workloads and the massive compute requirements associated with those workloads.  As a standalone server, the DGX delivers a solid scale-up architecture for data/compute-intensive workloads, and NVIDIA’s announced acquisition of Mellanox with its high-speed interconnects will only enhance that position and help enable a new realm of scale-out architectures as well. 

One of Bright Cluster Manager’s most popular features is its built-in monitoring system. It is lightweight and efficient, and it works right out of the box. But what people generally don’t know is that they can use Bright to monitor non-Bright nodes; nodes that were neither provisioned by, nor managed by, Bright. The Bright Lightweight CMDaemon can be used to monitor and healthcheck auxiliary servers that support the cluster but aren’t part of it, for example, database servers, authentication servers, and file servers.

I recognize that most readers of this blog are pretty well versed in the complexities of managing large scale clusters for high performance computing, artificial intelligence, etc.  But it might not be so obvious to your colleagues or others that need to understand why commercial-grade software with enterprise-class support is so critical to efficiently deploying and managing these complex systems.  Or why managing geographically disbursed clusters is so challenging.  Or why understanding who is using the cluster resources and how allocating those resources efficiently is crucial to the ROI of these very large investments in hardware, applications, and data centers. 

It was a great privilege for Bright to be an invited guest of Dell EMC at two recent industry events.

We are excited to share with you that Bright Computing will be attending the Advanced Scale Forum this year as a bronze sponsor. The conference is being held April 15 – 17 at the Ponte Vedra Inn & Club, in Ponte Vedra FL, and we expect it to be an exciting event!

Ansible is a simple, agentless IT automation tool that anyone can use. In this blog post, we will look at automating head node and software image tasks using Ansible.

Picking up from last week, I illustrated a common dialogue between IT administrators and executive leadership concerning the decision to build an HPC cluster management solution. Now that the green light has been given to build, all of that money you “saved” by not using commercial cluster management software allowed you to buy more hardware to support jobs from end users, right?  You can see the extra servers on the floor, so you must be providing your users with more capacity for work, right?  Maybe not. Does the do-it-yourself approach you developed tell you precisely which system resources are actually being used by end users, and for which jobs?  Or, are users requesting more resources for their jobs than they really need and sitting on them (unused), preventing other users from gaining access to do real work?  And one more thing … how much server/system resource is being inefficiently consumed by the processes of your do-it-yourself cluster management solution at the expense of real work for users?  The point is, that cluster that appears to be 95% utilized is very likely to be far less productive than you think. 

Having just returned to my home base in the UK after attending the 2^nd running of Supercomputing Asia, where I had the pleasure of meeting and speaking with many of Bright’s customers and partners attending the show, I started thinking about how our business got started in the Region.   

Everyone in the computer business knows just how hot the machine learning (ML) space is today.  The promise, as well as the demands, being placed on the AI data scientist by their companies, are numerous.  For most of these people, their GPU laden computers used to run analysis are viewed as just a tool. Often, each data scientist is provided with their own powerful computer and they don’t want to be burdened with the need to operationalize these computers. Simply put, they just want to run their jobs as quickly as possible.