By Lionel Gibbons | November 15, 2016 | HPC
In my last post, I showed you some of the results of Bright’s latest user survey. In particular we looked at the size of deployed clusters, workload manager preference, and the the high percentage of clusters running GPUs.
Our survey covered a number of other topics, so this time we’ll look at what survey respondents told us about their plans to expand their cluster use, how much uptime they see, and what parts of their infrastructure cause them the biggest headaches.
Let’s take a look.
Despite the fact that most respondents were operating fairly large clusters, most plan to expand. Sixty-nine percent of those surveyed said they planned to add more nodes to their clusters in the next 12 months. 28% said they planned to add more clusters.
The plans for those added nodes is varied. Thirty-five percent plan to run big data services on their clusters within 12 months. Thirty-one percent will run GPUs and deep learning applications. And users aren’t ignoring the cloud. Thirty-four percent of respondents said they planned to add OpenStack to their infrastructure within a year, and 28% have a hybrid cloud strategy in place.
With all of those mission-critical services running on their clustered infrastructure, it’s no surprise people are focused on maximizing uptime. This is one place Bright customers (the people surveyed here) have an advantage. Nearly 70% of Bright customers experience 99.9% uptime on their clusters. For reference, 55 minutes of down time in a month works out to 99.873% uptime.
Another 10% report getting 99.5% uptime. That’s nearly 80% of Bright users achieving 99.5% uptime or better. Customers tell us that much of the credit for their ability to provide the levels of service users demand belongs to the fact that they manage their systems using Bright.
Nobody achieves 100% uptime. That would be nice, of course, but things can and do go wrong from time to time. We asked customers which areas were most likely to be the source of problems. The responses indicate that trouble usually comes from hardware-related issues (36%), but software provides its share of excitement too. System software was pegged as the source of trouble about 29% of the time, and application software was to blame roughly 25% of the time.
When asked how they benefited from using Bright to manage their clouds and clusters, users said ease of use was key. Seventy percent picked ease of use as their top benefit. Time savings came in second with 57%, and having a centralized place to manage all of their clusters ranked third at 51%.
While there is always room to improve, we’re proud of the fact that 8 out of 10 respondents said they would recommend Bright to a colleague. Chances are that if you are reading this, your organization operates some clouds or clusters. You may want to consider the recommendation of Bright customers and see what Bright could do to help you succeed.