We’ve been living in a scale-out world for analytic workloads during the last handful of years. MPP data warehouses, clustered NoSQL, Hadoop® and Spark™, falling collectively under the big data umbrella, all readily come to mind. When better performance, support for a greater number of users or increased capacity is required, we scale out with server nodes comprising compute, memory and storage resources. But unless you operate Amazon Mechanical Turk, scaling the number of people as analytic needs grow is something to be avoided.
One topic ESG analyst Nik Rouda will discuss in our joint webinar, How to Realize Analytics Value Faster, is current research showing how few organizations are planning significant growth in IT staff. This is unsurprising if you consider that many enterprises, discounting the Facebooks and LinkedIns of the world, want to focus on their core competencies and do not set out to build large IT fiefdoms. Their main business is finance, healthcare, energy or the like, not managing large-scale IT infrastructure. But is it realistic to embark on an analytics project without commensurate growth in staffing? How can we reconcile the adoption of scale-out big data — especially with Hadoop, Spark and NoSQL requiring non-traditional skill sets — with the desire to hold IT staffing steady?
The first thing to realize is that even if organizations want to add people with the requisite big data skills, they may not be able to do so predictably. While vendors and universities are working to close the skills gap, it remains a challenge to hire experienced professionals in many geographies. A solution that boosts productivity, so that organizations can do more with less, is needed. Cray set out to do just that with the Urika-XA™ extreme analytics platform, which unifies hardware, software and management into a ready-made platform for big data analytics.
The Urika-XA platform enables users to get up and running immediately with Hadoop and Spark. There is no need to build out a big data cluster from scratch, so users enjoy faster time-to-value and require less in-house expertise and effort. Cray multiplies the productivity of an existing IT team by delivering a turnkey, converged platform that shifts the burden of configuring and supporting a complete system away from the organization. End users gain access to an analytics environment as quickly as possible, and IT accomplishes this by leveraging Cray’s foundational work. This promise of simplicity has resonated with many organizations we’ve spoken with, especially those who have experienced first hand the complexity of standing up a Hadoop cluster themselves.
Reducing ongoing management overhead for a system is another lever in improving IT productivity. One of the objectives of the Urika-XA system design is the consolidation of analytic workloads onto a single high-performance platform. Its high compute and memory density allows a single Urika-XA rack to replace three typical racks in the datacenter. As a multi-purpose platform, diverse classes of analytics can be accommodated in one place, avoiding cluster sprawl. A smaller overall footprint requires less effort to manage. With the Urika-XA system, organizations also benefit from a unified management pane and single point of support from Cray. With this management-friendly design, Cray empowers IT teams to support serious analytics without requiring significant growth in IT staffing.
Scale-out architectures are inherently more complicated to operate. Assuming you don’t have the wherewithal to scale your people proportionally to your Hadoop clusters, you need to select solutions that maximize the productivity of your existing team to have the best chance of success. Let us know how we can help.
The post Scale Out Analytics, Not People appeared first on Cray Blog.
