Hadoop vs. Spark? That’s the very outdated question that data analytics experts are asking these days. I’ll tell you up front: when faced with only those two options, it may be the lesser of two evils, but Spark is the answer. Although it is in its infancy and is rather immature, Spark will continue to outperform Hadoop. And given that it has a simpler programming model, it has not only a performance advantage over Hadoop but also an ease-of-use advantage over Hadoop. What Spark doesn’t have yet is a stability advantage over Hadoop, but eventually it should have that as well. Especially now that IBM is throwing its considerable weight behind Spark. However, the trap that organizations today are falling into is thinking along the lines of, “Well, all we have is Hadoop and Spark, so I have to pick one!” If you’ve followed Ryft over the past several months, you’ll know that we strongly disagree with that. We stand firmly behind the knowledge that homogeneous x86-based solutions aren’t the be-all and end-all of big data analytics. We want organizations to have the ability to streamline and accelerate their analytics infrastructures with the power of heterogeneous computing, which can help give them the performance and insights they need. This is why we launched the 1U Ryft ONE earlier this year, which ingests, stores and analyzes 48 TB of batch and streaming data with 100X performance when compared to today’s high performance servers. But what if you’ve already thrown your blood, sweat, tears, budget and time behind building out a Spark cluster, only to come to the realization that it still doesn’t give you the real-time insights you need? What if the slow realization has come that you can’t obtain the valuable insights from Internet of Things (IoT) sensor data, which are becoming such a critical piece of making better business decisions? Is accelerating performance really out of reach for the masses? At Ryft, we don’t think so. This week, we announced the Ryft Connector for Spark, which seamlessly integrates the Ryft ONE into new or existing Spark ecosystems to increase Spark performance for a variety of critical data analytics functions by more than 100X while streamlining the infrastructure required to extract insights from high volume, velocity and variety big data. Spark, while a definite improvement over Hadoop, is still severely limited by the often homogeneous x86-clustered or GPU-clustered hardware that it is typically run on. Days, weeks or even months of data transport, indexing and ETL bottlenecks will stifle Spark’s ease of use and performance, especially when you invariably find that your data doesn’t fit nicely in local RAM, which is so critical to x86-based and GPU-based performance. The result is that these architectures revert to slower disk-based and network-based transfers, which effectively bring your data analytics infrastructure to a grinding halt. By natively integrating Ryft ONE into a Spark cluster, Spark users can configure the Ryft ONE to be a Spark preferred node for specific data analytics functions. The end result is that these critical data analytics functions perform exponentially faster than—and without the bottlenecks of—the cluster infrastructure’s other nodes. The Ryft Connector for Spark solves the biggest challenges facing businesses trying to get faster insights from data: increasing performance, reducing complexity, simplifying usability and delivering efficiency gains. Reach out and let us know if you are struggling with performance issues either with Spark or another analytics ecosystem, and we’ll help you with a full audit to determine how you can streamline, accelerate and simplify your big data analysis using the high-performance 1U Ryft ONE appliance. Leave a Reply Cancel reply Your email address will not be published. Required fields are marked *Comment Name * Email * Website Save my name, email, and website in this browser for the next time I comment.