Powering Insight-Driven Enterprises with GigaSpaces’ InsightEdge and Intel
Powering Insight-Driven Enterprises with GigaSpaces’ InsightEdge and Intel
Read this transcripted interview by Jake Smith, who talks to Tal Doron, director of solution architecture at GigaSpaces.
Join the DZone community and get the full member experience.Join For Free
Insight for I&O leaders on deploying AIOps platforms to enhance performance monitoring today. Read the Guide.
GigaSpaces’ InsightEdge platform and Intel are set on delivering fast-data analytics through in-memory computing to accelerate data access, processing, and provide the right platforms to build artificial intelligence applications. In short, InsightEdge is an open-source, in-memory insight platform that unifies fast-data analytics and real-time applications via microservices architecture. In this talk, converted to a transcript for my readers, I'll be talking about how GigaSpaces is utilizing Intel® Xeon® Scalable processors along with Intel® Optane™ technology to reduce storage access latencies and improve application throughput for mission-critical, real-time analytics and actionable insights.
Jake: Good morning, good afternoon, or good evening. Wherever you may be listening to this Chip Chat. My name is Jake Smith, and you've joined us for a chip chat conversation in the cloud. I'm with Tal Doron, director of solution architecture at GigaSpaces. Tal, welcome.
Tal: Hi Jake. Nice to meet you.
Jake: Nice to meet you as well. Can you tell us a little bit about GigaSpaces and also about yourself?
Tal: GigaSpaces has been around for the past 17 years. One of the original companies adopting in-memory technology. Basically allowing organizations, enterprises, financial, transportation, and other verticals to build their applications on top of our platforms. Simplifying and obviously increasing performance of the whole architecture. I came from the R&D, I also went with my career to BI companies, business intelligence, and now focusing on pre-sales and biz dev and seeing a lot of a traditional multi-tier architecture in many of the enterprises, and I think the world is kind of shifting from that traditional approach to a more big data and, even more specifically, fast data architecture.
Jake: Let's talk about how In-Memory architects benefits end users and our customers?
Tal: So, using In-Memory architecture or memory computing is basically leveraging distributed platform to increase performance, lower costs, or should I say, lower TCO, and at the end of the day, basically it's simplifying everything by leveraging a microservices platform that takes care of everything you would basically do in order to keep the architecture working as fast and as highly available as possible.
Jake: Where is it that GigaSpaces can add the most value through analytics for their customers?
Tal: In the past, GigaSpaces were mainly focused on providing transformational opportunities. Basically, how do we squeeze long hours or batch processes into the millisecond range? How do we allow aggregation, correlations, and patterns, and I think what we are kind of looking at now is an insights platform. That is the current evolution of GigaSpaces. We are taking the past 17 years of in-memory computing know-how and applied it to building one platform to serve many aspects of the fast data challenges.
Jake: What kind of vertical markets, you know, if our customer works in transportation or one of our listeners works in oil and gas, how can GigaSpaces help them? Because I know you guys have some leadership positions in some key verticals. Do you want to talk about that a little bit?
Tal: Many verticals are actually investigating how to leverage their data. Then just the data, the archive data, to create actionable insights to allow this data to increase business performance, to cut down costs, and how to make money. Financial services, healthcare, telco, retail, IOT, all those verticals have something in common; they have a lot of data. Whether we are talking on-premise or public cloud. We're talking about ingesting hundreds of thousands of transactions per second and the question is know how to actually store the data but how to leverage the data for predictions, for descriptive analytics and other use cases.
Jake: So let's talk a little bit about the Insight Edge Platform. You guys have developed it, it does target some verticals, but it really is addressing customer challenges. Can you talk a little it about it?
Tal: Yeah, sure. Customer challenges, the way we see the industry, enterprises are slow design, not computation. So, how do you transform traditional architecture, enterprise architecture, multi-tier architecture, the same way we transform the multi-tier architecture into the 17 years of In-Memory computing XAP. How do we transform it into a fast data platform, into an insight platform? So the kinds of challenges we are seeing are how to reduce hours of data simulations into seconds.
How do we build IoT Edge and IoT HUB, how do we implement predictive analytics and anomalies detection? How do we trigger a transactional workflow according to the prediction data? These are the kind of challenges we are seeing in the industry where the more high-end enterprises, especially we see that in finance, trying to build live risk result stores. Not as a matter of aggregating and storing the data but actually being able to turn around and query the data in real time, the transactional data.
These are the kind of challenges we're seeing in the industry. Everybody is trying to solve them, we believe, or should I say, from other articles that I'm reading, about 20% to 30% are actually being successful in being data-driven.
Jake: We have had a long-standing relationship between Intel and GigaSpaces. Can you talk for our listeners so that they understand what are some of the things that we're working on?
Tal: Insight Edge is basically an insights platform. The idea here is to unify the different tiers, especially if we're talking about the big data complex workflow into one platform that can handle many of the aspects. Together with Intel, we are able to actually approach three of those aspects. The first one is on the hardware part, leveraging Intel's new NVMe and Optane technology. So we actually do off-heap persistence from the memory, which as we know, is a costly hardware. We're able to persist the data to the disk. Now, with the new Intel technology, we're talking about 8 to 10 times performance improvements for throughput, operations per second. We have done many benchmarks with previous NVMe technology and the new Optane drives, and it's pretty amazing.
On the software side, we were actually using bigDL, Intel's bigDL for deep learning. Now as we all know, deep learning is something that is very costly. Building your own customized hardware GPU-based farm is expensive.
Leveraging Intel's software on top of our platform, we are able to leverage traditional CPU's for deep learning, using our distributed platform. Actually, we have quite a few nice benchmarks, field studies, and use cases.
Jake: So you touch on something, Tal, that I'm very curious about. On-premise, on the cloud, does GigaSpaces really care where the analytics happen?
Tal: That's a good question. GigaSpaces as a software vendor, we don't really care where the analytics takes place. It could be on-premise, it could be on a virtual machine, it could be on the cloud, hybrid cloud. Currently, GigaSpaces are working on fully supported docker images for our software. So basically, you can deploy the software anywhere you want, obviously leveraging our RESTful API and customize RESTful API. Once you do that, I think there is no limit to where you can deploy this.
Jake: Can you talk a little about how InsightEdge is really transforming deep learning and data analytics?
Tal: Yeah, well, InsightEdge, as I mentioned before, is an “Insights Platform”, we're leveraging Apache Spark, which is a distributed analytical framework basically. It's amazing. However, it does have a few drawbacks. We believe we've solved that by leveraging our distributed in-memory data fabric underneath the hood of Apache Spark and by adding that and a few other solutions into the mix, we're able to produce the right platform for machine learning, deep learning and all that is converged together with transactional processing. That is the future of organizations when talking about analytical roadmaps.
Specifically for deep learning, our platform has increased performance. We're talking operations per second on arbitration in an extreme way. Leveraging our off-heap persistence, we are able to scale into the multi-terabytes range and enable Spark to access warm data in memory and also on SSD, and of course, by leveraging Intel's BigDL we are enabling deep learning to be actually used on top of commoditized servers, leveraging CPUs rather than expensive GPUs and GPU farms, so anyone can actually scale into deep learning. It doesn't have to expensive, it doesn't have to be a slow and agonizing process. It's just a matter of taking an “Insights Platform” and take a few data scientists, a few developers and start developing.
Jake: Outstanding. You know, our listeners are really going to get a lot from that because we want our listeners and we want users around the world to begin using GigaSpaces and Intel Xeon and Intel Optane technologies immediately. So what are ways that our listeners can find out more information about GigaSpaces than we were able to cover today, Tal?
Tal: We have the gigaspaces.com sites. All the information can be found on the site. Solutions, download the product, documentation, white papers and benchmarks of course. I would recommend some of the recent webinars we had and of course, listeners are always welcome to send an email whether to GigaSpaces or to me directly. I'd be more than happy to answer any question anyone has around their “Insights Platform” or the integration together with Intel technologies.
Jake: Well, on behalf of Intel and our "Chip Chat: Conversations in the Cloud," our production team, my name is Jake Smith, it has been my honor to be here today with Tal Doron, director and solution architecture at GigaSpaces. For our listeners around the world, we wish you a good morning, good afternoon, and good evening wherever you may be.
Opinions expressed by DZone contributors are their own.