One of the fastest growing areas for new investments in recent years has been Big Data Analytics. New companies are springing up and old stalwarts are revamping their offerings to provide more efficient ways to systematically extract information from huge volumes of data. While some applications have struggled, new applications like Apache Spark, Hadoop and Map Reduce have grown in stature. RXP hardware accelerated search can help improve the efficiency of Big Data Systems by offloading a key part of the search process from expensive, power hungry host processors to the highly optimized, massively parallel search engine, RXP.
Today’s explosion in the volume of data to be processed has caused some innovative companies to take a deeper look at their architectures. Legacy systems still bring data from hard drives or SSD drives across a network or PCIe bus to be processed by the x86 or Arm based host processors but in many cases this is inefficient. Processing data as close as possible to where it is stored can provide dramatic increases in throughput and reduce latencies, cost and power. By deploying compute right next to an SSD drive, searches can be performed without passing the data back to the host processor, only the search results need be passed back.
Therefore, with computational storage, networks and PCIe buses no longer constitute a bottleneck as less than 10% of the original data is typically passed back to the host. The SNIA organization has a special focus group dedicated to exploring the latest computational storage ideas in both ASICs and FPGA . Now Arm or RISC-V processor coupled with Titan’s RXP technology can search and sift through vast quantities of local data without any need to pass everything back to the host.
Artificial Intelligence, Machine Learning and Natural Language Processing (NLP) systems are in high demand. New algorithms and new silicon are being developed and released at a very fast rate but what many have in common is that the input data often needs to be prepared so that these engines operate at their maximum efficiency and capacity. RXP technology can be used to help identify and strip out extraneous data that adds little or no value to the data analysis.
By using RXP technology as a preprocessor, we have seen as much as 50% of the data being prefiltered out and therefore the next AI/ML or NLP stage of the operation can operate in a more optimal and efficient way.
An interesting use case for RXP technology is assisting with sentiment analysis of data such as Twitter feeds and online reviews. RXP can be used to systematically identify, extract, quantify and study any data where customers are posting their thoughts on products and services.
The huge volume of real-time data is ideally suited to be searched by RXP hardware accelerated search engines.
Financial Data Mining
The amount of data produced by online shopping, in-store purchases as well as stock trades and other transactions can provide a wealth of valuable insights into trends and the value of various sales and marketing activities on actual sales that are made. RXP technology can be used to help identify and classify key transactions so that advanced statistical, mathematical and artificial intelligence techniques can be used to predict future trends and behaviors.
With RXP hardware accelerated search technology, raw data can be better turned into valuable information.