HAMR Analytic Technologies
Redefining Predictive, Dynamic Big Data Analytics with HPC

High Performance Computing (HPC), due to its extreme computational power, finds its applications in solving today’s most complex problems linked to data storage and analysis. However as computing systems grow larger and more interconnected, the boundaries between the systems introduce stricter limitations on the flow of data between them. HAMR Analytic Technologies (HAMRTech), a Newark, DE based company, overcomes such limitations with its HAMR dynamic data analytics software engine that allows parallel execution of tasks on distributed computing systems. The company has coined the term FlowletTM to describe a data flow actor in a workflow that performs computation on an input dataset to produce output datasets. “HAMR's runtime uses Flowlet technology to enable complex processing graphs operating on in-memory data. Flowlets can read and write to files, store data in memory, and perform processing tasks,” says Rich Collier, CEO HAMRTech.

HAMR comes loaded with many unique features. It not only uses batch processes to handle existing chunks of data like other Big Data systems, but also streams the dynamic data from the cloud—resulting in real-time data analytics. “In fact one HAMR workflow can read data from both batch and streaming resources at the same time,” remarks Collier. HAMR performs fully in-memory data processing that eliminates the need to read and write data between disks. “Also, the analytics engine is very flexible—be it dynamic or stored data, it can flow through HAMR as long as it is represented using key value pairs,” he adds.

HAMR is different than similar systems (like Hadoop, MapReduce and Spark) as it is fundamentally a streaming engine, unlike others. HAMR seeks to reduce data loads by pushing data out of the system as quickly as possible, thereby minimizing the memory footprint. This allows users to process larger data sets in smaller systems. This feature of HAMR also has a significant performance impact. Reducing the memory footprint early means that there is less data to process, freeing processor resources to perform more computation. It solves the most complex problems involving an intricate workflow. Also, the HAMR interface allows the developer to work with data and provide answers at a low level with 10X performance gains over Spark and approximately 10X memory efficiency.

This is important for complex analytics operations like predictive analytics that are dependent on speed. “Our real-time data analytics enables IT organizations to move from backward looking reporting to forward looking predictive analytics,” says Collier.

HAMRTech takes the approach that the environment is not something that needs to be customized or optimized to run the product. “We have focused on the customer’s needs from the initial HAMR prototype. We‘ve built HAMR with the understanding that it would need to fit within existing IT big data infrastructures, or stand-alone,” he adds. For instance, one of HAMRTech’s customers had large amounts of streaming data that needed to be processed in a 12-hour timeframe. The process produced a key performance metric used for billing, demand forecasting, and revenue estimation. Faster processing of the data meant faster reaction to a demand spike, which translated to increased revenue and happier customers. The existing technology couldn’t process the data in the required time window and the algorithms were too complex to build. “HAMR provided a mechanism to build dynamic algorithms, process the data in real time and provide their businesses with much needed information, faster than any competitor,” extols Collier.

Our real-time data analytics enables IT organizations to move from backward looking reporting to forward looking predictive analytics

Moving forward, HAMRTech will maintain a focus on the latest advancements in technology to deliver state-of-the-art products, while improving ease of use. “We are always ready to build the best solutions to better serve the customers in this constantly changing big data market,” concludes Collier.