High-Performance Data Management on HPC Infrastructure

Industrial organisations involved

Cherrydata is a start-up that offers technology and solutions for highperformance data intelligence, with exceptionally fast response time and extremely low hardware
(HW) requirements.
This performance edge is granted by AyraDB, a next-generation database with outstanding performance and scalability.
This allows Cherrydata’s customers to work on their business KPIs with no technical or economic scalability limitations. It also reduces energy consumption, with a positive environmental impact

Technical/scientific Challenge

To enable a variety of big data applications, the market needs highly efficient data management technologies: satellite data are notoriously big, raising challenges not only from the perspective of analytic applications, but also to be efficiently stored and retrieved.

AyraDB was designed to address these challenges.

The goal of the present research is to perform a large-scale test of AyraDB on Davinci-1 supercomputer, belonging to Leonardo Company.

These tests can act as a Proof of Concept (PoC) of thetechnical features of AyraDB and of its suitability for challenging applications.

The Solution

Tests were performed on Davinci-1 with a total of 500 runs on a number of 8-core servers growing from 1 to 20, a data size growing from 10 GB to 500 GB.

Results show that AyraDB reaches 106 requests/s with 13 servers, with consistent read/write performance, beating the state of the art by a 5x factor.

AyraDB has not only best-in-class cost-to-performance ratio. Tests demonstrated that it has linearly scalable performance: if HW capacity doubles, AyraDB’s performance doubles. Competing solutions require more than double capacity to double performance.

Above a certain threshold, their performance stops increasing even if the infrastructure provides additional capacity, the database cannot exploit it.

In fact, AyraDB can use additional capacity with no upper bounds to scalability.

Business impact

The PoC has provided third-party performance benchmarks. Obtaining third-party performance benchmarks is a time-consuming and costly activity and the PoC has provided these benchmarks,
acting as proof of the performance of AyraDB beyond Cherrydata’s own claims.

This is clearly important for Cherrydata and can boost adoption by helping overcome the skepticism that is inevitably associated with a new technology. In the context of an IT deep tech is even more critical. The quasi monopoly of global cloud providers and their aggressive marketing have created the belief that they provide the best possible technology.

With this PoC, it was proved that it is possible to substantially improve the quality of cloud services and that greater performance can be obtained with a technology architecture that makes a better use of RAM and, thus, reduces costs by a factor of 20.

The benefits

  • Performance improvement of data management
  • Cost reduction of cloud services
  • Enabling technical and economic feasibility of several big data applications
  • Energy savings

Images Courtesy: Cherrydata