This release includes support for Java 11 and for edge-to-cloud environments.
The Workflows and Distributed Computing team at the Barcelona Supercomputing Center is proud to announce a new release, version 2.0, of dataClay, a distributed object store that avoids the need to have different data models in volatile memory and in persistent storage. It performs the calculations directly on the object store, without having to copy the data to the application space, thus avoiding both the time and energy costs associated with data transport. On the other hand, when working on a single model, the effort and time required to transform the data disappear, thus reducing the possibility of errors and improving efficiency of execution.
The new release comes with support for federating independent dataClay instances, thus providing a shared object space among different machines. This new feature, together with the ability to run on constrained devices such as Raspberry Pi or Nvidia Jetson boards, makes dataClay suitable not only for HPC clusters but also for edge-to-cloud deployments. In this context, dataClay has been adopted in the mF2C, CLASS and ELASTIC research projects, where it provides data management capabilities across the edge-to-cloud continuum.
dataClay 2.0 provides support for Java 11, while still giving backward compatibility to Java 8. Additionally, it includes improvements in the deployment, and performance optimizations both for Java and Python.