RETHINK big news roundup: open-source optical networks, Deep Learning on GPU clusters and the ideal memristor
In this edition of our news roundup, we bring you the following news:
The ideal memristor
In the last news roundup, there were some interesting news related to the memristor. This time researchers at Michigan Technological University have made an ideal memristor. According to them, an ideal memristor is the one that has an even relationship between current and voltage. This new technology may as well start a new era of faster and less energy consuming memories.
The full article can be found in the Michigan Tech website.
AMD releases open-source GPU tools for HPC
AMD has released a new toolbox to work with their GPUs and it has also been open-sourced. These tools have been released with the HPC community in mind. This may be very advantageous for developers that are considering switching from Nvidia to AMD.
Read the full article at The Register website.
The new Internet era with open-source optical networks
According to some scientists at the University of Bristol, future innovations and development in the Internet as it stands nowadays will require huge efforts because current infraestructures do not support independent development. The physical layer of current networks is too intricate for traditional programmers, however, open sourcing these networks and hiding the physical layer complexities would aid developers so that they would concentrate more on their specific problems.
Read the full article at NetworkWorld website.
The future of optical networks
A new Bandwidth Variable Transmitter (BVT) has been created by a group of researchers funded by the European Union in the IDEALIST project. This new device will be a vital component of larger optical networks and it is key towards bringing the future of more adaptable optical networks that will avoid current restrictions and enable innovation to the end users.
Read the full article on CORDIS website.
Google Cloud Dataproc is now publicly available
Google Cloud Dataproc is a platform that was developed so that customers could spin up clusters as needed in a convenient way isolating the data processing tasks from what actually matters. They let the users concentrate more on the data analytics tasks instead.
Read the full article at datanami website.
Deep Learning on GPU clusters
Yahoo has just released a new framework called CaffeOnSpark merging various technologies such as Hadoop, Spark and Caffe. Before the release of this framework, these techniques had to be applied separately in different clusters and this implied that data had to be moved from one cluster to another. Now the CaffeOnSpark framework offers users the opportunity to apply Deep Learning techniques in single Hadoop and Spark based clusters.
Read the full article at The Platform website.