DOME Open Users Platform Event

October 29th, 2014

ASTRON, Dwingeloo

General information

We are proud to invite you to the next DOME Open Users Platform Event, October 29th  in the Van de Hulst auditorium at ASTRON.


Challenge us


Small and medium enterprises, larger companies and knowledge institutes are cordially invited to put forward specific issues which might come to a solution using DOME technologies. In other words: what IT-related constraints are now holding you back in developing products and services, which you think the DOME team could help you overcome?
During the interactive sessions, the DOME team will review the issues you propose. If there is a viable match with one or more of the DOME research streams, the aim would be to proceed with R&D on a collaborative basis.  
For more information on the DOME research streams, please scroll down. We’re expecting some lively discussions during various interactive sessions on October 29th.


DOME research


Even though the DOME project aims at future exascale applications, the underlying technologies are widely applicable. Among other things, they are about efficient and low-power systems and components.
Below, we have listed some typical DOME technologies and approaches. This should help you in judging whether there a possible match between DOME research and you particular challenge. Please note, this is only a cross-section of our activities. If you have any doubts whether or not a topic fits, don’t hesitate to put it forward so that we can discuss. We’d rather have one too many, than miss out on an interesting line of thought.
For a recent update on each of the seven research streams, see DOME newsletter 4.

1) Algorithms and machines: we have made a computing model with which we can estimate the required cost and power, given a particular hardware system, and given particular algorithms. This approach can be used efficiently and effectively in system design of powerful compute applications. As verified experimentally, this model gives a fair prediction of required compute resources.

2) Access patterns: we have created a model for tiered data storage. It gives an optimum breakdown of required memory types (solid state, disk, tape) given access time constraints and cost constraints. This helps us define reliable high-volume and low-cost storage systems. We are also researching new storage technology components, such as phase change memory and modern tape technologies.

3) Nanophotonics: we have tested transmitting wide-band radio signals over optical fibre showing a higher reduction in data transport cost. We also have demonstrated very high density optical connections to electronic printed circuit boards. We have selected different technologies for short distance and long distance data transport.

4) Microservers: as presented at the DOME Microserver Premiere on July 3rd in Dwingeloo, the DOME team has created a very powerful yet low-power microserver unit, applicable in many technological and business areas. Microserver units will be made available for promising application areas. One of the technological breakthroughs is the use of power-efficient hot-water cooling.

5) Accelerators: we have shown that dedicated processing functions can be implemented very efficiently if the algorithms and the compute platforms (accelerators) are carefully designed and chosen. This goes well beyond the standard PC accelerator applications.

6) Compressed sensing and novel algorithms: for our specific high volume and compute-intensive applications we are looking into fundamentally new algorithmic approaches, making use of the fact that information buried in data usually has structure.

7) Real-time communication models for exascale computing: we have recently started research on real-time communication with the aim to drastically reduce the energy needed for high-volume data transport. Here, we're looking at various (RDMA) industry standards.

 

 

 

 


E-mail: info@dome-exascale.nl