"Gradient" was applied by Nvidia Grid K2 to virtualization of graphics

On December 19, 2015 the Gradiyent Ltd company announced use of NVIDIA technologies for geological exploration solving of tasks^[1].

Before the project

In 2009 the company used accelerators of Tesla for calculations of distribution of seismic waves in difficult geological environments for search and investigation of deposits of hydrocarbons.

In 2015 the Gradiyent company faced a problem – the volumes of numerical models calculated using a cluster on the GPU platform became too big for their transfer on jobs of specialists for the analysis and visualization.

In process of complication of the tasks solved by us also volumes of numerical models of the geological environment which we use for explorative tasks grow. When our customers, first of all JSC Gazprom and JSC Lukoil, develop harder and harder areas in the geological relation, from us work with models to one billion cells is required. If calculations of models of such volume on a cluster based on GPU went without problems, then our software solution executed in client-server technology did not cope with the analysis and especially visualization of these volumes. Evgeny Biryaltsev, the CEO of Gradiyent-technology LLC company – the engineering center Gradiyent Ltd, -

Broadcast of hundreds of gigabytes of data of modeling on an Ethernet network 1Gb from a cluster to a workplace became the main problem of analysts. The problem was aggravated with territorial dispersion of divisions of the company across the territory of Russia, respectively access to a computing cluster is possible only on the Internet. It constrained use of technologies of Gradiyent company.

Project Results

The task was resolved with availability of the NVIDIA Grid cards to support of virtualization of graphics in a cloud. Using specialists of the Kazan branch SRISA RAS, the software package of Gradiyent Ltd is transferred to Citrix XenDesktop environment. As a part of a complex of technical means acts the virtualization server on the basis of two NVIDIA GRID K2 cards.

The solution helped to exclude a data transmission problem on client side. Now the client receives a ready video flow from the server, and requirements to a data network do not exceed requirements for viewing videos of Youtube. At the same time remote users of the Gradient system have continuous access to data of any volume.