Ogma 5.0 release: A state-of-the-art GPU layout

We’ve been promising something big in the works for Ogma. We’re happy to announce that, with our latest release of Ogma 5.0, that new big thing is here. The latest, state-of-the-art iteration of our GPU layout now makes it possible to calculate the coordinates of hundreds of thousands of nodes in a few seconds simply using the client’s web browser and a mid-market graphic card. 

AI and data analytics software can directly benefit from this breakthrough improvement, as no heavy client like Gephi is required, and no server-side computation is needed contrary to Graphistry.

Read on to learn more about how the new layout works and what it means for your graph visualizations and exploration.

In interactive graph visualizations, there are two major performance bottlenecks: rendering and layout. Layouts are based on the state-of-the-art physical simulation algorithms and are notoriously hard to implement and to tune. 

For complex algorithms, a popular modern approach is parallelization on the graphical card (GPU), where parts of the algorithm are identified that can be run in parallel. Then the algorithm is redesigned completely to be run on the graphics card that is capable of running thousands of small programs at the same time.

About a year ago, we initially released a GPU version of the layout, showing how hardware alone could make the computations many times faster. This initial version of the GPU layout worked well, but it had limitations. The gains in speed could only be achieved on graphs with fewer than 10,000 nodes.

After much research on our end, we are happy to present the next generation of the GPU layout. It combines the power of parallel computations on GPU and a state-of-the-art scientific algorithm, delivering up to a 40x speed increase to the existing CPU implementation. It’s even faster than desktop applications like Gephi.

It opens a new page for interactive visualizations with Ogma since it can handle in seconds what previously took minutes to compute. Now, you can:

• Start the exploration of your data on bigger datasets
• Stop worrying about re-layouts, as they are fast even on graphs of hundreds of thousands of nodes

The new algorithm brings more than just performance. In addition to being so much faster, it also delivers much better precision and quality, which you can see in the increased symmetry and cluster shape quality.

With this new tool in our hands, we are beginning a journey into big graph exploration - something we have wanted to unlock for years. So, there is more in the works. This year, we have a steady plan to continue working on performance and quality, learning more about the challenges of drawing bigger graphs. 

Play with GPU layouts in the Ogma playground.