OmniSci Takes on Big Data with GPU Accelerated Analytics Technology

How ideas turn into reality can be a fascinating story, as is the case with OmniSci CEO Todd Mostak and how his company got its start.

Mostak traveled to Syria and Egypt to teach English and learn Arabic after finishing his undergrad studies at MIT. While taking computer science classes, he was studying spatial temporal data. This was around the time of the Arab Spring (2011) and when he got back to the U.S. he began studying social media as it was a catalyst for what was happening in the Middle East.

“I got excited about social media, especially with Twitter as a way of understanding what was happening on the ground, what people were saying, who is saying, what, when and where things were happening, which were geocoded,” said Mostak in an interview with GISCafe Voice. “I began harvesting millions of tweets, and with my computer science background I was basically trying to join them together, a very geospatially intensive activity.”

Mostak wanted to understand the people who started the revolution, who were more secular, etc. He wanted to know did they come from richer or more rural areas, what age groups participated mostly, which he addressed with geo-related regression.

“I was using PostGres for a lot of it, plus stuff I did myself and on top of it I wanted to do real time visualization of the geospatial data,” said Mostak. “I found it to be incredibly challenging and just to get a single query answered could take overnight and I would wake up and realize I had done it all wrong.”

Mostak got interested in video games accelerated by graphics cards and took an MIT database course. He realized he could use graphics cards that have thousands of cores instead of tens of cores of CPUs to basically power big data analytics or database and visualization platform. The GPUs have a lot of compute power but also visualization and innate rendering capabilities. The bottom line: this could be used to render big data, especially geospatial data.

https://www.omnisci.com/demos/flights/#/dashboard/4

This work at MIT on GPU accelerated analytics technology allowed you to access billions of points, hundreds of millions of polygons, lines, etc. The company MapD (now Omnisci), was spun out of that research in late 2013. Now OmniSci, the company employs 110 people and  is based in San Francisco.

60% of OmniSci’s use cases revolve around spatio-temporal use cases, such as automotive, oil and gas, federal, financial services, and telco.

In the beginning the OmniSci platform was just SQL but because geospatial became such a big use case they added OGC functionality to develop a more holistic geospatial comprehensive product.

One of OmniSci’s earliest customers, Verizon, now an investor, uses the platform for network planning and authorization and is a big proponent of 5G. 5G, Mostak says, is the same thing but at a much bigger scale. The towers are small and have small range, but it becomes a big data information problem that’s very geospatial. Verizon needs to find out where subscribers are, where it’s best to place these towers or infrastructure to maximize coverage and very small detail such as where they need a tower to have coverage in a certain building.

“It’s massive to put billions of dollars into the rollout of this and so it’s a really important for them to make sure they’re spending their money appropriately,” said Mostak.

A partner and investor is NVIDIA. OmniSci has been working with Charter Communications, a large American carrier, that is building a network where subscribers join their Wi-Fi access points to connect to Charter’s Wi-Fi.  “We are to put Wi-Fi spots in locations so people have to offload onto other carriers spots and they pay for that,” said Mostak. “This ensures people can stay on their network and get better quality service also.”

https://www.omnisci.com/demos/ships/#/dashboard/1

In the auto industry, data is streaming off autonomous vehicles, sometimes at 8,000 variables multiple times per second, with a very high precision GPS. “They’re trying to figure out things like people swerving in the road, and we need to alert them that there may be an obstruction in road,” said Mostak. “There may be geospatial planning through road planners so people can see where there are peak areas of traffic and plan infrastructure accordingly. The use cases are: when do people grab the wheel with a semi self-driving car? Does it tend to happen in certain places in the road, or a certain time of day when light conditions differ? It’s like a Q&A for autonomous systems, visualize where there are hiccups in system so they can improve it.”

A group of ex-FEMA people want to do disaster response in real time and want to bring in hundreds of different feeds. “It could be social media, data from the power grid, power lines are down, etc., could be emergency responders’ locations, for the next Katrina. Effectively these teams can see in real time where they need to deploy resources.”

OmniSci shares joint customers with Esri, where many customers are looking for a faster Esri.

“Esri is great but try to scale 100,000 points that can really dramatically slow the system and cause issues,” said Mostak. “Some of the standard connectivity around WMS  with our backend can actually serve up real time rendering in ways that Esri can consume. There are more integrations we’re doing together and we can go to market in a coordinated fashion a thousand times faster and more scalable.”

Geospatial used to be more of a niche, Mostak said, where issues such as where utilities planned to put their light poles, has become a big data problem. IoT, such as cell phones, satellite data, anything telematics from a plane, boat or car, social media, are implicitly geocodable. If you know someone’s address, you can put in a geocode, store location for retail, for targeted advertising, and store optimization.

Legacy systems can’t scale without GPUs and a focused approach, and Esri is looking for a way to scale as more customers are dealing with hundreds of billions of records.

In many organizations, geospatial data is used with other types of data. GIS tools have not been strong at the other types of data. BI platforms may be very good at the general data analytics but not strong at the geospatial specific visualization analytics. “Neither of them can scale to the big data volumes we’re seeing today,” said Mostak. “We’re seeing roughly 40 percent growth per year. We’re looking for solutions that can derive both those geospatial and non-geospatial contexts.”

Tags: crowdsourcing, data, ESRI, geospatial, GIS, GPS, intelligence, location, mapping, maps, mobile, social media

Categories: 3D Cities, analytics, data, emergency response, field GIS, geocoding, government, GPS, handhelds, location intelligence, public safety, retail, spatial data, transportation

This entry was posted on Thursday, July 4th, 2019 at 9:25 pm. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.