Technologists have constructed distributed programs designed to course of quite a lot of information sorts for a lot of use instances. We now have key-value shops, relational databases, doc databases, graph databases, and even time-series database. However there’s one information and database sort that has largely eluded the palms of expert builders: geospatial information. The parents at Wherobots who’re behind the Apache Sedona mission wish to treatment that scenario.
As an affiliate professor of laptop science at Arizona State College within the 2010s, Mohamed “Mo” Sarwat taught his college students concerning the several types of databases and distributed programs. He lined the varied attributes of the programs, equivalent to Apache Spark and Apache Flink, together with their strengths, weaknesses, and the tradeoffs that inherent on this line of labor. However there was one thing lacking when it got here to geospatial information.
“I checked out all of the programs that I taught, and all of the programs that I constructed and researched in my coaching within the discipline, and none of them handled geospatial as a first-class citizen,” Sarwat tells BigDATAwire. “They’re general-purpose programs, which is nice…However they didn’t present help for geospatial information, for any facets of the bodily world, although a lot of the information exists out there’s collected from the bodily world.”
That’s to not say there have been no functions designed for geo-spatial information. There are a selection of standard geographic data programs (GIS) functions available on the market. Nevertheless, whereas these GIS apps are extensively adopted, they usually don’t present the type of distributed information administration and processing capabilities that as we speak’s large geospatial information calls for, Sarwat says.
“I used to be taking a look at these two worlds,” Sarwat continues. “You had the geospatial world on one hand and the information and the infrastructure work then again, and so they have been talking completely different languages and have been going too many alternative instructions.”
Confronted with an absence out there, Sarwat and his ASU colleague, Jia Yu, who was within the PhD program, did what untold variety of technologists have performed earlier than them: They determined to construct it themselves.
A New Geospatial Framework
In 2017 at ASU, after in depth trial and error, the pair launched a framework referred to as GeoSpark that prolonged the Apache Spark framework with help for geospatial information and processing.
The software program is designed to effectively ingest, rework, and course of giant quantities of geospatial information, equivalent to that generated by satellites, GPS, telephones, cameras, and different sensors.
“It’s Apache Spark, however for bodily world information,” Sarwat says. “As we have been stepping into this type of area, we tried lots of issues, and it didn’t work out. That’s why the market did chunk on it [Sedona], as a result of it didn’t exist in any respect. We lastly discovered one thing that may assist us do this, and that’s why there’s lots of traction for the software program.”
Apache Sedona capabilities as a scalable information warehouse for geospatial information. Fashionable GIS instruments capabilities like enterprise intelligence instruments that lets customers work together with geospatial information in a really detailed approach, however which lack the underlying distributed engine that allows customers to work with very giant geospatial information units.
Builders can make the most of Apache Sedona by way of normal software programmer interfaces (APIs) for Python, which is the preferred approach to entry Sedona, or optionally by way of Spatial SQL, which is an extension of the SQL normal. The open supply mission additionally contains a software program improvement package (SDK) that Java and Scala builders can incorporate into their work. 
There are intricacies to dealing with geospatial information that different forms of distributed engines don’t face. As an illustration, it’s very troublesome to kind and index geospatial information, Sarwat says.
“Numerous this information is definitely polygonal geometries which might be very, very intricate,” he says “Consider boundaries–not even static boundaries, like state or counties. I’m speaking about boundaries of buildings. I’m speaking about even transferring boundaries, off automobiles or off transferring objects. It’s not simply an X column and a Y column, for instance, in a desk. It’s rather more difficult than that.”
Processing these boundaries includes filtering objects and figuring out how they intersect with one another. These geometric computations are very compute intense, and it simply doesn’t work with conventional computing paradigms, Sarwat says.
“It’d work, however will probably be very gradual, very inefficient, and will not even scale to the dimensions of the information and the dimensions of compute to run on that information,” he says.
Enter Wherobots
The downloads of GeoSpark began at just some tons of at first, nevertheless it rapidly cranked up into the hundreds and shortly the hundreds of thousands. In 2020, Sarwat and Yu submitted Sedona to the Apache Software program Basis, and as of July 2025, Apache Sedona had been downloaded 15 million instances. The uptake shocked them.
“To be utterly sincere, after we launched it as teachers on the college, me and my college students, thought perhaps like just a few different folks the world over and different universities would begin utilizing it,” Sarwat says. “We realized there’s a hole, however we didn’t notice how large of a spot that was. The market was very thirsty for a expertise like that.”
In response to the rampant enthusiasm for his or her mission, Sarwat and Yu did what an untold variety of technologists have additionally performed by means of historical past: They determined to create an organization round it. In 2022, they co-founded Wherobots to ship a hosted model of Apache Sedona, a la the connection Databricks initially had with Apache Spark.
As an alternative of making an attempt to run geospatial workloads as person outlined capabilities (UDFs) in a knowledge warehousing setting, equivalent to Oracle, Databricks, or Snowflake, they’ll run the workload as a regular operate in a Apache Sedona cluster and get large efficiency positive factors. In the event that they transfer their Sedona workload to Wherobots serverless cloud, which options greater than 300 pre-built raster and vector capabilities equivalent to map matching, geostatistics, and map tiles, they’ll see one other 30% to 50% in efficiency positive factors, Pruden says.
Huge Geospatial Use Instances
The beauty of bettering the processing of geospatial information is the variety of functions that may be constructed. From insurance coverage and actual property to logistics and social media, there are all kinds of the way geospatial information may be integrated into an software. Because the variety of information factors goes up, so too does the load on the underlying information infrastructure, which is the place Apache Sedona and its Apache Spark-based information processing capabilities are available in.
For instance, the last-mile supply downside is a large problem for corporations like Amazon that try to ship packages to billions of individuals all over the world. The amount of deliveries instances the dimensions of the supply squad instances the dimensions of the developed world equals a serious computational downside for Amazon. However due to Apache Sedona, Amazon is ready to deal with the problem.
Amazon offered on their use of Apache Sedona through the AWS re:Invent convention final yr, says Wherebots Director of Advertising Ben Pruden.
“They’re taking on this information from satellite tv for pc imagery, from aerial imagery, like from drones, from GPS traces coming off of their vehicles which might be taking packages to your home, streetside imagery,” Pruden says, “and so they deliver it right into a system that’s largely powered by Sedona to do a really giant graph after which conflation of their information units to take care of these actually up to date maps of the world.”
Apache Sedona is vital for offering detailed map representations that Amazon drivers use to get immediately the suitable place inside prospects’ homes or condominium advanced, Pruden says. “Or if you happen to’re out in a rural space, perhaps there’s a very lengthy driveway that isn’t apparent that it’s a must to determine the place to drive down,” he says. “They’re getting ready all that information throughout the whole planet after which serving that again in order that their drivers can put it to use.”
One other early adopter of Apache Sedona is Overture Maps Basis, which is constructing an open reference map of the whole world. The group began out operating Apache Sedona on Spark, and prior to now six months has been migrating to the Wherobots platform, Pruden says.
“Organizations like Overture and quite a lot of others are utilizing each our open supply and in addition more and more Wherobots to research and ship perception and create information merchandise for information concerning the bodily world,” he says.
Whereobots, which relies in Scottsdale, Arizona, continues to be ramping up cloud operations on AWS, which the corporate says is an in depth accomplice. The corporate raised $21.4 million in enterprise funding in November. Within the meantime, the corporate is trying ahead to the following frontier for geospatial information: integration with AI.
“Up to now, AI has been actually good with language, responding to us. We work together with it, and it’s unbelievable,” Sarwat says. “However up till as we speak, AI doesn’t have an excellent data of the bodily world, like the right way to purpose concerning the bodily world normally…And that is what we’re specializing in, on how can we offer a knowledge engine that may make that type of information AI prepared and make it very comprehensible by AI.”
Associated Objects:
5 Methods Huge Geospatial Information Is Driving Analytics Within the Actual World
How Geospatial Information Drives Perception for Bloomberg Customers