Aerospike Adds GeoSpatial Data to Speedy NoSQL
Organizations that are looking to leverage location data in their operational apps have another option in Aerospike, which today announced support for geospatial data in its NoSQL database.
Aerospike is a key-value data store designed to run within a combination of RAM and Flash storage. The database, which essentially bypasses the file system to speed up processing, is widely used among ad tech companies thanks to its capability to process mixed read and write workloads at high speeds with low latencies.
Adding support for geospatial data—specifically, support for geoJSON, Google’s S2 library, and geo-hashing—will give customers new capabilities to keep track of objects moving about the Internet of Thigns (IoT), says AeroSpike vice president Alvin Richards.
“Geospatial data has been underserved,” Richards tells Datanami. “While there are databases out there, NoSQL and relational, that provided geospatial capabilities, they don’t provide the kind of velocity that Aerospike customer are normally used to.”
Aerospike databases often run upwards of 30TB in size, are asked to process up to 100 billion objects, and tackle a million transactions per second. With geospatial data sitting alongside other supported data types—like strings, integers, binaries, and lists—Aerospike customers will be able to track movement in the world like never before.
“What this really means is you can built geospatial at scale with high throughput, mixed reads and write with very latency, and that opens up applications and use cases in this Internet of Moving Things,” Richards says. “We’re seeing this as the next trend in geospatial data where you need this kind of throughput, low latency and reliability.”
Geospatial data will be very beneficial to customers in the ad-tech business, Richards says. “You can imagine wandering around a region or a store. They want to set up target campaign to your location based on all the other data they have. They want to take geolocation as part of that,” he says.
“Also in terms of segmenting markets in order to understand how devices are used within a household and how devices are used outside the household. Geo-spatial is another component of their machine learning and algorithms that they want to start processing at the speed they are [currently processing] cookies.”
Aerospike 3.7 also brings a new list manipulation capability that will help customers accelerate their business processing, whatever that might be. Instead of pulling the lists of data out of Aerospike to manipulate them, which takes time and bandwidth, Aerospike customers can simply manipulate the list as it sits on the database server.
“What we find with customers is they need a richness of data,” Richards says. “You can’t have a vanilla record like you get in relational database. You need to supplement it, whether it’s geospatial or scaler attributes, or lists of information, whether time-series [data] or ranking or various other ways to augment that record.”
Fraud detection is one area that will be enhanced with the list manipulation capability. “In order to make decisions on whether you accept the credit payment or not, you need to look at sets of information, patterns of information,” Richards says. “You want to combine your offline model, feeding back into real time and blend the data together so you can make a better policy decision.”
Finally, version 3.7 brings new adaptive clustering capabilities that improve how Aerospike runs in environments that have widely variable performance characteristics, such as cloud environments.
Aerospike offers a community edition of its database that is open source, and sells licenses for an enterprise version that has more advanced capabilities.