Aerospike update aims to improve database performance
With the addition of expression indexes that enable users to streamline the data discovery process, the vendor is providing capabilities that could force competitors to respond.
Aerospike recently launched expression indexing capabilities aimed at reducing the amount of memory that applications use when searching databases for relevant data, enabling users to lower operational costs of machine learning and AI development.
Database indexing is a means of structuring data to make it easy to locate within a broader database table. Expression indexes are specialized indexes that further hone searches by narrowing them down to functions -- also known as expressions -- applied to one or more columns within database tables. Examples of expression indexes include breaking down data within columns based on characteristics such as code patterns or data and time.
With expression indexes, databases can limit scans based on user prompts to only data that matters, which avoids the need to scan all records within a table. Their benefits include more efficient queries that reduce the amount of memory needed to discover data, subsequently lowering the cost of such searches.
Aerospike's expression indexing capabilities were made generally available as part of Aerospike 8.1, the latest version of the vendor's database platform, and are a valuable addition for the vendor's users, according to Stephen Catanzano, an analyst at Enterprise Strategy Group, now part of Omdia.
Aerospike's expression indexes … mean that indexes can be smaller because they exclude records that don't match the expression, which in turn reduces memory usage, making queries run faster. That's very significant for customers.
Stephen CatanzanoAnalyst, Enterprise Strategy Group
"Aerospike's expression indexes … mean that indexes can be smaller because they exclude records that don't match the expression, which in turn reduces memory usage, making queries run faster," he said. "That's very significant for customers."
Based in Mountain View, Calif., Aerospike is a multimodal database vendor whose platform supports data types including documents, key-values, graphs and vectors.
New capabilities
Aerospike's addition of expression indexing capabilities was driven by rising interest in developing AI and machine learning applications, according to Srini Srinivasan, the vendor's founder and chief technology officer.
OpenAI's November 2022 launch of ChatGPT marked significant improvement in generative AI technology. Since then, enterprises have increased their investment in developing AI and machine learning tools.
Meanwhile, with growing volumes of relevant, high-quality data as the foundation of AI and machine learning applications, most data management vendors have reacted by adding capabilities aimed at simplifying the process of discovering data and feeding it into development pipelines.
"Due to the fast adoption of AI and machine learning applications, the scale and functionality needed to access data is becoming increasingly sophisticated," Srinivasan said. "Expression indexes are one way of indexing large amounts of data. … Essentially, [they] will increase the performance of the system at a large scale for these complex AI applications.
Aerospike is not the first database vendor to provide support for expression indexes.
Tech giants Oracle, IBM and Microsoft are among those offering expression indexing capabilities within some of their relational database products. Among NoSQL database specialists, however, Aerospike is perhaps unique in providing full support for expression indexing, according to Catanzano.
As a result, competing NoSQL database vendors such as Redis and Couchbase might need to respond.
"Data centers are running out of space and hitting physical limits, so any solution that can offer better performance with fewer resources will be quite attractive and put migration pressure on other players that do not have this ability," Catanzano said.
Even beyond competition, other NoSQL database vendors might need to add expression indexing capabilities based on user demand, Catanzano continued.
"The specific use case of this capability in NoSQL data stores like Aerospike seems to be directly related to the massive growth in compute requirements for AI workloads," he said. "In this context, what was previously a nice-to-have feature may have become a need-to-have" feature.
Matt Aslett, an analyst at ISG Research, similarly said that expression indexing is a potential way for Aerospike to stand apart from competitors as database providers attempt to meet the demands of increasing AI and machine learning workloads.
While indexing is a fundamental component of any database management system, vendors can provide different types of indexing that align with their architectural approaches and the use cases they serve, he noted.
"New indexing approaches are becoming increasingly important given the volumes of data that need to be processed at speed to fulfill the inferencing requirements of intelligent operational applications driven by AI," Aslett said. "Aerospike's expression indexes approach matches the company's focus on use cases requiring high-performance real-time read-write capabilities."
Looking ahead
Following the launch of support for new types of OLTP applications and expression indexing capabilities in February, database performance will continue to be a focus for Aerospike over the final months of 2025, according to Srinivasan.
Specifically, the vendor plans to add operational improvements through its Kubernetes Operator as well as new performance and scale features that address customer demands, he said.
Catanzano, meanwhile, recommended that Aerospike add integrations with AI frameworks to better enable developers to easily build AI and machine learning. In addition, he suggested that Aerospike add AI-powered automation capabilities to expression indexes that further improve data discovery.
"Aerospike may want to consider AI-powered adaptive indexing that can automatically optimize expression indexes based on query patterns and data characteristics," he said. "They may also want to consider building a library of pre-built expression templates for common AI ad machine learning workloads.
Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.
