your123 - stock.adobe.com

Qdrant update adds security measures for AI development

The vector database specialist's update includes features that enable secure AI development such as role-based access control and an API for managing cloud deployments.

Vector database specialist Qdrant's latest update includes security, automation and real-time observability capabilities aimed at enabling enterprises to safely scale AI development, including generative AI.

Vectors are numerical representations of data -- automatically assigned by algorithms -- that enable users to give structure to unstructured data such as text and images so that such data can be searched for and discovered. In addition, vectors enable similarity searches so that relevant data can be found to inform models and applications.

Generative AI requires high volumes of relevant, high-quality data for accuracy. As a result, vectors have gained popularity in the two years since OpenAI's launch of ChatGPT marked significant improvement in generative AI capabilities and spurred a wave of AI development.

The latest Qdrant Cloud update, unveiled on Tuesday, features an API that simplifies managing cloud deployments, security features including managing hybrid cloud deployments that enable organizations to keep sensitive information on-premises, and an integration with OpenMetrics to observe performance and usage in real time.

With data security a significant concern for many organizations developing generative AI applications, Qdrant's update addresses important enterprise needs, according to Donald Farmer, founder and principal of TreeHive Strategy.

Development teams are placing increasingly high demands on vector databases to enable more customized multimodal and agentic AI capabilities.
Donald FarmerFounder and principal, TreeHive Strategy

"Development teams are placing increasingly high demands on vector databases to enable more customized multimodal and agentic AI capabilities," he said. "However, they are also very concerned with issues of privacy and data sovereignty, so ... this Qdrant release, especially with hybrid cloud, plays well to those needs."

Based in Berlin, Qdrant is an open source database vendor whose platform enables users to store and search vectorized data.

Vendors such as Pinecone and Redis are also vector database specialists. Meanwhile, more broad-based data management vendors, including tech giants AWS and Oracle, have made vector capabilities part of their database strategies.

New capabilities

With generative AI development requiring large amounts of data, it is imperative that data management tools, including vector databases, maintain performance speed as workload demands increase.

As part of a push to provide enterprise-grade vector database capabilities, Qdrant in January added support for GPU-powered vector indexing. GPUs deliver better performance and cost efficiency than CPUs.

A chart shows the differences between vector search and traditional search.

In addition to performance, security is a major concern for many enterprises as they develop and deploy AI applications.

Organizations need to combine their proprietary data with generative AI models such as OpenAI's GPT-4o and Google Gemini to develop generative AI applications that understand their operations. When data is combined with generative AI models, it is moved from storage in a database through pipelines that deliver it to the model for training.

Public clouds can be more susceptible to data breaches than hybrid clouds and on-premises private clouds. Some organizations, therefore, prefer to develop their AI pipelines in hybrid or on-premises environments.

After addressing performance concerns with its January update, Qdrant is now continuing its push to provide a more enterprise-grade platform by addressing security concerns, according to André Zayarni, Qdrant's CEO and co-founder.

"Qdrant's focus has always been on high-performance vector search," he said. "Now, with enterprises moving AI workloads into large-scale production, these applications must also meet stringent security, compliance and operational requirements."

Regarding the impetus for addressing data security following performance improvements, user feedback was a main motivator, Zayarni continued.

"Customer and community feedback ... always is a core driver for our roadmap," he said. "To be specific, the primary push for these features came from the need to scale AI applications from early production to large-scale enterprise deployment while ensuring security, efficiency and reliability."

Specific new Qdrant Cloud features include the following:

  • Cloud API, a feature that simplifies managing cloud deployments by enabling administrators to manage clusters, authentications and configurations.
  • Automated deployments and scaling the Cloud API, and support for infrastructure-as-code workflows such as Terraform.
  • Cloud role-based access control (RBAC) that enables administrators to set fine-grained permissions for managing clusters, billing and hybrid deployments.
  • An integration with OpenMetrics, a suite of observability tools built on the open source Prometheus monitoring framework, so that users can track query performance and latency, request volumes, CPU usage, and other metrics.

While each feature aims to provide users with enterprise-grade capabilities, cloud RBAC is perhaps the most significant, given that the number of people within organizations accessing and using vectorized data is rapidly increasing, according to Zayarni.

"Data engineers, machine learning teams, IT security and even business users need controlled access to vector databases, without sacrificing security or agility," he said. "Cloud RBAC [ensures] the right people have the right level of access."

Farmer, meanwhile, highlighted the Cloud API because it simplifies managing complex hybrid deployments.

"As teams roll out their own hybrid solutions, it's good for teams to be able to deploy infrastructure as code, especially when they are creating their own hybrid solutions, because they need to be able to prototype, deploy and test numerous configurations to get it right," he said.

In addition, Qdrant's recent addition of support for GPUs was significant, Farmer continued. There is currently a shortage of GPUs. Qdrant enables customers to use GPUs from any vendor, so provided that customers can find available GPUs, they can use those to run Qdrant workloads.

"I would normally say it gives users the freedom to choose, but right now, it's more the ability to work with whatever they are lucky enough to get their hands on," Farmer said.

Next steps

Qdrant began in 2021 as an open source project. Qdrant Cloud, a commercial version of the vendor's platform, was later launched in February 2023.

Now, as evidenced by the recent additions of support for GPUs to improve performance and the new features that add security needed in industries such as finance and healthcare, Qdrant is improving its platform to make it more enterprise-grade.

Toward that end, developing more ways to enable deployment flexibility is a priority, according to Zayarni. In addition, further performance improvements to lower latency, reduce manual intervention and enable users to develop deeper insights are part of Qdrant's roadmap.

"We will continue [investing in] enterprise-grade AI infrastructure," Zayarni said.

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.

Dig Deeper on Data management strategies