Jason Stamper
Jason Stamper (European Research Manager, Data Intelligence, Integration and Platforms for AI, IDC Europe)

In the realm of data platform decision-making, organizations typically consider several dimensions when making their choices. These encompass aspects including functionality, performance, scalability, cost, flexibility, and alignment with specific use cases.

The following are some of the key criteria of data platform decision-making. It’s worth noting that one of the most-hyped databases right now, in support of AI, is vector databases. We’ll explain why.

Data Model and Schema Flexibility

Organizations assess whether the database supports their data model requirements. Some may need the flexibility of schema-less or schema-on-read models. Others may require the rigidity of a relational model. The choice depends on factors like the structure of the data — is it simply rows and columns of numbers? is it a mix of images, videos, and documents? — and the need for agility to adapt to evolving schemas.

With the rise of Hadoop, many organizations began to store more of their data for analysis, now or in the future. Open source Hadoop offered data storage on commodity hardware, while more traditional proprietary data warehouses were almost certainly far more expensive. The trouble is that Hadoop lacked a schema — a structure for the data warehouse — making it harder to extract the data when you need it (though workarounds are now available).

As mentioned above, vector databases are garnering a lot of attention because of the rise of AI. Reasons for this include:

  • Efficient Similarity Search or Nearest Neighbor Search: Vector databases are optimized for nearest neighbor search, a fundamental operation in many AI applications such as recommendation systems, image retrieval, and natural language processing.
  • High-Dimensional Data Handling: AI models, especially in NLP and computer vision, generate high-dimensional embedding vectors. Vector databases can store and index these vectors efficiently, allowing for rapid querying and analysis.
  • Semantic Search: By leveraging embedding vectors that capture semantic information, vector databases enable more intuitive and relevant search results compared to traditional keyword-based searches.
  • Multimodal Search: Vector databases support the integration of various data types (text, images, audio) by converting them into a common vector space, allowing for unified search and analysis across different modalities.
  • Clustering and Classification: Vector databases support operations like clustering and classification directly on the stored vectors, aiding in tasks such as customer segmentation, anomaly detection, and pattern recognition.

Vector databases are pivotal for AI because they provide the necessary infrastructure to store, manage, and query high-dimensional vectors efficiently. This capability is foundational for enabling fast, scalable, and intelligent AI systems across various applications and industries.

Performance and Scalability

Performance considerations include factors like query speed, throughput, latency, and concurrency. Scalability relates to the ability of the database to handle growing volumes of data and increasing user loads without sacrificing performance. Organizations evaluate whether the database can scale horizontally (adding more servers) or vertically (upgrading existing servers).

Consistency and Durability

Consistency refers to the degree to which data remains in a consistent state across distributed systems, especially in the event of failures or concurrent transactions. Durability relates to the ability of the database to ensure that committed transactions persist even in the face of system failures. Organizations weigh the trade-offs between consistency, availability, and partition tolerance based on their application requirements.ACID is key to relational, transactional databases. ACID compliance refers to a set of properties that ensure the reliability and integrity of transactions in a database system. The acronym ACID stands for Atomicity, Consistency, Isolation, and Durability, each representing a fundamental aspect of transaction management.

ACID is spoken of in somewhat hushed tones by NoSQL vendors. When pushed, some will say they offer “ACID-like” compliance. For many modern use cases, ACID-like is good enough. But speak to a database developer dealing with transactional systems — like core banking systems — and they will tell you their regulators and other stakeholders require “pure” ACID compliance. Compliance with ACID standards can help organizations meet regulatory requirements and maintain data governance.

Data Integrity and Security

Organizations prioritize databases that provide robust mechanisms for maintaining data integrity (e.g., through constraints, transactions, and validations) and enforcing security (e.g., encryption, authentication, authorization, and auditing). Compliance with regulatory requirements such as GDPR, HIPAA, or PCI-DSS may also influence database selection.

Ease of Development and Maintenance

This encompasses factors like developer productivity, ease of learning, availability of tools and libraries, and support for programming languages and frameworks. Organizations seek databases that streamline the development process, facilitate debugging and monitoring, and minimize operational overhead.

Total Cost of Ownership (TCO)

TCO considerations include both up-front costs (e.g., licensing fees, hardware costs) and ongoing expenses (e.g., maintenance, support, scaling). Organizations evaluate databases based on their ability to deliver value relative to their costs over the entire life cycle of the system.

Ecosystem and Integration

Organizations assess the database’s ecosystem, including its compatibility with existing infrastructure, integration with other systems (e.g., data warehouses, analytics platforms, the cloud), availability of third-party tools and services, and community support. Integration capabilities influence factors such as data migration, interoperability, and extensibility. There is also the issue here of deployment venue: on premises, in the cloud, hybrid, or multicloud.


By evaluating data platforms along these lines, organizations can make informed decisions that align with their business objectives, technical requirements, and constraints. Vector databases are certainly one of the hottest tickets in town in support of AI — but different use cases have different priorities.

Spread the love