Analysts believe that very few buyers will be able to get their hands on Pinecone’s new serverless vector database, called Pinecone Serverless.
“If you can get the same functionality from the database you’re already using and managing your data with, why would you set up and manage a separate database, even if it has the scalability benefits of serverless?” says Principal Doug Henschen. said. Analyst at Constellation Research.
Outside of mainstream vector databases such as Milvus, Weaviate, and Chroma, vector embedding and search features have either already been added or will soon be available to database service providers such as MongoDB, Couchbase, Snowflake, and Google BigQuery, among others.
“The addition of vector embeddings and search makes it difficult for startup, vector-only databases to develop a large market,” Henschen said.
According to experts, vector databases and vector search are two technologies that developers use to convert unstructured information into vectors; it is now more commonly referred to as docking.
The scalability advantage of vector search has also helped it gain favor among developers building generative AI-based applications, as you can feed more data into a large language model (LLM) and the model can produce more accurate answers when needed. This makes top layer implementation more efficient.
But the chief analyst said he wasn’t convinced that vector databases like Pinecone, which are drawing more attention to developers and data scientists working on AI, would force businesses to pay for an additional database service used solely for the development of AI. based applications.
Flat IT budgets may raise Pinecone concerns
Moreover, the launch of Pinecone Serverless comes at a time when organizations’ IT budgets continue to remain stagnant.
“Although there is a lot of interest in generative AI, budgets are not yet increasing accordingly,” said Tony Baer, principal analyst at dbInsight.
“Flat budgets can be attributed to the immaturity of the field; Selection of everything from tools to underlying models to runtime services is still in its infancy, and aside from co-pilots and natural language querying, companies are still on the learning curve of identifying winning use cases,” Baer added.
In addition to fueling demand for productive AI, Pinecone expects its new serverless database to help organizations reduce the cost and need to manage infrastructure.
The company stated that cost reduction is possible by separating reading, writing and storage, adding that the database aims to reduce latency by adopting an architecture in which vector clustering lies on top of blob storage.
According to Baer, the new vector indexing gives Pinecone an advantage over other vector and operational databases. Pinecones support nearly a dozen index species, the analyst said.
Baer says that the serverless feature of the database is also the need of the hour.
“The nature of rollback augmented generation (RAG) workloads is that they have the characteristics of any query-driven workload (think analytics) in that they are inherently bursty. Without serverless, Baer says, customers will typically be able to provision “just in case” capacity that is likely to fall silent.” He said he should.