Pinecone launches serverless edition of its vector database on AWS – SiliconANGLE

Vector database startup Pinecone Systems Inc. said today the serverless version of its platform is launching in general availability on Amazon Web Services Inc.’s public cloud platform after a successful beta testing phase.

Pinecone is the creator of an advanced vector database that can dynamically store, transform and index billions of high-dimensional data points, enabling it to respond rapidly and accurately to queries such as nearest-neighbor search.

The company launched the original, server-based version of its vector database back in 2021, saying that it was aimed at artificial intelligence and machine learning applications. The main purpose, it said at the time, is to provide a way for developers to store and consume the enormous amounts of data required for AI training. With the emergence of generative AI in the last couple of years, Pinecone’s vector database has gotten significant attention thanks to its ability to act as a long-term memory store for AI chatbots.

Unlike relational databases that store structured information in rows and columns, vector databases are designed to house unstructured data, which is stored as high-dimensional data points represented by vectors, or an array of numbers. One of their primary functions is to enable similarity searches that aim to find vectors that are most similar to a given query vector, which is done using techniques such as cosine similarity or Euclidean distance.

With the launch of Pinecone Serverless on AWS, the company is leveraging a cloud computing execution model where the cloud provider dynamically manages the allocation and provisioning of servers. The advantage of this model is that developers don’t have to waste time and energy provisioning the underlying cloud infrastructure for the database, so they can bring their applications to market faster. Pinecone also claims that the serverless approach can reduce the underlying costs of running its vector database by up to 98%.

According to Pinecone’s own research, making unstructured data available for context retrieval can help AI models to reduce the frequency of unhelpful responses, or those that incorrectly respond or simply fail to respond to a question, by 50%. In other words, large language model developers can significantly improve the quality of their products by making more data accessible to them.

One challenge associated with vector databases is that storing large amounts of information within them can quickly become prohibitively expensive. But Pinecone addresses this by storing and searching AI-generated representations of unstructured information that encapsulate the meaning of the original content in a machine-readable format.

This not only improves performance, but makes it more cost-efficient for keyword-based searches. Moreover, the serverless architecture automatically provisions the storage resources to save time and money for users.

Another major advantage of Pinecone Serverless is that its architecture separates reads, writes and storage to further reduce costs. It relies on vector clustering that sits atop blob storage — that is, storage for unstructured data — for low-latency searches of data stores, using purpose-built indexing and retrieval algorithms. There’s also a multitenant compute layer for on-demand availability, which enables Pinecone Serverless to support thousands of users simultaneously.

Pinecone Serverless customers will be able to take advantage of AWS PrivateLink, now in preview, which provides a way for them to connect to the cloud database over a private link, avoiding the risks of sending traffic over the public internet.

The company said more than 20,000 organizations have already used Pinecone Serverless since it became available in beta in early February. Moreover, it has a fast-growing ecosystem, the company said, instigating with a range of AI development tools such as Anyscale, Amazon Bedrock, Confluent, Langchain, Mistral, Monte Carlo, Qwak, Together.ai and Vectorize.

Pinecone Serverless is generally available now in AWS’ us-west-2, us-east-1, and eu-west-1 regions, with more regions to come in the future. Though it’s exclusive to AWS for now, Pinecone said it will expand its serverless offering to Microsoft Azure and Google Cloud Platform later this year.

Image: GDJ/Pixabay