Vector databases are fundamental to artificial intelligence; they provide the triangulated reasoning (or more) algorithmic connections that enable us to build inference. But as important as they are, users (that’s software engineers… and actual users) may assume that these data behemoths are blessed with infinite resource power.
Of course, no technology runs without an appropriate level of supervisory mechanics.
As such, AWS is now detailing serverless GPU acceleration and auto-optimization for vector index in Amazon OpenSearch Service, the company’s managed service for running, scaling and monitoring OpenSearch clusters.
Optimal Trade-Offs
This acceleration and auto-optimization is engineered to help data science teams build large-scale vector databases faster and run them at lower costs due to their ability to automatically optimize vector indexes for optimal trade-off points between search quality, speed and cost.
So how does it work?
AWS says that in the area of GPU acceleration, cloud-native engineers can now build vector database services up to 10 times faster at a quarter of the indexing cost when compared to non-GPU acceleration. If that sounds like a slightly over-exact and potentially contrived speed rating, we can understand the boost power by saying it means we create a billion-scale vector database in under an hour.
Fond of telling us that this will enable nice things like “time-to-market and innovation velocity”, but really it’s just straight play to encourage the adoption of vector search.
Vector Fields in Space & Time
The auto-optimization controls here mean engineers can tinker with the controls to find the best balance between search latency, quality and memory requirements for any given vector field (the number of points in space being analyzed by the vector database, also potentially including their direction) without needing vector expertise.
This optimization is also said to help achieve better cost savings and recall rates when compared to default index configurations, while manual index tuning can take weeks to complete.
“[Users] can use these capabilities to build vector databases faster and more cost-effectively on OpenSearch Service. You can use them to power generative AI applications, search product catalogs and knowledge bases and more. You can enable GPU acceleration and auto-optimization when you create a new OpenSearch domain or collection, as well as update an existing domain or collection,” detailed AWS, in a technical blog.
GPU Acceleration
When users enable GPU acceleration on their OpenSearch Service domain or Serverless collection, OpenSearch Service automatically detects opportunities to accelerate vector indexing workloads. This acceleration helps build the vector data structures in OpenSearch Service domain or Serverless collection.
Engineers do not need to provision the GPU instances, manage their usage or pay for idle time. OpenSearch Service securely isolates accelerated workloads to a team’s domain or Amazon Virtual Private Cloud (Amazon VPC) instance within their account.
Users pay only for useful processing through the OpenSearch Compute Units (OCU) Vector Acceleration pricing.
Ingestion Digestion
Cloud engineers can also use the new vector ingestion feature to ingest documents from Amazon Simple Storage Service (Amazon S3), generate vector embeddings, optimize indexes automatically and build large-scale vector indexes. During the ingestion, auto-optimization generates recommendations based on a team’s selected vector fields and the related indexes of their OpenSearch Service domain or Serverless collection.
Manually configuring mappings is a thing of the past, at least at this level of the AWS universe.
