# Build OpenSearch Vector Indexes faster with NetApp Instaclustr and NVIDIA cuVS [Blog](/blog/)>[Technology](/blog/category/technical/)>Build OpenSearch Vector Indexes faster with NetApp Instaclustr and NVIDIA cuVS Build OpenSearch Vector Indexes faster with NetApp Instaclustr and NVIDIA cuVS ============================================================================== September 21, 2026 | By [ Alex Bunday](https://www.instaclustr.com/blog/author/abunday/) [ ](https://x.com/intent/tweet?text=Build%20OpenSearch%20Vector%20Indexes%20faster%20with%20NetApp%20Instaclustr%20and%20NVIDIA%20cuVS&url=https://www.instaclustr.com/blog/build-opensearch-vector-indexes-faster-with-netapp-instaclustr-and-nvidia-cuvs/) [ ](https://www.linkedin.com/shareArticle?mini=true&url=https://www.instaclustr.com/blog/build-opensearch-vector-indexes-faster-with-netapp-instaclustr-and-nvidia-cuvs/&title=&summary=Build%20OpenSearch%20Vector%20Indexes%20faster%20with%20NetApp%20Instaclustr%20and%20NVIDIA%20cuVS&source=) **GPU-accelerated vector index builds for managed OpenSearch, designed to deliver faster performance across AWS, Azure, and Google Cloud.** NetApp today announced the general availability of Vector Index Acceleration for OpenSearch on the NetApp® Instaclustr Managed Platform. Leveraging open source [NVIDIA cuVS,](https://developer.nvidia.com/topics/ai/generative-ai/cuvs) this new feature is designed to enable GPU-accelerated vector index building. Customers can easily add a managed GPU node via the Instaclustr Console, allowing vector index builds to run on the GPU while searches continue seamlessly on existing CPU nodes. Internal NetApp Instaclustr benchmark tests show that adding just one managed GPU node reduced total vector index build time by up to 6.7×, with the force merge step accelerated by up to 24×. Built in collaboration with NVIDIA, Vector Index Acceleration is a fully managed, multi-cloud capability and is now available for provisioning via the console, API or Terraform. “Customers consistently share that while vector search is effective, scaling index builds has become a major bottleneck in time and infrastructure cost,” said Sayan Saha, Senior Director of Product Management for NetApp Instaclustr. “Vector Index Acceleration removes this barrier, delivering NVIDIA GPU performance as a managed cluster node—without the burden of driver management, patching, or added operational overhead.” What is GPU-accelerated vector indexing for OpenSearch? ------------------------------------------------------- GPU-accelerated vector indexing for OpenSearch uses GPU performance to speed up the process of building vector indexes, while search continues to run on existing CPU nodes. With NetApp Instaclustr Vector Index Acceleration, customers add a managed NVIDIA GPU node to their OpenSearch cluster so vector index builds can complete faster without requiring teams to manage GPU infrastructure themselves. Why vector indexing matters for AI search and RAG ------------------------------------------------- As RAG, semantic search, recommendation engines, and agentic AI move from pilots into production, vector indexes can scale from millions to billions of embeddings. CPU-only index builds can force a trade-off between slower release cycles and higher infrastructure spend. At this scale, building the indexes that make vectors searchable becomes one of the most compute-intensive parts of running OpenSearch, putting pressure on both teams and budgets. Vector Index Acceleration changes that. See how is works in this [2 minute demo](https://youtu.be/HSy_zLLC1h0). What you get ------------ - **Speed:** Cut total vector index build time by up to 6.7×, with the force merge step, the most compute-intensive part of building a vector index, accelerated by up to 24×, based on internal NetApp Instaclustr benchmark testing on Instaclustr-managed infrastructure. - **Cost efficiency:** Improve the cost efficiency of each index build compared with CPU-only approaches, without over-provisioning CPU capacity that sits idle between builds. - **Reduced overhead and simple deployment:** Add a GPU node to your cluster through the Instaclustr Console. NetApp Instaclustr manages drivers, patching, monitoring, security updates, and lifecycle, so teams can benefit from NVIDIA GPU performance without building their own GPU operations practice. - **Multi-cloud freedom:** Deploy managed OpenSearch with Vector Index Acceleration wherever your workloads run. Powered by NVIDIA cuVS ---------------------- Vector Index Acceleration is built on NVIDIA cuVS, the GPU-accelerated vector search library that contributes to the OpenSearch project and the broader search ecosystem. This collaboration gives customers three clear advantages: 1. **Technology designed for scale:** cuVS is actively maintained by NVIDIA and adopted across the vector search ecosystem, giving customers a library designed for scale rather than a one-off integration. 2. **GPU performance, managed end to end:** NVIDIA cuVS delivers the index build acceleration, while NetApp Instaclustr delivers it as a managed GPU node within the customer’s OpenSearch cluster, with the same provisioning, monitoring, billing, and support experience as every other node on the platform. 3. **Production ready:** Co-developed by NetApp Instaclustr and NVIDIA, cuVS helps ensure optimized performance at scale, ready to use through a managed service experience. How does Vector Index Acceleration work? ---------------------------------------- Customers add an NVIDIA GPU node to their OpenSearch cluster through the Instaclustr Console or using an API, using the same provisioning flow as any other node size on the platform. When a vector index build is triggered, the work is routed to the GPU node, where NVIDIA cuVS builds a graph-based index using the CAGRA algorithm, which is designed to take advantage of GPU parallelism. The graph is then converted into a standard HNSW-compatible format. Search continues on the CPU nodes in the cluster, while the vector index can be rebuilt faster and more often. ![vector acceleration index for OpenSearch diagram]() *Figure 1 – Vector data is routed to a dedicated GPU to build a vector index while data nodes continue serving search queries and other operations* Nothing changes about your existing workflow. OpenSearch APIs, query syntax, and the serving path all remain the same. NetApp Instaclustr manages the GPU node, including drivers, patching, monitoring, and lifecycle, so teams get the performance benefit without needing to stand up their own GPU operations capability. “What makes this exciting is how easy it is to implement from a developer’s perspective. It’s not like we have to rebuild our setup. Having that kind of managed power running on the backend is great,” said Brett Bittke, Senior Programmer and Lead App Developer, Your Store Wizards. What this means for customers ----------------------------- Platform and data engineeringBillion-scale vector growth means proportional CPU spend and operational burdenAdd a managed GPU node to the cluster and offload index builds without building an in-house GPU operations practice WhoBeforeAfterPlatform and data engineeringBillion-scale vector growth means proportional CPU spend and operational burdenAdd a managed GPU node to the cluster and offload index builds without building an in-house GPU operations practiceApplication developers (RAG, semantic search)Index builds gate releases and embedding model iterationGo from idea to production faster. Re-index faster and help reduce infrastructure-related delays when shipping AI features.Multi-cloud enterprisesGPU-accelerated vector indexing requires separate infrastructure per cloudUse the same managed capability and consistent workflow across AWS, Azure, and Google Cloud“Rebuilding indexes at scale is incredibly resource-intensive, but these new processing speeds change the game. The ability to complete index refreshes in minutes rather than hours opens up major opportunities for us to keep our search experience incredibly accurate and real-time,” said Brett Bittke, Senior Programmer and Lead App Developer, Your Store Wizards. **Performance and cost** ------------------------ **Internal NetApp Instaclustr using NVIDIA benchmark:** NetApp Instaclustr benchmarked a 3-node OpenSearch cluster (3x r6g.4xlarge) running CPU-only indexing against the same cluster with a single managed GPU node added (3x r6g.4xlarge + 1x A10G GPU, g5.4xlarge). Adding the GPU node cut total build time from 505.3 minutes to as low as 75.0 minutes, up to 6.7× faster, and reduced the force merge step from about 438.9 minutes to as low as 18.1 minutes, up to 24× faster. ![vector index build time chart]() *Figure 2 – Vector index build time and force merge time, CPU-only vs. GPU-accelerated (internal NetApp Instaclustr benchmark, single managed GPU node added)* **Get started** --------------- Vector Index Acceleration for OpenSearch is generally available today on the NetApp Instaclustr Managed Platform for supported OpenSearch versions and cloud regions. Existing Instaclustr OpenSearch customers can add a GPU node to supported clusters through the Instaclustr Console, API or Terraform while new customers can provision OpenSearch clusters with Vector Index Acceleration from the start. Get started with a [free trial](https://console2.instaclustr.com/signup) today. ### Frequently Asked Questions - Can OpenSearch use GPUs for vector indexing? + Yes. With NetApp Instaclustr Vector Index Acceleration for OpenSearch, customers can add a managed NVIDIA GPU node to accelerate vector index builds while search continues on existing CPU nodes. - How does NVIDIA cuVS accelerate OpenSearch vector indexing? + NVIDIA cuVS builds a graph-based index using the CAGRA algorithm, which is designed to take advantage of GPU parallelism. The graph is then converted into a standard HNSW-compatible format. - Why are CPU-only vector index builds a bottleneck? + As AI search, RAG, semantic search, recommendation engines, and agentic AI workloads scale from millions to billions of embeddings, CPU-only index builds can slow release cycles and increase infrastructure spend. - Does Vector Index Acceleration change the OpenSearch workflow? + No. OpenSearch APIs, query syntax, and the serving path remain the same. NetApp Instaclustr manages the GPU node, including drivers, patching, monitoring, and lifecycle. - Is Vector Index Acceleration available across clouds? + Yes. Vector Index Acceleration for OpenSearch is available as a managed, multi-cloud capability across supported OpenSearch versions, cloud regions, and deployment options on AWS and GCP. Azure is not supported at this time. ### About the author **[Alex Bunday](https://www.instaclustr.com/blog/author/abunday/)** | Product Manager Alex Bunday is a Product Manager with a strong focus on security and search within the Instaclustr Managed Platform, driving value through innovative solutions in the open source community. With expertise in managing product lifecycles and engaging stakeholders, Alex leverages data-driven insights to continuously improve and deliver impactful results. His collaborative approach empowers teams to develop and ship enhancements that meet customer needs and strengthen product offerings. [ Add Instaclustr as a preferred source on Google ](https://google.com/preferences/source?q=instaclustr.com) ![mail icon]()#### Get the latest articles for open sourceIn your inbox Sign up now ### Related content [ AI Search for OpenSearch®: Unlocking next-generation search Unlock next-gen semantic and multimodal AI search with OpenSearch. Our managed platform simplifies AI-powered search for smarter, ... ](https://www.instaclustr.com/blog/ai-search-for-opensearch-unlocking-next-generation-search/) [ Beyond keyword search: A guide to AI-powered search in OpenSearch Learn how OpenSearch AI search works across semantic, hybrid, neural sparse, multimodal, RAG, and OpenSearch vector search methods... ](https://www.instaclustr.com/blog/beyond-keyword-search-a-guide-to-ai-powered-search-in-opensearch/) [ Optimizing OpenSearch vector storage and search for faster AI applications ](https://www.instaclustr.com/courses/opensearch/optimizing-opensearch-vector-storage-and-search-for-faster-ai-applications/) ×Sign upto ourNewsletter -----------------------