OpenSearch Vector Engine is now disk-optimized for low price, correct vector search

OpenSearch Vector Engine can now run vector search at a 3rd of the price on OpenSearch 2.17+ domains. Now you can configure k-NN (vector) indexes to run on disk mode, optimizing it for memory-constrained environments, and allow low-cost, correct vector search that responds in low lots of of milliseconds. Disk mode offers a cost-effective different to reminiscence mode whenever you don’t want close to single-digit latency.

On this put up, you’ll study the advantages of this new characteristic, the underlying mechanics, buyer success tales, and getting began.

Overview of vector search and the OpenSearch Vector Engine

Vector search is a way that improves search high quality by enabling similarity matching on content material that has been encoded by machine studying (ML) fashions into vectors (numerical encodings). It permits use instances like semantic search, permitting you to think about context and intent together with key phrases to ship extra related searches.

OpenSearch Vector Engine permits real-time vector searches past billions of vectors by creating indexes on vectorized content material. You possibly can then run searches for the highest Ok paperwork in an index which might be most much like a given question vector, which might be a query, key phrase, or content material (equivalent to a picture, audio clip, or textual content) that has been encoded by the identical ML mannequin.

Tuning the OpenSearch Vector Engine

Search functions have various necessities by way of velocity, high quality, and price. For example, ecommerce catalogs require the bottom doable response occasions and high-quality search to ship a constructive buying expertise. Nevertheless, optimizing for search high quality and efficiency features usually incurs price within the type of extra reminiscence and compute.

The best steadiness of velocity, high quality, and price will depend on your use instances and buyer expectations. OpenSearch Vector Engine offers complete tuning choices so you can also make good trade-offs to attain optimum outcomes tailor-made to your distinctive necessities.

You need to use the next tuning controls:

• Algorithms and parameters – This consists of the next:
  • Hierarchical Navigable Small World (HNSW) algorithm and parameters like ef_search, ef_construct, and m
  • Inverted File Index (IVF) algorithm and parameters like nlist and nprobes
  • Actual k-nearest neighbors (k-NN), often known as brute-force k-NN (BFKNN) algorithm
• Engines – Fb AI Similarity Search (FAISS), Lucene, and Non-metric House Library (NMSLIB).
• Compression methods – Scalar (equivalent to byte and half precision), binary, and product quantization
• Similarity (distance) metrics – Interior product, cosine, L1, L2, and hamming
• Vector embedding varieties – Dense and sparse with variable dimensionality
• Rating and scoring strategies – Vector, hybrid (mixture of vector and Finest Match 25 (BM25) scores), and multi-stage rating (equivalent to cross-encoders and personalizers)

You possibly can alter a mixture of tuning controls to attain a various steadiness of velocity, high quality, and price that’s optimized to your wants. The next diagram offers a tough efficiency profiling for pattern configurations.

Tuning for disk-optimization

With OpenSearch 2.17+, you possibly can configure your k-NN indexes to run on disk mode for high-quality, low-cost vector search by buying and selling in-memory efficiency for larger latency. In case your use case is happy with ninetieth percentile (P90) latency within the vary of 100–200 milliseconds, disk mode is a superb possibility so that you can obtain price financial savings whereas sustaining excessive search high quality. The next diagram illustrates disk mode’s efficiency profile amongst different engine configurations.

Disk mode was designed to expire of the field, decreasing your reminiscence necessities by 97% in comparison with reminiscence mode whereas offering excessive search high quality. Nevertheless, you possibly can tune compression and sampling charges to regulate for velocity, high quality, and price.

The next desk presents efficiency benchmarks for disk mode’s default settings. OpenSearch Benchmark (OSB) was used to run the primary three exams, and VectorDBBench (VDBB) was used for the final two. Efficiency tuning greatest practices had been utilized to attain optimum outcomes. The low scale exams (Tasb-1M and Marco-1M) had been run on a single r7gd.massive information node with one reproduction. The opposite exams had been run on two r7gd.2xlarge information nodes with one reproduction. The % price discount metric is calculated by evaluating an equal, right-sized in-memory deployment with the default settings.

These exams are designed to display that disk mode can ship excessive search high quality with 32 occasions compression throughout a wide range of datasets and fashions whereas sustaining our goal latency (beneath P90 200 milliseconds). These benchmarks aren’t designed for evaluating ML fashions. A mannequin’s affect on search high quality varies with a number of components, together with the dataset.

Disk mode’s optimizations beneath the hood

Once you configure a k-NN index to run on disk mode, OpenSearch mechanically applies a quantization method, compressing vectors as they’re loaded to construct a compressed index. By default, disk mode converts every full-precision vector—a sequence of lots of to hundreds of dimensions, every saved as 32-bit numbers—into binary vectors, which symbolize every dimension as a single-bit. This conversion leads to a 32 occasions compression charge, enabling the engine to construct an index that’s 97% smaller than one composed of full-precision vectors. A right-sized cluster will maintain this compressed index in reminiscence.

Compression lowers price by decreasing the reminiscence required by the vector engine, but it surely sacrifices accuracy in return. Disk mode recovers accuracy, and subsequently search high quality, utilizing a two-step search course of. The primary section of the question execution begins by effectively traversing the compressed index in reminiscence for candidate matches. The second section makes use of these candidates to oversample corresponding full-precision vectors. These full-precision vectors are saved on disk in a format designed to scale back I/O and optimize disk retrieval velocity and effectivity. The pattern of full-precision vectors is then used to enhance and re-score matches from section one (utilizing actual k-NN), thereby recovering the search high quality loss attributed to compression. Disk mode’s larger latency relative to reminiscence mode is attributed to this re-scoring course of, which requires disk entry and extra computation.

Early buyer successes

Clients are already operating the vector engine in disk mode. On this part, we share testimonials from early adopters.

Asana is enhancing search high quality for patrons on their work administration platform by phasing in semantic search capabilities by OpenSearch’s vector engine. They initially optimized the deployment through the use of product quantization to compress indexes by 16 occasions. By switching over to the disk-optimized configurations, they had been in a position to doubtlessly cut back price by one other 33% whereas sustaining their search high quality and latency targets. These economics make it viable for Asana to scale to billions of vectors and democratize semantic search all through their platform.

DevRev bridges the basic hole in software program firms by immediately connecting customer-facing groups with builders. As an AI-centered platform, it creates direct pathways from buyer suggestions to product growth, serving to over 1,000 firms speed up progress with correct search, quick analytics, and customizable workflows. Constructed on massive language fashions (LLMs) and Retrieval Augmented Era (RAG) flows operating on OpenSearch’s vector engine, DevRev permits clever conversational experiences.

“With OpenSearch’s disk-optimized vector engine, we achieved our search high quality and latency targets with 16x compression. OpenSearch affords scalable economics for our multi-billion vector search journey.”

– Anshu Avinash, Head of AI and Search at DevRev.

Get began with disk mode on the OpenSearch Vector Engine

First, it’s essential to decide the assets required to host your index. Begin by estimating the reminiscence required to help your disk-optimized k-NN index (with the default 32 occasions compression charge) utilizing the next system:

Required reminiscence (bytes) = 1.1 x ((vector dimension rely)/8 + 8 x m) x (vector rely)

For example, when you use the defaults for Amazon Titan Textual content V2, your vector dimension rely is 1024. Disk mode makes use of the HNSW algorithm to construct indexes, so “m” is without doubt one of the algorithm parameters, and it defaults to 16. In the event you construct an index for a 1-billion vector corpus encoded by Amazon Titan Textual content, your reminiscence necessities are 282 GB.

When you’ve got a throughput-heavy workload, it’s essential to be sure that your area has ample IOPs and CPUs as effectively. In the event you observe deployment greatest practices, you should utilize occasion retailer and storage efficiency optimized occasion varieties, which can usually give you ample IOPs. It’s best to at all times carry out load testing for high-throughput workloads, and alter the unique estimates to accommodate for larger IOPs and CPU necessities.

Now you possibly can deploy an OpenSearch 2.17+ area that has been right-sized to your wants. Create your k-NN index with the mode parameter set to on_disk, after which ingest your information. If you have already got a k-NN index operating on the default in_memory mode, you possibly can convert it by switching the mode to on_disk adopted by a reindex job. After the index is rebuilt, you possibly can downsize your area accordingly.

Conclusion

On this put up, we mentioned how one can profit from operating the OpenSearch Vector Engine on disk mode, shared buyer success tales, and supplied you recommendations on getting began. You’re now set to run the OpenSearch Vector Engine at as little as a 3rd of the price.

To study extra, discuss with the documentation.

In regards to the Authors

Dylan Tong is a Senior Product Supervisor at Amazon Net Companies. He leads the product initiatives for AI and machine studying (ML) on OpenSearch together with OpenSearch’s vector database capabilities. Dylan has a long time of expertise working immediately with prospects and creating merchandise and options within the database, analytics and AI/ML area. Dylan holds a BSc and MEng diploma in Pc Science from Cornell College.

Vamshi Vijay Nakkirtha is a software program engineering supervisor engaged on the OpenSearch Venture and Amazon OpenSearch Service. His main pursuits embrace distributed methods.