Vector Databases 101 - An introduction to the world of Vector Databases

1 | © Copyright 2023 Zilliz
1
Stephen Batifol | Zilliz
Unstructured Data Meetup #1, @GetYourGuide
Vector Databases 101

2
Stephen Batifol
Developer Advocate, Zilliz
stephen.batifol@zilliz.com
https://www.linkedin.com/in/stephen-batifol/
https://twitter.com/stephenbtl
Speaker

3
Open Source
Deploy fully managed or “Bring Your
Own Cloud” (BYOC)
Commercial Offerings
Zilliz Cloud
Optimized Milvus with essential data and
security tools for a high-performing vector
search platform
VECTOR SEARCH
ENGINE
VECTORDB
BENCHMARK TOOL
VECTOR DATABASE
SEMANTIC CACHE
FOR LLM QUERIES
GPT-Cache
Product Portfolio
GUI for Milvus

4
Couple of Customers

5
Cloud
Service
Provider
Data Platform
GenAI Tooling
Chip
Manufacturer
Partner with Industry Leaders

6
01 What is Similarity Search?
CONTENTS
03
04 Vector Database Architecture
02 Where do You Get Vectors From?
How are Vector Databases Different?

7
01 Vector Databases 101 Review

8
- Unstructured Data is 80% of data
- Vector Databases are really good with
unstructured data
- Examples of Unstructured Data include text,
images, videos, audio, etc
Why Vector Databases?

9
Traditional databases were built on exact search

10
…which misses context, semantic meaning, and user intent
VS.
Apple
VS.
Rising dough
VS.
Change car tire
Rising Dough
Proofing Bread
✔
❌

11
But wait! There’s more!

12
Vector
Databases
Where do Vectors Come From?

13
Embeddings Models

14
02 Vector Database Use Cases

15
Retrieval Augmented
Generation (RAG)
Expand LLMs' knowledge by
incorporating external data sources
into LLMs and your AI applications.
Match user behavior or content
features with other similar ones to
make effective recommendations.
Recommender System
Search for semantically similar
texts across vast amounts of
natural language documents.
Text/ Semantic Search
Image Similarity Search
Identify and search for visually
similar images or objects from a
vast collection of image libraries.
Video Similarity Search
Search for similar videos, scenes,
or objects from extensive
collections of video libraries.
Audio Similarity Search
Find similar audios in large datasets
for tasks like genre classification or
speech recognition
Molecular Similarity Search
Search for similar substructures,
superstructures, and other
structures for a specific molecule.
Anomaly Detection
Detect data points, events, and
observations that deviate
significantly from the usual pattern
Multimodal Similarity Search
Search over multiple types of data
simultaneously, e.g. text and
images
Common AI Use Cases

16
Reverse Image Search - Paintings

17
Text Search on Wikipedia

18
03 How do Vector
Databases Work?

19
Example Entry

20
04 What is Similarity Search?

21
Image from Nvidia
Vector Search Overview

22
Vector Similarity Measures: L2 (Euclidean)
Queen = [0.3, 0.9]
King = [0.5, 0.7]
d(Queen, King) = √(0.3-0.5)2
+ (0.9-0.7)2
= √(0.2)2
+ (0.2)2
= √0.04 + 0.04
= √0.08 ≅ 0.28

23
Vector Similarity Measures: Inner Product (IP)
Queen = [0.3, 0.9]
King = [0.5, 0.7]
Queen · King = (0.3*0.5) + (0.9*0.7)
= 0.15 + 0.63 = 0.78

24
Queen = [0.3, 0.9]
King = [0.5, 0.7]
Vector Similarity Measures: Cosine
𝚹
cos(Queen, King) = (0.3*0.5)+(0.9*0.7)
√0.32
+0.92
* √0.52
+0.72
= 0.15+0.63 _
√0.9 * √0.74
= 0.78 _
√0.666
≅ 0.03

25
Important Notes
- Cosine, IP, and L2 are all the SAME rank order.
- They differ in use case
- L2 for when you need magnitude
- Cosine for orientation
- IP for magnitude and orientation
- OR
- Cosine = IP for normalized vectors

26
Inverted File Index
Source:
https://towardsdatascience.com/similarity-search-with-ivfpq-9c6348fd4db3

27
HNSW
Source:
https://arxiv.org/ftp/arxiv/papers/1603/1603.09320.pdf

28
GPU Index

29
Scalar Quantization

30
Product Quantization
Source:
https://towardsdatascience.com/product-quantization-for-similarity-search-2f1f67c5fddd

31
Picking an Index
● 100% Recall – Use FLAT search if you need 100% accuracy
● 10MB < index_size < 2GB – Standard IVF
● 2GB < index_size < 20GB – Consider PQ and HNSW
● 20GB < index_size < 200GB – Composite Index, IVF_PQ or
HNSW_SQ
● Disk-based indexes

32
05
How are Vector
Databases Different
from Other Databases?

33
Why Not Use a SQL/NoSQL Database?
• Inefficiency in High-dimensional spaces
• Suboptimal Indexing
• Inadequate query support
• Lack of scalability
• Limited analytics capabilities
• Data conversion issues
TL;DR: Vector operations are too computationally intensive for
traditional database infrastructures

34
Why Not Use a Vector Search Library?
• Have to manually implement filtering
• Not optimized to take advantage of the latest hardware
• Unable to handle large scale data
• Lack of lifecycle management
• Inefficient indexing capabilities
• No built in safety mechanisms
TL;DR: Vector search libraries lack the infrastructure to help you scale,
deploy, and manage your apps in production.

35
What is Milvus/Zilliz ideal for?
• Advanced filtering
• Hybrid search
• Durability and backups
• Replications/High Availability
• Sharding
• Aggregations
• Lifecycle management
• Multi-tenancy
• High query load
• High insertion/deletion
• Full precision/recall
• Accelerator support (GPU,
FPGA)
• Billion-scale storage
Purpose-built to store, index and query vector embeddings from unstructured data at scale.

36
06 Vector Database Architecture

37
Meta Storage
Root Query Data Index
Coordinator Service
Proxy
Proxy
etcd
Log Broker
SDK
Load Balancer
DDL/DCL
DML
NOTIFICATION
CONTROL SIGNAL
Object Storage
Minio / S3 / AzureBlob
Log Snapshot Delta File Index File
Worker Node QUERY DATA DATA
Message Storage
Access Layer
Query Node Data Node Index Node
High-level overview of Milvus’ Architecture

38
Questions?
Give Milvus a Star! Chat with me on Discord! Meetup, May 7
@Delivery Hero

40
Milvus Architecture: Differentiation
1. Cloud Native, Distributed System Architecture
2. True Separation of Concerns
3. Scalable Index Creation Strategy with 512 MB Segments

41
Vector Databases are purpose-built to handle
indexing, storing, and querying vector data.
Milvus & Zilliz are specifically designed for high
performance and billion+ scale use cases.
Takeaway:

42
Examples
Chunk Size=32, Overlap=10

43
Examples

44
Examples

45
Examples
SentenceSplitter

46
We’ve built technologies for various types of
use cases
Compute Types
Support different types of
compute powers, such as
AVX512, Neon for SIMD
execution, quantization &
cache-aware optimization,
and GPU
Leverage specific strengths
of each hardware type
efficiently, ensuring
high-speed processing and
cost-effective scalability for
diverse application needs
Search Types
Provide diverse search
types such as top-K ANN,
Range ANN, hybrid ANN
and metadata filtering
Enable unparalleled query
flexibility and accuracy,
allowing developers to
tailor their data retrieval
needs
Multi-tenancy
Enable multi-tenancy
through collection and
partition management
Allow for efficient resource
utilization and customizable
data segregation, ensuring
secure and isolated data
handling for each tenant
Index Types
Offer a diverse range of 11
index types, including
popular ones like HNSW,
IVF, PQ, and GPU index
Empower developers with
tailored search
optimizations, catering to
specific performance and
accuracy needs

47
Tests shows consistent query performance when
scaled from 65 million to 1 billion vectors

48
ANN Benchmark has recognized Milvus as the
performance leader among vector database players

49
The Zilliz Cloud Platform
Data Consumers
Data Applications
Data Sources
Unstructured Data
Zilliz Cloud
Managed Milvus Vector Database
Integrations
High performance
vector search
Store, Index, and
query
Low latency with
high recall
Supports real-time
responses
Vector search
index diversity
Efficient search &
scalability
Multiple similarity
metrics
Flexibility in querying
& adaptability
Built-in data
filtering
Efficient data
retrieval
On transit data
encryption
Enterprise security &
support
Images
Video
Documents
Audio
User Generated
Content
Recommendation
systems
Search (text,
image, audio)
Chatbots
Anomaly
detection
DNA sequence
classification

50
Data Consumers
Data Sources Zilliz Cloud
Integrations
Recommendation
systems
Search (text,
image, audio)3
Chatbots
Anomaly
detection
DNA sequence
classification
Images
User Generated
Video
Documents
Audio
Hosted Milvus Vector Database Data Applications
Unstructured Data
Zilliz Cloud Platform
✔ High performance vector search ⏤ Store, Index, and query
✔ Low latency with high recall ⏤ Supports real-time responses
✔ Vector search index diversity ⏤ Efficient search & scalability
✔ Multiple similarity metrics ⏤ Flexibility in querying & adaptability
✔ Built-in data filtering ⏤ Efficient data retrieval
✔ On transit data encryption ⏤ Enterprise security & support

51

52
Meta Storage
Root Query Data Index
Coordinator Service
Proxy
Proxy
etcd
Log Broker
SDK
Load Balancer
DDL/DCL
DML
NOTIFICATION
CONTROL SIGNAL
Object Storage
Minio / S3 / AzureBlob
Log Snapshot Delta File Index File
Worker Node QUERY DATA DATA
Message Storage
VECTOR
DATABASE
Access Layer
Query Node Data Node Index Node
Milvus Architecture

Vector Databases 101 - An introduction to the world of Vector Databases

Recommended

Recommended

More Related Content

Similar to Vector Databases 101 - An introduction to the world of Vector Databases

Similar to Vector Databases 101 - An introduction to the world of Vector Databases (20)

More from Zilliz

More from Zilliz (12)

Recently uploaded

Recently uploaded (20)

Vector Databases 101 - An introduction to the world of Vector Databases