intro
Find out everything you need to know about the concept of vector databases — what they are, how they work, and when should you use them.
Databases. Many of our readers will know that there are multiple different types of databases, including non-relational, relational, vector databases.
Ask any software engineer about databases and you will not only hear about a couple of database management systems (e.g., MySQL Server, MongoDB, TimescaleDB, and the like), but also about database types.
Some will start walking you through how NoSQL databases don’t work with relational data sets, some may walk you through ACID or BASE functionality and the difference between them. Some of them are built by the world’s best engineers and work flawlessly (once they’re employed properly, that is), some are still ridden with weird errors and glitches.
Regardless, databases have a purpose—they store data. Some of them store data in a JSON format, some store data in a structured manner, and some use vector embeddings. This is the case of a vector database!
Explaining Vector Databases
Vector databases are just what they sound like—they’re a type of database that stores data as vectors. For some, vectors may associate with links to data from indexes (in the sense that the database doesn’t store the data itself, but stores “links” to the data itself), some may associate them with geospatial operations as a vector is an object with a magnitude and a direction.
Here’s an example for easier understanding. Assume this scenario:
The point here is very simple—the search works using vector databases. Using vector databases saves your server tremendous amounts of processing time because your data is “matched” to the query only once, and since embeddings of the data is stored in a vector database, searches are blazing fast. Got it?
What Kind of Vector Databases Exist?
Right, you know what a vector database is. Cool. The next question you’ve probably already asked yourself is the following—what vector databases exist and which one do I choose for my specific use case? Does DbVisualizer support any of them? Which of those databases are supported by DbVisualizer?
Let’s start by answering the first question. There are multiple vector databases you can choose from, and these databases include, but are not limited to:
| Vector Database | About | Supported in DbVisualizer? |
|---|---|---|
| https://www.elastic.co/elasticsearch/vector-database | One of the most popular choices for those looking for a vector database. An efficient way to store vector embeddings at scale—known for combining text-based search and vector search for an easy and stress-free searching process. | Yes |
| https://cassandra.apache.org/_/index.html | Vector search is a feature available from Apache Cassandra 5.0 and is used to find content within larger data sets. | Yes |
| https://clickhouse.com | ClickHouse is a real-time OLAP database. ClickHouse has many features unique to itself and one of those includes a vector search capability, enabling ClickHouse to be used as a vector database. Learn more about ClickHouse’s support for vector datasets https://clickhouse.com/blog/vector-search-clickhouse-p2. | Yes |
| https://developers.cloudflare.com/vectorize/ | At present, CloudFlare’s Vectorize database is in a public beta stage, but all developers using the Workers Paid plan can give it a try. CloudFlare is working on its vector database offering to help you build applications with powerful similarity, recommendation, anomaly detection, or other capabilities based on data. | No, this is a very specific offering for CloudFlare customers. |
There are many other vector databases you can choose from and we’ve only mentioned a few. Anyway, most of these databases (excluding those with specific use cases, such as Vectorize) serve the same purpose, but in different ways. Many of the most widely used vector databases are available in DbVisualizer and you can search through the list of the available databases we support.
When to Use a Vector Database?
Another question you may have may be related to whether it’s actually useful for your team to use a vector database. The answer is simple and quite straightforward, too—use a vector database if you need to find data that is similar to other data. Some examples? When looking for images similar to other images based on a specific factor, when building a search engine for movies and grouping movies based on actors, IMDB ratings, or something else, etc.
As such, vector databases are not for everyone, and they do have their downsides too:
We’re coming to a conclusion that vector databases are useful if you need to find data similar to other data and you want to store data in a vector format to facilitate blazing-fast searches by saving processing time.
If you’re after something else, do have a look through our supported databases and choose the one that’s right for you.
Summary
Vector databases store data as vectors to minimize the processing time of your server by matching data to the query only once. They do have their use cases and they’re widely used to facilitate searches through data similar to other data, but not for many other things.
Do you use such databases? What kind of vector databases do you use? Let us know on Twitter, and until next time.
Frequently Asked Questions
What are vector databases?
Vector databases are a type of database that stores data as vector objects—they’re used to find data similar to other data.
What’s the Most Popular Vector Database?
The most popular vector database is ElasticSearch.
When Should I Use a Vector Database?
Use a vector database when your queries are simple, you don’t have the need for complex data types, and you need to store data as vector objects.
Does DbVisualizer support vector databases?
Yes, DbVisualizer supports many vector databases including ElasticSearch, ClickHouse, Apache Cassandra, and many more. For the full list of databases supported by DbVisualizer, visit our database support page.
