Register now open for the virtual Milvus Community Conf2020!Join us on Oct.17th, 2020.

Product FAQ

Is Milvus free of charge?

Milvus is an open-source project, and hence is free-of-charge.

Please adhere to Apache License 2.0, when using Milvus for reproduction or distribution purposes.

Does Milvus support non-x86 architecture?

No, it does not.

Does Milvus support CRUD operations on vectors?

Yes. To update a vector, you can delete it and then insert a new one.

Can Milvus handle datasets of up to a 100-billion scale?

By deploying Mishards, a cluster sharding middleware for Milvus, you can process datasets of up to a 100-billion scale.

Where does Milvus store imported data?

Vectors imported into Milvus are stored locally at milvus/db/tables/.

Metadata can be stored in either MySQL or SQLite. See Manage Metadata with MySQL for more information.

Why can't I find vectors on SQLite or MySQL?

Milvus stores vectors and indexes directly in the disk as files, not in SQLite or MySQL. It uses SQLite or MySQL to store metadata of the vectors instead.

Can I use SQL Server or PostgreSQL to store metadata in Milvus?

No, we only support storing metadata using SQLite or MySQL.

Does Milvus' Python SDK have a connection pool?

Python SDKs corresponding to Milvus v0.9.0 or later have a connection pool. There is no upper limit on the default number of connections in a connection pool.

Does Milvus support inserting while searching?

Yes.

Is there a graphical tool for managing Milvus?

As of Milvus v0.7.0, we have provided Milvus Enterprise Manager as a graphical tool for managing Milvus.

Can I export data from Milvus?

We do not have a dedicated tool as yet. You can call get_entity_by_id to get the intended vectors by ID.

Why do the retrieved vectors suffer precision loss after the get_entity_by_id method call?

Milvus stores and processes each dimension of a vector in single-precision floating-point format (accurate to seven decimal places). Therefore, if the original format of each dimension is double-precision floating-point (accurate to sixteen decimal places), you will see a precision loss.

Should I specify entity IDs when importing vectors or have Milvus generate them for me?

Either way is fine. But please note that entity IDs in the same collection must be either user-generated or Milvus-generated. Can't be both.

Is there a length limit on the self-defined entity IDs?

Entity IDs must be non-negative 64-bit integers.

Is there a volume limit on the vectors inserted each time?

Vectors inserted each time must not exceed 256 MB.

Why is the top1 result of a vector search not the search vector itself, if the metric type is inner product?

This occurs if you have not normalized the vectors when using inner product as the distance metric.

Does the size of a collection affect vector searches in one of its partitions, especially when it holds up to 100 million vectors?

No. If you have specified partitions when conducting a vector search, Milvus searches the specified partitions only.

Does Milvus load the whole collection to the memory if I search only certain partitions in that collection?

No, Milvus only loads the partitions to search.

Are queries in segments processed in parallel?

Yes. But the parallelism processing mechanism varies with Milvus versions.

Suppose a collection has multiple segments, then when a query request comes in:

  • CPU-only Milvus processes the segment reading tasks and the segment searching tasks in pipeline.
  • On top of the abovementioned pipeline mechanism, GPU-enabled Milvus distributes the segments among the available GPUs.

See How Does Milvus Schedule Query Tasks for more information.

How to choose an index in Milvus?

It depends on your scenario. See Select Vector Search Tool and How to Choose an Index in Milvus for more information.

Can Milvus create different types of index for different partitions?

No, you cannot. Indexes are created at the collection level, not at the partition level.

Can Milvus create different types of index in the same collection?

No, you cannot. Although a collection can hold various types of data, the same collection can use only one index type.

Does Milvus create new indexes after vectors are inserted?

Yes. When the inserted vectors grow to a specified volume, Milvus creates a new segment and starts to create an index file for it at the same time. The building of the new index file does not affect the existing index files.

Does IVFSQ8 differ from IVFSQ8H in terms of recall rate?

No, they have the same recall rate for the same dataset.

What is the difference between FLAT index and IVF_FLAT index?

IVFFLAT index divides a vector space into nlist clusters. If you keep the default value of nlist as 16384, Milvus compares the distances between the target vector and the centers of all 16384 clusters to get nprobe nearest clusters. Then Milvus compares the distances between the target vector and the vectors in the selected clusters to get the nearest vectors. Unlike IVFFLAT, FLAT directly compares the distances between the target vector and each and every vector.

Therefore, when the total number of vectors approximately equals nlist, IVFFLAT and FLAT has little difference in the way of calculation required and search performance. But as the number of vectors grows to two times, three times, or n times of nlist, IVFFLAT index begins to show increasingly greater advantages.

See Select Vector Search Tool for more information.

Why do I see a surge in memory usage when conducting a vector search immediately after an index is created?

This is because:

  • Milvus loads the newly created index file to the memory for the vector search.
  • The original vector files used to create the index are not yet released from the memory, because the size of original vector files and the index file has not exceeded the upper limit specified by cache.cache_size.

Can I update index_file_size and metric_type after creating a collection?

No, you cannot.

What is the interval at which Milvus flushes data to the disk?

Milvus automatically flushes data to disk at intervals of one second.

If I have set preload_collection, does Milvus service start only after all collections are loaded to the memory?

Yes. If you have set preload_collection in server_config.yaml, Milvus' service is not available until it loads all specified collections.

In what way does Milvus flush data?

Milvus loads inserted data to the memory and automatically flushes data from memory to the disk at fixed intervals. You can call flush to manually trigger this operation.

We recommend that you configure write nodes to using GPU-enabled Milvus and read nodes to using CPU-only Milvus. If you can have only one write node, you can configure this node to using GPU-enabled Milvus for creating indexes and configure read nodes to using CPU-only Milvus.

Does Mishards support RESTful APIs?

No, it does not.

What is normalization? Why is normalization needed?

To normalize a vector is to uniformly set the length of all vectors to 1. If you have normalized the vectors in the same space, then the top k nearest vectors returned using Euclidean distance (L2) are identical to the the nearest vectors returned using inner product (IP).

See Wikipedia for more information.

Why do I get different results using Euclidean distance (L2) and inner product (IP) as the distance metric?

Check if the vectors are normalized. If not, you need to normalize the vectors first. Theoretically speaking, similarities worked out by L2 are different from similarities worked out by IP, if the vectors are not normalized.

Is there a limit on the total number of collections and partitions?

Yes. The total number of collections and partitions must not exceed 4,096.

Why do I get fewer than k vectors when searching for topk vectors?

Among the indexes that Milvus supports, IVFFLAT and IVFSQ8 implement the k-means clustering method. A data space is divided into nlist clusters and the inserted vectors are distributed to these clusters. Milvus then selects the nprobe nearest clusters and compares the distances between the target vector and all vectors in the selected clusters to return the final results.

If nlist and k are large and nprobe is small, the amount of vectors in the nprobe clusters may be less than k. Therefore, when you search for the topk nearest vectors, the number of returned vectors is less than k.

To avoid this, try setting nprobe larger and nlist and k smaller.

See Index Types for more information.

Still have questions?

You can:

  • Check out Milvus on GitHub. You're welcome to raise questions, share ideas, and help others.
  • Join our Slack community to find more help and have fun!
Edit
© 2019 - 2020 Milvus. All rights reserved.