mirror of
https://github.com/qdrant/landing_page.git
synced 2026-10-01 00:48:32 +02:00
update tutorials
This commit is contained in:
generall
@@ -0,0 +1,169 @@
|
||||
---
|
||||
title: Multitenancy
|
||||
weight: 12
|
||||
aliases:
|
||||
- ../tutorials/multiple-partitions
|
||||
---
|
||||
# Configure Multitenancy
|
||||
|
||||
**How many collections should you create?** In most cases, you should only use a single collection with payload-based partitioning. This approach is called multitenancy. It is efficient for most of users, but it requires additional configuration. This document will show you how to set it up.
|
||||
|
||||
**When should you create multiple collections?** When you have a limited number of users and you need isolation. This approach is flexible, but it may be more costly, since creating numerous collections may result in resource overhead. Also, you need to ensure that they do not affect each other in any way, including performance-wise.
|
||||
|
||||
## Partition by payload
|
||||
|
||||
When an instance is shared between multiple users, you may need to partition vectors by user. This is done so that each user can only access their own vectors and can't see the vectors of other users.
|
||||
|
||||
1. Add a `group_id` field to each vector in the collection.
|
||||
|
||||
```http
|
||||
PUT /collections/{collection_name}/points
|
||||
|
||||
{
|
||||
"points": [
|
||||
{
|
||||
"id": 1,
|
||||
"payload": {"group_id": "user_1"},
|
||||
"vector": [0.9, 0.1, 0.1]
|
||||
},
|
||||
{
|
||||
"id": 2,
|
||||
"payload": {"group_id": "user_1"},
|
||||
"vector": [0.1, 0.9, 0.1]
|
||||
},
|
||||
{
|
||||
"id": 3,
|
||||
"payload": {"group_id": "user_2"},
|
||||
"vector": [0.1, 0.1, 0.9]
|
||||
},
|
||||
]
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
client.upsert(
|
||||
collection_name="{collection_name}",
|
||||
points=[
|
||||
models.PointStruct(
|
||||
id=1,
|
||||
payload={"group_id": "user_1"},
|
||||
vector=[0.9, 0.1, 0.1],
|
||||
),
|
||||
models.PointStruct(
|
||||
id=2,
|
||||
payload={"group_id": "user_1"},
|
||||
vector=[0.1, 0.9, 0.1],
|
||||
),
|
||||
models.PointStruct(
|
||||
id=3,
|
||||
payload={"group_id": "user_2"},
|
||||
vector=[0.1, 0.1, 0.9],
|
||||
),
|
||||
]
|
||||
)
|
||||
```
|
||||
2. Use a filter along with `group_id` to filter vectors for each user.
|
||||
|
||||
```http
|
||||
POST /collections/{collection_name}/points/search
|
||||
|
||||
{
|
||||
"filter": {
|
||||
"must": [
|
||||
{
|
||||
"key": "group_id",
|
||||
"match": {
|
||||
"value": "user_1"
|
||||
}
|
||||
}
|
||||
]
|
||||
},
|
||||
"vector": [0.1, 0.1, 0.9],
|
||||
"limit": 10
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
from qdrant_client import QdrantClient, models
|
||||
|
||||
client = QdrantClient("localhost", port=6333)
|
||||
|
||||
client.search(
|
||||
collection_name="{collection_name}",
|
||||
query_filter=models.Filter(
|
||||
must=[
|
||||
models.FieldCondition(
|
||||
key="group_id",
|
||||
match=models.MatchValue(
|
||||
value="user_1",
|
||||
),
|
||||
)
|
||||
]
|
||||
),
|
||||
query_vector=[0.1, 0.1, 0.9],
|
||||
limit=10,
|
||||
)
|
||||
```
|
||||
## Calibrate performance
|
||||
|
||||
The speed of indexation may become a bottleneck in this case, as each user's vector will be indexed into the same collection. To avoid this bottleneck, consider _bypassing the construction of a global vector index_ for the entire collection and building it only for individual groups instead.
|
||||
|
||||
By adopting this strategy, Qdrant will index vectors for each user independently, significantly accelerating the process.
|
||||
|
||||
To implement this approach, you should:
|
||||
|
||||
1. Set `payload_m` in the HNSW configuration to a non-zero value, such as 16.
|
||||
2. Set `m` in hnsw config to 0. This will disable building global index for the whole collection.
|
||||
|
||||
```http
|
||||
PUT /collections/{collection_name}
|
||||
|
||||
{
|
||||
"vectors": {
|
||||
"size": 768,
|
||||
"distance": "Cosine"
|
||||
},
|
||||
"hnsw_config": {
|
||||
"payload_m": 16,
|
||||
"m": 0
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
from qdrant_client import QdrantClient, models
|
||||
|
||||
client = QdrantClient("localhost", port=6333)
|
||||
|
||||
client.recreate_collection(
|
||||
collection_name="{collection_name}",
|
||||
vectors_config=models.VectorParams(size=768, distance=models.Distance.COSINE),
|
||||
hnsw_config=models.HnswConfigDiff(
|
||||
payload_m=16,
|
||||
m=0,
|
||||
),
|
||||
)
|
||||
```
|
||||
|
||||
3. Create keyword payload index for `group_id` field.
|
||||
|
||||
```http
|
||||
PUT /collections/{collection_name}/index
|
||||
|
||||
{
|
||||
"field_name": "group_id",
|
||||
"field_schema": "keyword"
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
client.create_payload_index(
|
||||
collection_name="{collection_name}",
|
||||
field_name="group_id",
|
||||
field_schema=models.PayloadSchemaType.KEYWORD
|
||||
)
|
||||
```
|
||||
|
||||
## Limitations
|
||||
|
||||
One downside to this approach is that global requests (without the `group_id` filter) will be slower since they will necessitate scanning all groups to identify the nearest neighbors.
|
||||
@@ -0,0 +1,295 @@
|
||||
---
|
||||
title: Optimize Resources
|
||||
weight: 11
|
||||
aliases:
|
||||
- ../tutorials/optimize
|
||||
---
|
||||
|
||||
# Optimize Qdrant
|
||||
|
||||
Different use cases have different requirements for balancing between memory, speed, and precision.
|
||||
Qdrant is designed to be flexible and customizable so you can tune it to your needs.
|
||||
|
||||
|
||||
|
||||
Let's look deeper into each of those possible optimization scenarios.
|
||||
|
||||
## Prefer low memory footprint with high speed search
|
||||
|
||||
The main way to achieve high speed search with low memory footprint is to keep vectors on disk while at the same time minimizing the number of disk reads.
|
||||
|
||||
Vector quantization is one way to achieve this. Quantization converts vectors into a more compact representation, which can be stored in memory and used for search. With smaller vectors you can cache more in RAM and reduce the number of disk reads.
|
||||
|
||||
To configure in-memory quantization, with on-disk original vectors, you need to create a collection with the following configuration:
|
||||
|
||||
```http
|
||||
PUT /collections/{collection_name}
|
||||
|
||||
{
|
||||
"vectors": {
|
||||
"size": 768,
|
||||
"distance": "Cosine"
|
||||
},
|
||||
"optimizers_config": {
|
||||
"memmap_threshold": 20000
|
||||
},
|
||||
"quantization_config": {
|
||||
"scalar": {
|
||||
"type": "int8",
|
||||
"always_ram": true
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
from qdrant_client import QdrantClient
|
||||
from qdrant_client.http import models
|
||||
|
||||
client = QdrantClient("localhost", port=6333)
|
||||
|
||||
client.recreate_collection(
|
||||
collection_name="{collection_name}",
|
||||
vectors_config=models.VectorParams(size=768, distance=models.Distance.COSINE),
|
||||
optimizers_config=models.OptimizersConfigDiff(memmap_threshold=20000),
|
||||
quantization_config=models.ScalarQuantization(
|
||||
scalar=models.ScalarQuantizationConfig(
|
||||
type=models.ScalarType.INT8,
|
||||
always_ram=True,
|
||||
),
|
||||
),
|
||||
)
|
||||
```
|
||||
|
||||
`mmmap_threshold` will ensure that vectors will be stored on disk, while `always_ram` will ensure that quantized vectors will be stored in RAM.
|
||||
|
||||
Optionally, you can disable rescoring with search `params`, which will reduce the number of disk reads even further, but potentially slightly decrease the precision.
|
||||
|
||||
```http
|
||||
POST /collections/{collection_name}/points/search
|
||||
|
||||
{
|
||||
"params": {
|
||||
"quantization": {
|
||||
"rescore": false
|
||||
}
|
||||
},
|
||||
"vector": [0.2, 0.1, 0.9, 0.7],
|
||||
"limit": 10
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
from qdrant_client import QdrantClient
|
||||
from qdrant_client.http import models
|
||||
|
||||
client = QdrantClient("localhost", port=6333)
|
||||
|
||||
client.search(
|
||||
collection_name="{collection_name}",
|
||||
query_vector=[0.2, 0.1, 0.9, 0.7],
|
||||
search_params=models.SearchParams(
|
||||
quantization=models.QuantizationSearchParams(
|
||||
rescore=False
|
||||
)
|
||||
)
|
||||
)
|
||||
```
|
||||
|
||||
## Prefer high precision with low memory footprint
|
||||
|
||||
In case you need high precision, but don't have enough RAM to store vectors in memory, you can enable on-disk vectors and HNSW index.
|
||||
|
||||
```http
|
||||
PUT /collections/{collection_name}
|
||||
|
||||
{
|
||||
"vectors": {
|
||||
"size": 768,
|
||||
"distance": "Cosine"
|
||||
},
|
||||
"optimizers_config": {
|
||||
"memmap_threshold": 20000
|
||||
},
|
||||
"hnsw_config": {
|
||||
"on_disk": true
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
from qdrant_client import QdrantClient, models
|
||||
|
||||
client = QdrantClient("localhost", port=6333)
|
||||
|
||||
client.recreate_collection(
|
||||
collection_name="{collection_name}",
|
||||
vectors_config=models.VectorParams(size=768, distance=models.Distance.COSINE),
|
||||
optimizers_config=models.OptimizersConfigDiff(memmap_threshold=20000),
|
||||
hnsw_config=models.HnswConfigDiff(on_disk=True)
|
||||
)
|
||||
```
|
||||
|
||||
In this scenario you can increase the precision of the search by increasing the `ef` and `m` parameters of the HNSW index, even with limited RAM.
|
||||
|
||||
```json
|
||||
...
|
||||
"hnsw_config": {
|
||||
"m": 64,
|
||||
"ef_construct": 512,
|
||||
"on_disk": true
|
||||
}
|
||||
...
|
||||
```
|
||||
|
||||
The disk IOPS is a critical factor in this scenario, it will determine how fast you can perform search.
|
||||
You can use [fio](https://gist.github.com/superboum/aaa45d305700a7873a8ebbab1abddf2b) to measure disk IOPS.
|
||||
|
||||
## Prefer high precision with high speed search
|
||||
|
||||
For high speed and high precision search it is critical to keep as much data in RAM as possible.
|
||||
By default, Qdrant follows this approach, but you can tune it to your needs.
|
||||
|
||||
Is is possible to achieve high search speed and tunable accuracy by applying quantization with re-scoring.
|
||||
|
||||
```http
|
||||
PUT /collections/{collection_name}
|
||||
|
||||
{
|
||||
"vectors": {
|
||||
"size": 768,
|
||||
"distance": "Cosine"
|
||||
},
|
||||
"optimizers_config": {
|
||||
"memmap_threshold": 20000
|
||||
},
|
||||
"quantization_config": {
|
||||
"scalar": {
|
||||
"type": "int8",
|
||||
"always_ram": true
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
from qdrant_client import QdrantClient
|
||||
from qdrant_client.http import models
|
||||
|
||||
client = QdrantClient("localhost", port=6333)
|
||||
|
||||
client.recreate_collection(
|
||||
collection_name="{collection_name}",
|
||||
vectors_config=models.VectorParams(size=768, distance=models.Distance.COSINE),
|
||||
optimizers_config=models.OptimizersConfigDiff(memmap_threshold=20000),
|
||||
quantization_config=models.ScalarQuantization(
|
||||
scalar=models.ScalarQuantizationConfig(
|
||||
type=models.ScalarType.INT8,
|
||||
always_ram=True,
|
||||
),
|
||||
),
|
||||
)
|
||||
```
|
||||
|
||||
There are also some search-time parameters you can use to tune the search accuracy and speed:
|
||||
|
||||
```http
|
||||
POST /collections/{collection_name}/points/search
|
||||
|
||||
{
|
||||
"params": {
|
||||
"hnsw_ef": 128,
|
||||
"exact": false
|
||||
},
|
||||
"vector": [0.2, 0.1, 0.9, 0.7],
|
||||
"limit": 3
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
from qdrant_client import QdrantClient, models
|
||||
|
||||
client = QdrantClient("localhost", port=6333)
|
||||
|
||||
client.search(
|
||||
collection_name="{collection_name}",
|
||||
search_params=models.SearchParams(
|
||||
hnsw_ef=128,
|
||||
exact=False
|
||||
),
|
||||
query_vector=[0.2, 0.1, 0.9, 0.7],
|
||||
limit=3,
|
||||
)
|
||||
```
|
||||
|
||||
- `hnsw_ef` - controls the number of neighbors to visit during search. The higher the value, the more accurate and slower the search will be. Recommended range is 32-512.
|
||||
- `exact` - if set to `true`, will perform exact search, which will be slower, but more accurate. You can use it to compare results of the search with different `hnsw_ef` values versus the ground truth.
|
||||
|
||||
## Latency vs Throughput
|
||||
|
||||
- There are two main approaches to measure the speed of search:
|
||||
- latency of the request - the time from the moment request is submitted to the moment a response is received
|
||||
- throughput - the number of requests per second the system can handle
|
||||
|
||||
Those approaches are not mutually exclusive, but in some cases it might be preferable to optimize for one or another.
|
||||
|
||||
To prefer minimizing latency, you can set up Qdrant to use as many cores as possible for a single request\.
|
||||
You can do this by setting the number of segments in the collection to be equal to the number of cores in the system. In this case, each segment will be processed in parallel, and the final result will be obtained faster.
|
||||
|
||||
```http
|
||||
|
||||
PUT /collections/{collection_name}
|
||||
|
||||
{
|
||||
"vectors": {
|
||||
"size": 768,
|
||||
"distance": "Cosine"
|
||||
},
|
||||
"optimizers_config": {
|
||||
"default_segment_number": 16
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
from qdrant_client import QdrantClient, models
|
||||
|
||||
client = QdrantClient("localhost", port=6333)
|
||||
|
||||
client.recreate_collection(
|
||||
collection_name="{collection_name}",
|
||||
vectors_config=models.VectorParams(size=768, distance=models.Distance.COSINE),
|
||||
optimizers_config=models.OptimizersConfigDiff(default_segment_number=16),
|
||||
)
|
||||
```
|
||||
|
||||
To prefer throughput, you can set up Qdrant to use as many cores as possible for processing multiple requests in parallel.
|
||||
To do that, you can configure qdrant to use minimal number of segments, which is usually 2.
|
||||
Large segments benefit from the size of the index and overall smaller number of vector comparisons required to find the nearest neighbors. But at the same time require more time to build index.
|
||||
|
||||
```http
|
||||
|
||||
PUT /collections/{collection_name}
|
||||
|
||||
{
|
||||
"vectors": {
|
||||
"size": 768,
|
||||
"distance": "Cosine"
|
||||
},
|
||||
"optimizers_config": {
|
||||
"default_segment_number": 2
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
```python
|
||||
from qdrant_client import QdrantClient, models
|
||||
|
||||
client = QdrantClient("localhost", port=6333)
|
||||
|
||||
client.recreate_collection(
|
||||
collection_name="{collection_name}",
|
||||
vectors_config=models.VectorParams(size=768, distance=models.Distance.COSINE),
|
||||
optimizers_config=models.OptimizersConfigDiff(default_segment_number=2),
|
||||
)
|
||||
```
|
||||
Reference in New Issue
Block a user