Hybrid Search is Slow

Support
,
1

Description

My db has two collections, when searching on the smallest collection (164k objects).
I obtain the following percentile delays:

p50: 1192.8 ms
p95: 2564.4 ms
p99: 3493.5 ms

(metrics based on 124 samples)

My code is simply:

start_time = time.perf_counter()
await artigo.query.hybrid(
  query=query,
  alpha=alpha,
  vector=vector,
  return_properties=return_properties,
  query_properties=query_properties, 
  return_metadata=wvc.query.MetadataQuery(score=True),
  limit=75,)
duration_ms = (time.perf_counter() - start_time) * 1000
metrics.record(duration_ms)

I would like to know what is taking so long is it the search or sending the data to my server?

My code runs Frankfurt and my Weaviate Cluster is in GCP Europe.

Server Setup Information

  • Weaviate Database Version: 1.34.5
  • Deployment Type: Shared
2

Hi @jpiabrantes,

Good day, and welcome to the Weaviate Community! :waving_hand:

If you’re experiencing latency when running queries, there can be several factors involved, such as network conditions, cluster resources, or workload patterns. To better understand what’s happening and investigate your specific cluster setup, we’d be happy to take a closer look.

If you’re using Weaviate Cloud, please feel free to create a support ticket with our Support Team so we can investigate this further. You can reach us at support@weaviate.io.

Thank you, we’re here to help!

1 Like