github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #1400] How to serve multiple simultaneous request in Ollama? #26503

New Issue
Closed
opened 2026-04-22 02:48:25 -05:00 by GiteaMirror · 10 comments
GiteaMirror commented 2026-04-22 02:48:25 -05:00
Owner
Copy Link

Originally created by @austin-starks on GitHub (Dec 6, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1400

Hello!

I want to deploy Ollama in the cloud server. The cloud server I'm renting is big enough to handle multiple requests at the same time with the models I'm using. However, Ollama queues the request. What specific changes do I need to make for this to be possible? And, is there any way for this to be an additional configuration option added to the Ollama repo?

Originally created by @austin-starks on GitHub (Dec 6, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1400 Hello! I want to deploy Ollama in the cloud server. The cloud server I'm renting is big enough to handle multiple requests at the same time with the models I'm using. However, Ollama queues the request. What specific changes do I need to make for this to be possible? And, is there any way for this to be an additional configuration option added to the Ollama repo?
GiteaMirror commented 2026-04-22 02:48:27 -05:00
Author
Owner
Copy Link

@technovangelist commented on GitHub (Dec 6, 2023):

For now Ollama is designed to provide a great experience on your local machine for a single user. It is designed to queue the request and then it will get to the next request after the current one is complete. We intend to look at making a better server experience in the future.

<!-- gh-comment-id:1843612664 --> @technovangelist commented on GitHub (Dec 6, 2023): For now Ollama is designed to provide a great experience on your local machine for a single user. It is designed to queue the request and then it will get to the next request after the current one is complete. We intend to look at making a better server experience in the future.
GiteaMirror commented 2026-04-22 02:48:27 -05:00
Author
Owner
Copy Link

@phalexo commented on GitHub (Dec 7, 2023):

If you wanted to do the work, you could probably set up a process pool, which would take work off its own queue and then managing multiple ollama(s) running against different GPUs. It is messy of course.

<!-- gh-comment-id:1845903319 --> @phalexo commented on GitHub (Dec 7, 2023): If you wanted to do the work, you could probably set up a process pool, which would take work off its own queue and then managing multiple ollama(s) running against different GPUs. It is messy of course.
GiteaMirror commented 2026-04-22 02:48:27 -05:00
Author
Owner
Copy Link

@ciliamadani commented on GitHub (Dec 19, 2023):

Hello @austin-starks, what alternatives are you considering if I may know, I'm in the same situation and would like to know the available options.
Thank you in advance.

<!-- gh-comment-id:1863456852 --> @ciliamadani commented on GitHub (Dec 19, 2023): Hello @austin-starks, what alternatives are you considering if I may know, I'm in the same situation and would like to know the available options. Thank you in advance.
GiteaMirror commented 2026-04-22 02:48:28 -05:00
Author
Owner
Copy Link

@ParisNeo commented on GitHub (Jan 9, 2024):

Maybe you can create multiple instances of the server with different port numbers

<!-- gh-comment-id:1882610703 --> @ParisNeo commented on GitHub (Jan 9, 2024): Maybe you can create multiple instances of the server with different port numbers
GiteaMirror commented 2026-04-22 02:48:29 -05:00
Author
Owner
Copy Link

@ParisNeo commented on GitHub (Jan 9, 2024):

For now Ollama is designed to provide a great experience on your local machine for a single user. It is designed to queue the request and then it will get to the next request after the current one is complete. We intend to look at making a better server experience in the future.

It is fine to have a single generation with a queue. But can you add an endpoint to ask fir que status?

I want to launch multiple instances of the server to solve this multi users problem. But i need to put a disparcher in front of them. I need to be able to know the state of each queue to dispatch the connections.

<!-- gh-comment-id:1882732693 --> @ParisNeo commented on GitHub (Jan 9, 2024): > For now Ollama is designed to provide a great experience on your local machine for a single user. It is designed to queue the request and then it will get to the next request after the current one is complete. We intend to look at making a better server experience in the future. It is fine to have a single generation with a queue. But can you add an endpoint to ask fir que status? I want to launch multiple instances of the server to solve this multi users problem. But i need to put a disparcher in front of them. I need to be able to know the state of each queue to dispatch the connections.
GiteaMirror commented 2026-04-22 02:48:29 -05:00
Author
Owner
Copy Link

@m0wer commented on GitHub (Jan 17, 2024):

Sounds like something https://github.com/vllm-project/vllm has sorted out (queuing + configurable number of workers).

<!-- gh-comment-id:1895745260 --> @m0wer commented on GitHub (Jan 17, 2024): Sounds like something https://github.com/vllm-project/vllm has sorted out (queuing + configurable number of workers).
GiteaMirror commented 2026-04-22 02:48:30 -05:00
Author
Owner
Copy Link

@ParisNeo commented on GitHub (Jan 18, 2024):

except vllm doesn't know how to run GGUF models and is very hungry in terms of memory consumption.

<!-- gh-comment-id:1898010168 --> @ParisNeo commented on GitHub (Jan 18, 2024): except vllm doesn't know how to run GGUF models and is very hungry in terms of memory consumption.
GiteaMirror commented 2026-04-22 02:48:30 -05:00
Author
Owner
Copy Link

@m0wer commented on GitHub (Jan 18, 2024):

except vllm doesn't know how to run GGUF models and is very hungry in terms of memory consumption.

Agreed. Would be great to have parallelism in Ollama instead.

<!-- gh-comment-id:1898067655 --> @m0wer commented on GitHub (Jan 18, 2024): > except vllm doesn't know how to run GGUF models and is very hungry in terms of memory consumption. Agreed. Would be great to have parallelism in Ollama instead.
GiteaMirror commented 2026-04-22 02:48:31 -05:00
Author
Owner
Copy Link

@easp commented on GitHub (Jan 18, 2024):

Depending on what level of concurrency you are after, quantization can be counterproductive. Using quantized models requires more compute, which ends up becoming the bottleneck at higher levels of concurrency.

<!-- gh-comment-id:1898877492 --> @easp commented on GitHub (Jan 18, 2024): Depending on what level of concurrency you are after, quantization can be counterproductive. Using quantized models requires more compute, which ends up becoming the bottleneck at higher levels of concurrency.
GiteaMirror commented 2026-04-22 02:48:32 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Jan 26, 2024):

Going to close this as a dupe of #358

<!-- gh-comment-id:1912842579 --> @pdevine commented on GitHub (Jan 26, 2024): Going to close this as a dupe of #358
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#26503
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?