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 #1927] Handling High traffic #1108

New Issue
Closed
opened 2026-04-12 10:51:19 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-12 10:51:19 -05:00
Owner
Copy Link

Originally created by @lauvindra on GitHub (Jan 11, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/1927

Assume I have ollama in server Tesla T4 GPU with 16GB Vram and 120 Ram, how many request can it handle in one second?

Originally created by @lauvindra on GitHub (Jan 11, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/1927 Assume I have ollama in server Tesla T4 GPU with 16GB Vram and 120 Ram, how many request can it handle in one second?
GiteaMirror commented 2026-04-12 10:51:20 -05:00
Author
Owner
Copy Link

@easp commented on GitHub (Jan 11, 2024):

That card apparently has ~320GB/s bandwidth. Tokens/s generated is approximately 8 bits/byte *320 GB/s / (# model parameters * # bits per parameter). For a q4 quantization of a 4 bit model that's probably about 100 tokens/s.

Ollama currently queues concurrent requests and processes them serially. This isn't an efficient way to processes concurrent requests.

<!-- gh-comment-id:1887970002 --> @easp commented on GitHub (Jan 11, 2024): That card apparently has ~320GB/s bandwidth. Tokens/s generated is approximately 8 bits/byte *320 GB/s / (# model parameters * # bits per parameter). For a q4 quantization of a 4 bit model that's probably about 100 tokens/s. Ollama currently queues concurrent requests and processes them serially. This isn't an efficient way to processes concurrent requests.
GiteaMirror commented 2026-04-12 10:51:21 -05:00
Author
Owner
Copy Link

@jimscard commented on GitHub (Jan 12, 2024):

If Ollama is intended to be used as a local LLM system, then queuing requests and processing them serially is appropriate. Enabling concurrent processing, e.g., a "server" scenario would significantly increase the attack surface and complexity.

<!-- gh-comment-id:1888483084 --> @jimscard commented on GitHub (Jan 12, 2024): If Ollama is intended to be used as a *local* LLM system, then queuing requests and processing them serially is appropriate. Enabling concurrent processing, e.g., a "server" scenario would significantly increase the attack surface and complexity.
GiteaMirror commented 2026-04-12 10:51:21 -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:1912841199 --> @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#1108
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?