This is really interesting. For SOTA inference systems, I've seen two general approaches:

* The "stack-centric" approach such as vLLM production stack, AIBrix, etc. These set up an entire inference stack for you including KV cache, routing, etc.

* The "pipeline-centric" approach such as NVidia Dynamo, Ray, BentoML. These give you more of an SDK so you can define inference pipelines that you can then deploy on your specific hardware.

It seems like LLM-d is the former. Is that right? What prompted you to go down that direction, instead of the direction of Dynamo?

It sounds like you might be confusing different parts of the stack. NVIDIA Dynamo for example supports vLLM as the inference engine. I think you should think of something like vLLM as more akin to GUnicorn, and llm-d as an application load balancer. And I guess something like NVIDIA Dynamo would be like Django.

llm-d is intended to be three clean layers:

1. Balance / schedule incoming requests to the right backend

2. Model server replicas that can run on multiple hardware topologies

3. Prefix caching hierarchy with well-tested variants for different use cases

So it's a 3-tier architecture. The biggest difference with Dynamo is that llm-d is using the inference gateway extension - https://github.com/kubernetes-sigs/gateway-api-inference-ext... - which brings Kubernetes owned APIs for managing model routing, request priority and flow control, LoRA support etc.

I would think that that the NVidia Dynamo SDK (pipelines) is a big difference as well (https://github.com/ai-dynamo/dynamo/tree/main/deploy/sdk/doc...), or am I missing something?

That's a good example - I can at least answer about why it's a difference: different target user.

As I understand the Dynamo SDK it is about simplifying and helping someone get started with Dynamo on Kubernetes.

From the user set we work with (large inference deployers) that is not a high priority - they already have mature deployment opinions or a set of tools that would not compose well with something like the Dynamo SDK. Their comfort level with Kubernetes is moderate to high - either they use Kubernetes for high scale training and batch, or they are deploying to many different providers in order to get enough capacity and need a standard orchestration solution.

llm-d focuses on helping achieve efficiency dynamically at runtime based on changing traffic or workload on Kubernetes - some of the things the Dynamo SDK encodes are static and upfront and would conflict with that objective. Also, large deployers with serving typically have significant batch and training and they are looking to maximize capacity use without impacting their prod serving. That requires the orchestrator to know about both workloads at some level - which Dynamo SDK would make more difficult.

In this analogy, Dynamo is most definitely not like Django. It includes inference aware routing, KV caching, etc. -- all the stuff you would need to run a modern SOTA inference stack.

You're right, I was confusing TensorRT with Dynamo. It looks like the relationship between Dynamo and vLLM is actually the opposite of what I was thinking -- Dynamo can use vLLM as a backend rather than vice versa.

What would be the benefit of this project over hosting VLLM in Ray?

I did a quick scan of the repo and didn't see any reference to Ray. Would this indicate that llm-d lacks support for pipeline parallelism?

We inherit any multi-host support from vLLM, so https://docs.vllm.ai/en/latest/serving/distributed_serving.h... would be the expected path.

We plan to publish examples of multi-host inference that leverages LeaderWorkerSets - https://github.com/kubernetes-sigs/lws - which helps run ranked serving workloads across hosts. LeaderWorkerSet is how Google supports both TPU and GPU multi-host deployments - see https://github.com/kubernetes-sigs/lws/blob/main/config/samp... for an example.

Edit: Here is an example Kubernetes configuration running DeepSeek-R1 on vLLM multi-host using LeaderWorkerSet https://github.com/kubernetes-sigs/wg-serving/blob/main/serv.... This work would be integrated into llm-d.

I believe this is a question you should ask about vLLM, not llm-d. It looks like vLLM does support pipeline parallelism via Ray: https://docs.vllm.ai/en/latest/serving/distributed_serving.h...

This project appears to make use of both vLLM and Inference Gateway (an official Kubernetes extension to the Gateway resource). The contributions of llm-d itself seems to mostly be a scheduling algorithm for load balancing across vLLM instances.

I wonder if this is preferable to kServe

llm-d would make sense if you are running a very large production LLM serving setup - say 5+ full H100 hosts. The aim is to be much more focused than kserve is on exactly the needs of serving LLMs. It would of course be possible to run alongside kserve, but the user we are targeting is not typically a kserve deployer today.

Do you think https://github.com/openai/CLIP can be ran on it? LLM makes me think of chatbots but I suppose because it's inference-based it would work. Somewhat unclear on what's the difference between LLMs and inference, I think inference is the type of compute LLMs use.

I wonder if inference-d would be a fitting name.

Inference is the process of evaluating a model ("inferring" a response to the inputs). LLMs are uniquely difficult to serve because they push the limits on the hardware.

The models we support come from the model server vLLM https://docs.vllm.ai/en/latest/models/supported_models.html, which has a focus on large generative models. I don't see CLIP in the list.