Skip to content

Add tiered KV cache offloading blog post - #304

Open
orozery wants to merge 4 commits into
vllm-project:mainfrom
orozery:tiered-kv-offloading-blog
Open

Add tiered KV cache offloading blog post#304
orozery wants to merge 4 commits into
vllm-project:mainfrom
orozery:tiered-kv-offloading-blog

Conversation

@orozery

@orozery orozery commented Aug 11, 2026

Copy link
Copy Markdown
Contributor

Covers the host-centric architecture, secondary tiers (filesystem, object storage, P2P), hybrid model support, observability, KV events, and performance benchmarks.

Covers the host-centric architecture, secondary tiers (filesystem,
object storage, P2P), hybrid model support, observability, KV events,
and performance benchmarks.

Signed-off-by: Or Ozeri <oro@il.ibm.com>
Signed-off-by: Or Ozeri <oro@il.ibm.com>
Signed-off-by: Or Ozeri <oro@il.ibm.com>
@orozery
orozery marked this pull request as ready for review August 11, 2026 16:13
Signed-off-by: Or Ozeri <oro@il.ibm.com>
Uses ZMQ for coordination and RDMA (via NIXL) for bulk data transfer.
All transfers are **host-to-host** — no accelerator memory involved on either side.

P2P transfers are triggered via `kv_transfer_params` request headers, typically managed by an external orchestrator such as [llm-d](https://github.com/llm-d/llm-d).

@liranschour liranschour Aug 17, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You mention kv_transfer_params but the code section below captures the kv-transfer-config cli. I suggest to add a small section describing the kv_transfer_params protocol for p2p/pd. You can take it from here: p2p-issue

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants