Prefix caching in vLLM under multi-tenant agent traffic
Nexus Labs implemented prefix caching in vLLM to improve latency for their multi-tenant agent workloads. The results showed a significant reduction in time-to-first-token (TTFT) for one tenant, while another tenant faced challenges due to their dynamic prompt structure. Adjustments were made to optimize performance, leading to improved efficiency across the board.
- ▪Prefix caching in vLLM reduced TTFT from 480ms to 110ms for Tenant A.
- ▪Tenant B's dynamic prompt structure initially resulted in a low cache hit rate of 0.3%.
- ▪After refactoring, Tenant B improved their TTFT from 510ms to 145ms with an 87% hit rate.
DEV.to (Top) files mainly under programming. We currently carry 4,924 of its stories.
Story provenance
Source · retrieval · rights · ranking — open for full record
inspect →
Story provenance
Attribution is not the same as permission. This drawer separates discovery metadata, excerpts, WeSearch-generated summaries, reuse status, and whether the publisher receives the visit. Nothing here claims a legal grant the publisher has not made.
Record
| Original publisher | DEV.to (Top) |
| Canonical URL | https://dev.to/marcuswwchen/prefix-caching-in-vllm-under-multi-tenant-agent-traffic-5e2j |
| Publication time | Tue, 26 May 2026 06:35:20 +0000 |
| Retrieval time | 2026-05-26T07:07:46.336Z |
| Last seen | 2026-05-26T07:07:46.336Z |
| Headline source | Publisher (no WeSearch rewrite) |
| Excerpt source | publisher body |
| Excerpt method | First ~120 words (~800 chars) of extracted publisher body, fair-use limited. |
| Summary | WeSearch · cerebras-chat (WeSearch summarizer) |
| Summary source text | contentText |
| Citation coverage | Summary is a WeSearch-generated derivative; primary citation is the original publisher URL. |
| Cluster | 9AL7VOaGhRWs |
| Cluster logic | Grouped by semantic title/content similarity across sources within a rolling window. Same-publisher template collisions are excluded from coverage comparison. |
| Ranking reason | Story pages are not engagement-ranked. Hub feeds use recency, with optional source-diversified chronological ordering (cap consecutive stories per source). No personalized ranking. |
| Publisher visit | Yes — open original |
| Substitutes article? | No — link-out required for full text |
Rights status (four layers)
WeSearch handling by dimension
| Indexing | May the item be indexed (stored, ranked, made findable)? | Allowed |
| Snippet | May a short excerpt of the publisher's text be shown? | Allowed |
| AI summary | May WeSearch generate its own short summary of the article? | Limited |
| Retrieval / RAG | May the content be exposed for third-party retrieval-augmented generation? | Not asserted |
| Model training | May the content be used to train AI models? | Not asserted |
| Commercial reuse | May the content be reused commercially? | Not permitted |
Basis: Derived from the published RSS/Atom feed. Contact: [email protected]. Reviewed: 2026-07-24.
Opening excerpt (first ~120 words) tap to expand
try { if(localStorage) { let currentUser = localStorage.getItem('current_user'); if (currentUser) { currentUser = JSON.parse(currentUser); if (currentUser.id === 3859428) { document.getElementById('article-show-container').classList.add('current-user-is-article-author'); } } } } catch (e) { console.error(e); } Marcus Chen Posted on May 26 Prefix caching in vLLM under multi-tenant agent traffic #llm #mlops #infrastructure #pytorch TL;DR: We turned on vLLM's prefix cache for our agent workloads at Nexus Labs and watched TTFT drop from 480ms to 110ms on one tenant and stay exactly the same on another. The split wasn't about traffic volume. It was about how each team templated their system prompts. The setup Our fine-tuning team serves 14 enterprise agents through a shared inference cluster.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at DEV.to (Top).
Discussion
0 commentsMore from DEV.to (Top)
