A 1.48B-request production trace and generator show LLM serving demand splits into systematic signatures: small dense models jitter, MoE models burst, and input-output length follows an inverted bowl for dense models.
Gonçalves, Idilio Drago, Alex B
1 Pith paper cite this work, alongside 20 external citations. Polarity classification is still indexing.
1
Pith paper citing it
20
external citations · OpenAlex
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
FineServe: A Fine-Grained Dataset and Characterization of Global LLM Serving Workloads
A 1.48B-request production trace and generator show LLM serving demand splits into systematic signatures: small dense models jitter, MoE models burst, and input-output length follows an inverted bowl for dense models.