AI & ML
A Flat KV Cache Storage Architecture for LLM Serving
arXiv cs.DC0 views
arXiv:2610.38355v2 Announce Type: replace Abstract: KV cache reuse accelerates LLM serving, but existing systems manage host DRAM and NVMe SSD as a tiered hierarchy where DRAM acts as an intermediate cache. Production LLM workloads break both premises of this hierarchy. First, bimodal reuse defeats DRAM temporal filtering, since workloads are dominated either by one-shot blocks with long reuse distances or by multi-gigabyte agentic sessions that exceed DRAM capacity, putting SSD on the critical read path. Second, DRAM staging and eviction-ordered layout bottleneck SSD restoration. Holding host staging buffers for in-flight prefixes causes contention under load that aborts prefetches, while incremental eviction scatters a session's blocks across the SSD and prevents large sequential reads. We present FlatCake, a flat KV storage architecture that exposes host DRAM and NVMe SSD as peer media in a unified 64-bit address space. Exploiting the fact that KV blocks are write-once, immutable, and reused together within a prompt or session, FlatCake places each block directly onto its target medium in session-contiguous, layer-major order at creation time and never migrates it. FlatCake streams coalesced SSD reads directly into a bounded GPU window via GPUDirect Storage and decouples DRAM and SSD request queues to prevent head-of-line blocking. We implement FlatCake in SGLang and compare it against host-staged (Mooncake) and GDS-enabled (Mooncake-GDS) tiered storage, as well as SSD-direct storage (Tutti). Across three production agentic, coding, and conversational traces with DeepSeek-V4-Flash, FlatCake reduces mean TTFT near the baselines' latency knees by $1.40\times$--$2.42\times$ over the fastest baseline. On the agentic Claude Code trace, FlatCake sustains a 91.6% cache hit rate under load (versus 71.1% for all three baselines) and delivers up to $2.16\times$ higher token throughput.
Source:https://arxiv.org/abs/2610.38355PreviousThe Cost of Synchronization in GPU Scale-Up DomainsNextMUSE: Multi-Tenant Model Serving With Seamless Model Updates
Related posts
Razorpay webhook not verified: anyone can fake a payment
0dev.to
From FAQPage Schema to Answerable AI Documentation
0dev.to
Wiring Android's MediaPipe LLM Inference API to a Multi-Turn Conversation Engine
0dev.to
What Is LLM Throughput? Tokens per Second, Concurrency, and Batching
0dev.to
Diese Woche haben mir zwei Tools etwas vorgeschlagen und es direkt abgeschickt.
0dev.to
AI Agent Observability: What to Log Before Production
0dev.to