<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:image="http://www.google.com/schemas/sitemap-image/1.1">
<url>
<loc>https://runtimereview.co</loc>
<lastmod>2026-11-18T20:13:08.300Z</lastmod>
<changefreq>daily</changefreq>
<priority>1</priority>
</url>
<url>
<loc>https://runtimereview.co/compare</loc>
<lastmod>2026-11-18</lastmod>
<changefreq>monthly</changefreq>
<priority>0.6</priority>
</url>
<url>
<loc>https://runtimereview.co/compare/vllm-vs-ollama</loc>
<lastmod>2026-11-18</lastmod>
<changefreq>monthly</changefreq>
<priority>0.8</priority>
</url>
<url>
<loc>https://runtimereview.co/sections/long-context-serving</loc>
<lastmod>2026-11-18T20:13:08.300Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://runtimereview.co/sections/distributed-serving</loc>
<lastmod>2026-11-13T21:37:27.575Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://runtimereview.co/sections/model-to-hardware-fit</loc>
<lastmod>2026-10-17T19:35:47.941Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://runtimereview.co/sections/benchmarking</loc>
<lastmod>2026-10-02T19:40:52.475Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://runtimereview.co/sections/inference-runtime</loc>
<lastmod>2026-09-26T06:00:18.951Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://runtimereview.co/sections/features</loc>
<lastmod>2026-09-18T06:51:02.322Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://runtimereview.co/authors/darius-okonkwo</loc>
<lastmod>2026-11-18T20:13:08.300Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.4</priority>
</url>
<url>
<loc>https://runtimereview.co/authors/mei-ling-tsao</loc>
<lastmod>2026-11-18T20:13:08.300Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.4</priority>
</url>
<url>
<loc>https://runtimereview.co/authors/priya-subramanian</loc>
<lastmod>2026-11-18T20:13:08.300Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.4</priority>
</url>
<url>
<loc>https://runtimereview.co/authors/tobias-wrenner</loc>
<lastmod>2026-11-18T20:13:08.300Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.4</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/eviction-policy-design-when-kv-cache-growth-saturates-gpu-memory-in-serving-clus</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/b9a3b627-dc47-497b-a07e-b23408879273.jpeg</image:loc>
</image:image>
<lastmod>2026-11-18T20:13:08.300Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/cross-request-kv-cache-persistence-patterns-for-session-aware-llm-serving</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/973f41ba-2b7e-4dad-9efe-27cacaf0dc95.jpeg</image:loc>
</image:image>
<lastmod>2026-11-17T19:43:31.798Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/per-request-kv-cache-budget-enforcement-as-the-primary-cost-lever-in-long-contex</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/5028aa92-609f-4032-8e62-849f5a95d98a.jpeg</image:loc>
</image:image>
<lastmod>2026-11-16T20:23:01.400Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/disaggregated-storage-backends-versus-monolithic-gpu-memory-the-infrastructure-s</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/eaf90fb5-d466-4899-8268-d80c9bb855b0.jpeg</image:loc>
</image:image>
<lastmod>2026-11-15T21:38:03.793Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/load-balancing-strategies-for-heterogeneous-multi-node-llm-serving-fleets</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/9aaa51a1-2573-40c3-af5b-0362a1445d1b.jpeg</image:loc>
</image:image>
<lastmod>2026-11-13T21:37:27.575Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/weight-streaming-architectures-that-eliminate-gpu-idle-time-during-model-loading</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/d11dc66a-d6e5-4013-96dc-d4e7885223c9.jpeg</image:loc>
</image:image>
<lastmod>2026-11-12T19:42:59.140Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/why-storage-latency-becomes-the-binding-constraint-in-production-llm-inference-p</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/ab8695e6-7979-49a2-9329-0288c35b5ecf.jpeg</image:loc>
</image:image>
<lastmod>2026-11-11T20:02:57.373Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/distributed-kv-cache-coordination-across-gpu-node-boundaries-in-multi-host-infer</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/2614235b-cb1b-4c73-8ebf-bbc7f276d30a.jpeg</image:loc>
</image:image>
<lastmod>2026-11-10T21:56:56.567Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/how-hyperscale-llm-api-teams-architect-kv-cache-management-under-real-throughput</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/c406831a-d94a-48fd-9d32-4f1a71082307.jpeg</image:loc>
</image:image>
<lastmod>2026-11-09T20:41:49.779Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/the-storage-layer-rethink-that-unlocks-context-scaling-beyond-what-gpu-memory-ca</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/df44c71b-6347-4927-b00b-c0ae0d855ffd.jpeg</image:loc>
</image:image>
<lastmod>2026-11-08T20:51:02.586Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/beyond-gpu-memory-walls-the-hidden-bottlenecks-throttling-long-context-transform</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/d15aecd6-7e16-4f57-9e8b-f50fba0c4dfc.jpeg</image:loc>
</image:image>
<lastmod>2026-11-07T20:21:19.404Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/tiered-storage-layout-decisions-that-determine-sustainable-throughput-ceilings-i</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/2ade1faa-e243-4154-9e64-bd325c6de0c0.jpeg</image:loc>
</image:image>
<lastmod>2026-11-06T21:26:51.477Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/nvme-tier-bandwidth-accounting-as-the-primary-driver-of-cost-per-token-in-llm-se</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/0beab167-c397-4a67-8727-0bfd8a264473.jpeg</image:loc>
</image:image>
<lastmod>2026-11-05T20:56:08.187Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/storage-first-architecture-decisions-for-llm-serving-systems-targeting-million-t</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/919dc133-1474-4968-8972-1cde52f68dd9.jpeg</image:loc>
</image:image>
<lastmod>2026-11-03T19:51:41.309Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/kv-cache-offloading-tools-ranked-by-production-inference-overhead</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/03d48d3b-33fc-4ee9-8147-e09acd04e9a7.jpeg</image:loc>
</image:image>
<lastmod>2026-11-02T21:32:27.235Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/kv-cache-storage-tiering-across-dram-hbm-and-ssd-in-million-token-serving-pipeli</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/3a77318c-2de2-4ff5-8e2d-529da7f8bb5c.jpeg</image:loc>
</image:image>
<lastmod>2026-10-31T19:51:19.423Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/multi-node-inference-cluster-network-topology-for-ai-workloads</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/cd13f96a-700d-4165-9e4a-cc4e35697266.jpeg</image:loc>
</image:image>
<lastmod>2026-10-29T20:22:00.331Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/disaggregated-kv-cache-architectures-across-gpu-memory-and-nvme-tiers</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/cb150edc-455e-46e5-a823-0b01937ad787.jpeg</image:loc>
</image:image>
<lastmod>2026-10-28T20:56:31.192Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/networked-nvme-fabrics-for-shared-kv-cache-in-multi-node-serving-clusters</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/f392da64-1e1b-4947-bd54-a93f04af1345.jpeg</image:loc>
</image:image>
<lastmod>2026-10-27T20:46:53.048Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/storage-bandwidth-saturation-points-that-collapse-llm-inference-slas-at-scale</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/15a987c5-118c-4281-a79f-71c768d2d478.png</image:loc>
</image:image>
<lastmod>2026-10-26T20:26:20.308Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/kv-cache-memory-pooling-and-allocation-strategies-for-high-throughput-llm-servin</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/1bf44ed1-8012-4ca5-9db9-bff5165edea5.jpeg</image:loc>
</image:image>
<lastmod>2026-10-25T20:31:31.341Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/why-long-context-llm-serving-is-a-storage-architecture-problem-before-a-model-pr</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/64fefbf8-4dc4-4aa6-8e5a-e4dd25d9eb71.jpeg</image:loc>
</image:image>
<lastmod>2026-10-24T18:35:47.550Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/rdma-over-infiniband-for-kv-cache-transfer-between-prefill-and-decode-nodes</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/f5bef28b-8620-4044-8245-7e6771d1ec17.jpeg</image:loc>
</image:image>
<lastmod>2026-10-23T19:15:59.311Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/tensor-parallelism-degree-selection-for-llama-and-mixtral-multi-gpu-deployments</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/946abadf-2d06-4e5f-96d0-eadb1b4714c8.jpeg</image:loc>
</image:image>
<lastmod>2026-10-22T19:41:24.518Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/attention-kernel-efficiency-at-long-sequence-lengths-in-flash-attention-3</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/ad656e66-d27e-44d6-846e-861a35392d2a.jpeg</image:loc>
</image:image>
<lastmod>2026-10-21T20:36:12.644Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/multi-round-conversation-state-management-at-scale-for-long-contexts</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/7b46afeb-91dc-474a-ab66-d4427f71c98f.jpeg</image:loc>
</image:image>
<lastmod>2026-10-20T19:21:48.448Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/prefix-caching-effectiveness-for-repeated-long-system-prompts</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/c6c70a41-4dee-4362-bd2b-f391aa2d5933.jpeg</image:loc>
</image:image>
<lastmod>2026-10-19T20:12:11.881Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/chunked-prefill-strategies-for-128k-token-inputs-in-vllm</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/76a6dd62-cc3f-40bf-afb6-da1a41f155d7.jpeg</image:loc>
</image:image>
<lastmod>2026-10-18T19:20:46.941Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/fp8-vs-bf16-inference-throughput-and-memory-tradeoffs-for-llama-and-qwen-models</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/02a12d47-089a-4191-8050-8f3f845160c8.jpeg</image:loc>
</image:image>
<lastmod>2026-10-17T19:35:47.941Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/gpu-direct-storage-integration-with-open-source-inference-runtimes</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/534869ac-8f19-4b68-aef2-eb87cae79585.jpeg</image:loc>
</image:image>
<lastmod>2026-10-16T20:16:33.297Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/llama-3-1-405b-multi-node-serving-hardware-requirements</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/7841ff1a-41cd-41cc-8764-11481f963eda.jpeg</image:loc>
</image:image>
<lastmod>2026-10-15T19:31:37.291Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/selecting-nvme-drive-count-and-topology-for-kv-cache-offloading-workloads</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/692ad109-4e5e-4214-a3a9-b34f7e21e819.jpeg</image:loc>
</image:image>
<lastmod>2026-10-13T19:21:57.414Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/deepseek-r1-moe-routing-and-nvme-bandwidth-requirements</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/3773925b-a365-4161-a85a-8093a31dda6b.jpeg</image:loc>
</image:image>
<lastmod>2026-10-12T20:16:25.591Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/qwen2-5-72b-tensor-parallelism-across-nvlink-vs-pcie-gpu-topologies</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/38498ef8-f664-4552-8242-e4dd92b7c00f.jpeg</image:loc>
</image:image>
<lastmod>2026-10-11T20:01:07.053Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/llama-3-70b-serving-on-h100-vs-a100-nvme-offloading-configurations</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/b46c8fca-a4a3-4d04-8931-c7774d4e6611.jpeg</image:loc>
</image:image>
<lastmod>2026-10-10T19:55:46.898Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/disaggregated-prefill-and-decode-for-long-context-request-handling</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/ee774cae-29e7-4a5a-ad6e-b93795ebc20c.jpeg</image:loc>
</image:image>
<lastmod>2026-10-09T19:30:15.171Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/nvme-offloading-for-kv-cache-in-long-context-vllm-deployments</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/bdd67954-c304-47ff-9b62-e61b4a8b51ec.jpeg</image:loc>
</image:image>
<lastmod>2026-10-08T20:21:21.857Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/kv-cache-memory-scaling-from-32k-to-1m-token-contexts</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/a2ca1089-f06c-4153-97a0-641b9c336738.jpeg</image:loc>
</image:image>
<lastmod>2026-10-07T19:36:02.628Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/pipeline-parallelism-latency-bubbles-and-mitigation-in-multi-node-inference</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/24955237-304f-4eb3-8f61-2caab3f9b3ef.jpeg</image:loc>
</image:image>
<lastmod>2026-10-06T20:25:48.032Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/kv-cache-compression-techniques-for-long-context-llm-serving</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/4827788b-0e1f-44f4-9a4b-c38a442ca7d9.jpeg</image:loc>
</image:image>
<lastmod>2026-10-04T19:50:16.575Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/mistral-7b-and-mixtral-8x7b-gpu-memory-footprint-comparison-at-scale</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/5246e65a-2d4e-4de4-91ec-597ef6817216.jpeg</image:loc>
</image:image>
<lastmod>2026-10-03T20:44:59.035Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/tokens-per-second-vs-latency-tradeoffs-in-throughput-optimized-serving-configs</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/0c69355d-17ad-4cf6-9cf9-a7bf3030a123.jpeg</image:loc>
</image:image>
<lastmod>2026-10-02T19:40:52.475Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/identifying-storage-i-o-as-the-bottleneck-in-gpu-utilization-profiles</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/d2038420-0015-4528-9b7e-3cf7d97fe974.jpeg</image:loc>
</image:image>
<lastmod>2026-10-01T20:15:13.289Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/measuring-time-to-first-token-and-inter-token-latency-under-sustained-load</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/c45babef-4fd7-46ea-a081-5866528cfffd.jpeg</image:loc>
</image:image>
<lastmod>2026-09-30T19:00:32.703Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/designing-a-kv-cache-throughput-harness-for-production-serving-nodes</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/a1d35592-1bd4-4174-9b5b-18aefc178f3c.jpeg</image:loc>
</image:image>
<lastmod>2026-09-30T17:26:24.566Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/benchmarking-nvme-sequential-and-random-read-performance-for-kv-cache-patterns</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/4008c5a7-41b5-4a2a-9f88-8a481f916363.jpeg</image:loc>
</image:image>
<lastmod>2026-09-30T16:26:02.845Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/reproducing-mlperf-inference-benchmarks-on-open-source-serving-stacks</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/0c73653c-0de8-4c43-b33e-28dbb7cbf571.jpeg</image:loc>
</image:image>
<lastmod>2026-09-29T07:35:30.909Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/mlperf-storage-benchmark-setup-and-interpretation-for-ai-inference-workloads</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/a2ffe067-005e-4737-ad52-a2f5507db3bd.jpeg</image:loc>
</image:image>
<lastmod>2026-09-27T06:05:01.137Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/speculative-decoding-overhead-in-production-vllm-deployments</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/71401f41-e5b6-435f-870a-288771c1d9d9.jpeg</image:loc>
</image:image>
<lastmod>2026-09-26T06:00:18.951Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/tensorrt-llm-paged-kv-cache-configuration-for-multi-lora-serving</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/a96068c1-1439-48de-930a-387a337e2e24.jpeg</image:loc>
</image:image>
<lastmod>2026-09-25T00:59:24.351Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/sglang-vs-vllm-throughput-under-long-context-request-mixes</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/24365dd1-f902-4001-b7da-16908aad243e.jpeg</image:loc>
</image:image>
<lastmod>2026-09-24T01:10:08.708Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/vllm-chunked-prefill-scheduling-and-ttft-optimization</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/01a0e2bd-35d4-49d8-90c0-efbda0f3e0da.jpeg</image:loc>
</image:image>
<lastmod>2026-09-22T19:06:39.547Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/llama-cpp-kv-cache-quantization-and-memory-tradeoffs</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/3b5c03f4-608b-4ac5-a830-c82cc00579e6.jpeg</image:loc>
</image:image>
<lastmod>2026-09-20T19:08:40.741Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/tensorrt-llm-in-flight-batching-vs-vllm-continuous-batching</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/0004cdf8-5b41-41c6-a007-8d4e65113960.jpeg</image:loc>
</image:image>
<lastmod>2026-09-18T19:24:18.889Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://runtimereview.co/posts/vllm-pagedattention-memory-allocation-under-concurrent-load</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/2cf2d27a-ffce-415c-a989-3e56c5a058ba/generated/ef0cf1c9-b3b3-4b5f-9264-d78688e8385a.jpeg</image:loc>
</image:image>
<lastmod>2026-09-18T06:51:02.322Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
</urlset>
