23 docs tagged with "scope:tech"

EKS GPU node strategy, Karpenter·KEDA·DRA resource management, NVIDIA GPU stack, AWS Neuron stack — the accelerated computing layer covering GPUs and AWS custom accelerators

Guide to Neuron SDK, Device Plugin, and NxD Inference for operating AWS custom AI accelerators (Trainium2/Inferentia2) on EKS

AWS Nitro System components, per-generation (v2–v6) networking changes, ENA driver/kernel requirements on EKS nodes, and PPS/CPS-focused performance tuning strategies

Unifying the three layers of inference caching (KV/Prefix, Prompt, Semantic) into a single decision framework with hit-rate targets, measurement points, and tuning levers for each layer.

Technical status and EKS application scenarios for GPU workload checkpoint/restore during Spot reclaim and scheduling events (Experimental)

Prefill/Decode separation architecture and NIXL common KV transfer engine, LeaderWorkerSet-based 700B+ large MoE model multi-node deployment guide

Optimal node strategies for GPU workloads across EKS Auto Mode, Karpenter, MNG, and Hybrid Nodes

Compares Istio multi-cluster, Cilium ClusterMesh, and Amazon VPC Lattice architectures for East-West (service-to-service) communication across EKS clusters along four axes — functionality, stability, operability, and cost — with guidance on which environment suits each and a migration path from Istio

Compares data plane architecture, mTLS, L7 policy, observability, performance overhead, and operational complexity of major service mesh solutions on EKS, with workload-specific selection criteria

2-Tier GPU autoscaling (KEDA·Karpenter), DRA compatibility, and operational lessons learned from large MoE model (GLM-5·Kimi K2.5) deployments for LLM serving

GPU resource management and cost optimization using Karpenter, KEDA, and DRA on EKS

Comparing SageMaker HyperPod Inference Operator's managed KV cache, intelligent routing, and DPD with a Tiered Gateway, and clarifying its role and limitations as an L2 inference routing layer.

vLLM·llm-d·MoE·NeMo — AI framework layer for actual model serving, distributed inference, and fine-tuning on GPUs

EKS architecture overview for maximizing LLM Inference performance — starting point for vLLM, KV Cache-Aware Routing, Disaggregated Serving, LWS multi-node, and GPU autoscaling

Summary of core technologies like vLLM PagedAttention, Continuous Batching, FP8 KV Cache, and comparison of llm-d/NVIDIA Dynamo KV Cache-Aware Routing and Gateway configuration

llm-d architecture concepts, KV Cache-aware routing, Disaggregated Serving, EKS Auto Mode integration strategy

The concept of LMCache — offloading KV cache beyond GPU memory to CPU and disk and sharing it across inference instances — and its relationship to vLLM prefix cache, NIXL, and kvaware routing.

A guide to the GPU infrastructure, inference framework, and inference optimization layers, with a single map of the end-to-end LLM inference request path and per-layer tuning levers — inference gateway, prefill/decode disaggregation, KV cache-aware routing, LMCache, and cache-hit strategy.

Architecture concepts, distributed deployment strategies, and performance optimization principles for Mixture of Experts models

NVIDIA NeMo Framework distributed training, fine-tuning, and TensorRT-LLM conversion architecture

Architecture and EKS integration for GPU Operator, DCGM, MIG, Time-Slicing, and Dynamo

LLM Gateway-level semantic caching strategy and implementation options comparison (GPTCache, Redis Semantic Cache, Portkey, Helicone, Bifrost+Redis)

vLLM PagedAttention, parallelization strategies, Multi-LoRA, and hardware support architecture