Folder: papers

attention
Azure SmartNICs
discrete event simulation
Dragonfly topology
elbencho
InfiniBand
IOR
LLM training
LLM training at scale
LLM training datasets
mdtest
Microsoft supercomputers
MTBF, FIT, and AFR
NFS
Nuclear energy
Scaling laws
Sustainability in HPC
thermodynamics
Transformers
Ultra Ethernet
3FS
AI datacenters
AI for science
AMD CDNA
AMD Helios
autoregressive decoding
availability
Azure infrastructure
B-tree
benchmarking
blobfuse
Broadcom Tomahawk 5
bullshitter
BXI
cables and connectors
Canadian sovereign AI
capex
checkpointing
Chesapeake Large Scale Analytics Conference
China
CIFAR
Coarse-Grained Reconfigurable Array
component reliability
continuous batching
Cray E2000
Cray EX
Cray GX
Cray K3000
Crusoe
DAOS
Darshan
DDN Infinia
decode
DeepSeek-R1
Denvr Dataworks
digital garden
Digital Research Alliance of Canada
distillation
Dragonfly+ topology
DRAM architecture
Duane Arnold
ECC schemes
EDA
energy based model
European sovereign AI
excursions
FalconFS
FASST
fio
foundation models for science
Frontier models
GekkoFS
Genesis
Government's role in AI
GPU terminology decoder ring
GPUaaS
GSP and SMC
GTC25
Hammerspace
head of line blocking
High-Performance Linpack
HPC vs AI
Hypertec
IBM SSS6000
IR7000
Job Mean Time To Interrupt
job reliability
jobs
KV cache
LCCF
Llama 4
LLM inferencing
Low-precision training
LPDDR5 Reliability
Lustre
MAI-1
Mamba
manufacturing level
memory bandwidth
Mercury
Meta Catalina
Meta Llama-3.1
Meta Movie Gen
Minipack2
MLPerf Storage
Model FLOPs Utilization
Mt. Diablo
Multi-plane topologies
multicluster training
NAIRR
natural gas
Network flow
Networking for LLM training
Neuromorphic computing
New Frontiers
Niagara
NVIDIA Vera
NVLink
Obsidian
Oil & gas
OpenAI x Microsoft
OPT-175B
Optical circuit switch
Palisades nuclear plant
PCIe Gen6
peak FLOPS
peer review
pNFS Flexible File Layout
pod
Podman
POSIX
power infrastructure
prefill
prefix caching
Productivity tools
QScale
quantum
RAG
Read-it-later apps
Reasoning models
reinforcement learning
Salishan
SC Conference
sequence parallelism
Signal modulation
silent data corruption
Slingshot
Slurm
Small Language Models
Social media platforms
sovereign AI
Spectrum-X Photonics
Stargate Project
startups
storage benchmarking is dumb
Storage for LLM training
Structured sparsity
Superintelligence
Synthetic data
System architect
Tensor cores and Matrix cores
test-time compute
Three Mile Island
training vs inference
VAST
VAST NFS
VAST vector database
vector database
Vision language model
VXLAN
WEKA
wisdom

Glenn's Digital Garden

Explorer

Folder: papers

Explorations of RDMA in LLM Systems

Future Generation High Performance Computing Center (FG-HPCC)

Revisiting Reliability in Large-Scale Machine Learning Research Clusters

Pre-Training GPT-4.5

Machines of Loving Grace

Nuclear finance will rely on consumers’ stomach for risk

The National Security Case for Public AI - Vanderbilt Policy Accelerator

The Intelligence Age

Recommendations on Powering Artificial Intelligence and Data Center Infrastructure

Carbon-Removal Firms Have One Very Big Backer. That’s a Problem

Datacenters to emit 3x more carbon dioxide because of genAI

Big Tech Is Rushing to Find Clean Power to Fuel AI’s Insatiable Appetite

A democratic vision for artificial intelligence must prevail over an authoritarian one.

FASST RFI

Azure Accelerated Networking: SmartNICs in the Public Cloud