Zobrazeno 1 - 10
of 63
pro vyhledávání: '"De Sensi, Daniele"'
Autor:
Bonato, Tommaso, Kabbani, Abdul, Ghalayini, Ahmad, Dohadwala, Mohammad, Papamichael, Michael, De Sensi, Daniele, Hoefler, Torsten
Most existing datacenter transport protocols rely on in-order packet delivery, a design choice rooted in legacy systems and simplicity. However, advancements in technology, such as RDMA, have made it feasible to relax this requirement, allowing for m
Externí odkaz:
http://arxiv.org/abs/2407.21625
Autor:
Luczynski, Piotr, Gianinazzi, Lukas, Iff, Patrick, Wilson, Leighton, De Sensi, Daniele, Hoefler, Torsten
Efficient Reduce and AllReduce communication collectives are a critical cornerstone of high-performance computing (HPC) applications. We present the first systematic investigation of Reduce and AllReduce on the Cerebras Wafer-Scale Engine (WSE). This
Externí odkaz:
http://arxiv.org/abs/2404.15888
Autor:
Bonato, Tommaso, Kabbani, Abdul, De Sensi, Daniele, Pan, Rong, Le, Yanfang, Raiciu, Costin, Handley, Mark, Schneider, Timo, Blach, Nils, Ghalayini, Ahmad, Alves, Daniel, Papamichael, Michael, Caulfield, Adrian, Hoefler, Torsten
With the rapid growth of machine learning (ML) workloads in datacenters, existing congestion control (CC) algorithms fail to deliver the required performance at scale. ML traffic is bursty and bulk-synchronous and thus requires quick reaction and str
Externí odkaz:
http://arxiv.org/abs/2404.01630
Publikováno v:
NSDI 2024
The allreduce collective operation accounts for a significant fraction of the runtime of workloads running on distributed systems. One factor determining its performance is the distance between communicating nodes, especially on networks like torus,
Externí odkaz:
http://arxiv.org/abs/2401.09356
Autor:
Blach, Nils, Besta, Maciej, De Sensi, Daniele, Domke, Jens, Harake, Hussein, Li, Shigang, Iff, Patrick, Konieczny, Marek, Lakhotia, Kartik, Kubicek, Ales, Ferrari, Marcel, Petrini, Fabrizio, Hoefler, Torsten
Publikováno v:
Proceedings of the 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24) Santa Clara, CA, USA April 16-18, 2024
Novel low-diameter network topologies such as Slim Fly (SF) offer significant cost and power advantages over the established Fat Tree, Clos, or Dragonfly. To spearhead the adoption of low-diameter networks, we design, implement, deploy, and evaluate
Externí odkaz:
http://arxiv.org/abs/2310.03742
Autor:
De Sensi, Daniele, Molero, Edgar Costa, Di Girolamo, Salvatore, Vanbever, Laurent, Hoefler, Torsten
The allreduce operation is an essential building block for many distributed applications, ranging from the training of deep learning models to scientific computing. In an allreduce operation, data from multiple hosts is aggregated together and then b
Externí odkaz:
http://arxiv.org/abs/2309.16214
Autor:
Khalilov, Mikhail, Chrapek, Marcin, Shen, Siyuan, Vezzu, Alessandro, Benz, Thomas, Di Girolamo, Salvatore, Schneider, Timo, De Sensi, Daniele, Benini, Luca, Hoefler, Torsten
Multi-tenancy is essential for unleashing SmartNIC's potential in datacenters. Our systematic analysis in this work shows that existing on-path SmartNICs have resource multiplexing limitations. For example, existing solutions lack multi-tenancy capab
Externí odkaz:
http://arxiv.org/abs/2309.03628
Autor:
De Sensi, Daniele, De Matteis, Tiziano, Taranov, Konstantin, Di Girolamo, Salvatore, Rahn, Tobias, Hoefler, Torsten
Cloud computing represents an appealing opportunity for cost-effective deployment of HPC workloads on the best-fitting hardware. However, although cloud and on-premise HPC systems offer similar computational resources, their network architecture and
Externí odkaz:
http://arxiv.org/abs/2210.15315
Autor:
Hoefler, Torsten, Bonato, Tommaso, De Sensi, Daniele, Di Girolamo, Salvatore, Li, Shigang, Heddes, Marco, Belk, Jon, Goel, Deepak, Castro, Miguel, Scott, Steve
Numerous microarchitectural optimizations unlocked tremendous processing power for deep neural networks that in turn fueled the AI revolution. With the exhaustion of such optimizations, the growth of modern AI is now gated by the performance of train
Externí odkaz:
http://arxiv.org/abs/2209.01346
Autor:
Di Girolamo, Salvatore, De Sensi, Daniele, Taranov, Konstantin, Malesevic, Milos, Besta, Maciej, Schneider, Timo, Kistler, Severin, Hoefler, Torsten
High-performance clusters and datacenters pose increasingly demanding requirements on storage systems. If these systems do not operate at scale, applications are doomed to become I/O bound and waste compute cycles. To accelerate the data path to remo
Externí odkaz:
http://arxiv.org/abs/2206.10007