Modern network hardware is able to meet the stringent bandwidth demands of applications
like GPU-accelerated AI. However, existing host network stacks offer a hard tradeoff …

Responding to the" datacenter tax" and" killer microseconds" problems for memory-intensive
datacenter applications, diverse solutions including Smart NIC-based ones have been …

Remote Direct Memory Access (RDMA) is widely used in high-performance computing
(HPC) and data center networks. In this paper, we first show that RDMA does not work well …

Modern datacenter applications widely exhibit multicast communication patterns.
Meanwhile, RDMA is emerging as the de-facto networking architecture to meet the stringent …

As datacenter networks continue to support a wider range of applications and faster link
speeds, they face the challenge of managing bursty traffic and transient congestion. End-to …

This paper challenges the well-established paradigm for building any-to-any networks for
training Large Language Models (LLMs). We show that LLMs exhibit a unique …

Gradient aggregation has long been identified as a major bottleneck in today's large-scale
distributed machine learning training systems. One promising solution to mitigate such …

Remote Direct Memory Access (RDMA) has become an important building block of modern
datacenter network (DCN) infrastructure given the merits of kernel bypass, zero memory …

RDMA, as a cutting-edge networking technology, has gained extensive adoption in large-
scale data centers due to its exceptional characteristics, such as low and stable latency, high …

This paper introduces PPT, a pragmatic transport that achieves comparable performance to
proactive transports while maintaining good deployability as reactive transports. Our key …