Abstract Context: The Internet of Things (IoT) comprises various computing devices that
operate on a non-standard platform and can connect to wireless networks to transmit data …

The rapid growth in both computational density and scale in AI models in recent years
motivates the construction of an efficient and reliable dedicated network infrastructure. This …

We present CASSINI, a network-aware job scheduler for machine learning (ML) clusters.
CASSINI introduces a novel geometric abstraction to consider the communication pattern of …

Congestion Control (CC) limits the transmission rate of data packets in the transport layer
protocol, eg, in TCP and QUIC, of a sending endpoint so as to mitigate the congestion in …

The conventional definition of fairness in congestion control is flow rate fairness. However,
Internet users typically care about flow completion times (FCTs) and flow rate fairness does …

Deep learning training (DLT), eg, large language model (LLM) training, has become one of
the most important services in multitenant cloud computing. By deeply studying in …

Congestion control is a key enabler for lossless Ethernet at scale. In this paper, we revisit
this classic topic from a new perspective, ie, understanding and exploiting the intrinsic …

Despite 40 years of active research on congestion control, there has been little or no
consideration of how it impacts the energy usage of end-hosts or networking equipment …

This paper argues that congestion control protocols in machine learning datacenters sit at a
sweet spot between centralized and distributed flow scheduling solutions. We present …

The increasing demand for Large Language Models (LLMs) across various applications has
led to a significant shift in the design of deep learning serving systems. Deploying LLMs …