Salvato in:
Dettagli Bibliografici
Autori principali: Huo, Pingyi, Devulapally, Anusha, Maruf, Hasan Al, Park, Minseo, Nair, Krishnakumar, Arunachalam, Meena, Akbulut, Gulsum Gudukbay, Kandemir, Mahmut Taylan, Narayanan, Vijaykrishnan
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2409.16633
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917786583826432
author Huo, Pingyi
Devulapally, Anusha
Maruf, Hasan Al
Park, Minseo
Nair, Krishnakumar
Arunachalam, Meena
Akbulut, Gulsum Gudukbay
Kandemir, Mahmut Taylan
Narayanan, Vijaykrishnan
author_facet Huo, Pingyi
Devulapally, Anusha
Maruf, Hasan Al
Park, Minseo
Nair, Krishnakumar
Arunachalam, Meena
Akbulut, Gulsum Gudukbay
Kandemir, Mahmut Taylan
Narayanan, Vijaykrishnan
contents Deep Learning Recommendation Models (DLRMs) have become increasingly popular and prevalent in today's datacenters, consuming most of the AI inference cycles. The performance of DLRMs is heavily influenced by available bandwidth due to their large vector sizes in embedding tables and concurrent accesses. To achieve substantial improvements over existing solutions, novel approaches towards DLRM optimization are needed, especially, in the context of emerging interconnect technologies like CXL. This study delves into exploring CXL-enabled systems, implementing a process-in-fabric-switch (PIFS) solution to accelerate DLRMs while optimizing their memory and bandwidth scalability. We present an in-depth characterization of industry-scale DLRM workloads running on CXL-ready systems, identifying the predominant bottlenecks in existing CXL systems. We, therefore, propose PIFS-Rec, a PIFS-based scheme that implements near-data processing through downstream ports of the fabric switch. PIFS-Rec achieves a latency that is 3.89x lower than Pond, an industry-standard CXL-based system, and also outperforms BEACON, a state-of-the-art scheme, by 2.03x.
format Preprint
id arxiv_https___arxiv_org_abs_2409_16633
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle PIFS-Rec: Process-In-Fabric-Switch for Large-Scale Recommendation System Inferences
Huo, Pingyi
Devulapally, Anusha
Maruf, Hasan Al
Park, Minseo
Nair, Krishnakumar
Arunachalam, Meena
Akbulut, Gulsum Gudukbay
Kandemir, Mahmut Taylan
Narayanan, Vijaykrishnan
Hardware Architecture
Distributed, Parallel, and Cluster Computing
Information Retrieval
Machine Learning
Deep Learning Recommendation Models (DLRMs) have become increasingly popular and prevalent in today's datacenters, consuming most of the AI inference cycles. The performance of DLRMs is heavily influenced by available bandwidth due to their large vector sizes in embedding tables and concurrent accesses. To achieve substantial improvements over existing solutions, novel approaches towards DLRM optimization are needed, especially, in the context of emerging interconnect technologies like CXL. This study delves into exploring CXL-enabled systems, implementing a process-in-fabric-switch (PIFS) solution to accelerate DLRMs while optimizing their memory and bandwidth scalability. We present an in-depth characterization of industry-scale DLRM workloads running on CXL-ready systems, identifying the predominant bottlenecks in existing CXL systems. We, therefore, propose PIFS-Rec, a PIFS-based scheme that implements near-data processing through downstream ports of the fabric switch. PIFS-Rec achieves a latency that is 3.89x lower than Pond, an industry-standard CXL-based system, and also outperforms BEACON, a state-of-the-art scheme, by 2.03x.
title PIFS-Rec: Process-In-Fabric-Switch for Large-Scale Recommendation System Inferences
topic Hardware Architecture
Distributed, Parallel, and Cluster Computing
Information Retrieval
Machine Learning
url https://arxiv.org/abs/2409.16633