A lossless switch for data acquisition networks

Grzegorz Jereczek,Miroslaw Walukiewicz,Giovanna Lehmann Miotto,David Malone

doi:10.1109/lcn.2015.7366370

Grzegorz Jereczek, Miroslaw Walukiewicz + Show 2 more

Open Access

https://doi.org/10.1109/lcn.2015.7366370

Copy DOI

Abstract

The recent trends in software-defined networking (SDN) and network function virtualization (NFV) are boosting the advance of software-based packet processing and forwarding on commodity servers. Although performance has traditionally been the challenge of this approach, this situation changes with modern server platforms. High performance load balancers, proxies, virtual switches and other network functions can be now implemented in software and not limited to specialized commercial hardware, thus reducing cost and increasing the flexibility. In this paper we design a lossless software-based switch for high bandwidth data acquisition (DAQ) networks, using the ATLAS experiment at CERN as a case study. We prove that it can effectively solve the incast pathology arising from the many-to-one communication pattern present in DAQ networks by providing extremely high buffering capabilities. We evaluate this on a commodity server equipped with twelve 10 Gbps Ethernet interfaces providing a total bandwidth of 120 Gbps.

Highlights

ATLAS [1] is a general-purpose particle detector designed to study particle collisions at the Large Hadron Collider (LHC) at CERN
Important is the presence of packet processing frameworks, which allow the design of dedicated network applications tailored for a specific configuration, like our lossless software switch for the ATLAS data acquisition (DAQ) network
In this paper we proposed the design of a lossless software based switch targeting high bandwidth data acquisition net works with the aim of preventing TCP throughput collapse due to incast

Summary

Introduction

ATLAS [1] is a general-purpose particle detector designed to study particle collisions at the Large Hadron Collider (LHC) at CERN. In [2] , treating the ATLAS DAQ network as a case study, we identified and described a major challenge typical for the these networks: incast It is perceived as a throughput collapse occurring when the number of servers sending data to clients increases past the ability of an Ethernet switch to buffer packets and has catastrophic consequences on the performance of the entire data acquisition system. In this paper we discuss whether this impediment can be possibly overcome by employing a group of commodity servers equipped with multiple Ethernet ports running a dedicated packet processing application This gives the opportunity to greatly extend the buffering capabilities (limited only by the amount of the DRAM memory available on the servers) and perform flexible optimisations tailored to the DAQ traffic patterns.

Objectives

Methods

Findings

Discussion

Conclusion