Browse Prior Art Database

A system and method for Software Defined Congestion Controlled Reliable Multicast Communication in the Cloud

IP.com Disclosure Number: IPCOM000246048D
Publication Date: 2016-Apr-29

Publishing Venue

The IP.com Prior Art Database

Abstract

Several data center applications such as Hadoop and Open-Stack Virtual Machine (VM) provisioning utilize group communication (one-to-many or many-to-many transfers). Since these applications require reliable and stable delivery, they fall back on Transmission Control Protocol (TCP) for all group communications. Even though multicast lends itself naturally to these group communication patterns, it has remained largely under-deployed in the Internet owing to concerns around reliability and stability . However, data center networks with their structured topologies and centralized control using Software Defined Networking (SDN) present an opportunity to address these concerns. This article discloses a system and method for using reliable and stable multicast for all group communications inside SDN based data centers. The disclosed system (named "Athena") offers multicast that enables loss recovery to ensure that no data is lost, and congestion control to ensure that other TCP flows in the network are not starved.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 10% of the total text.

Page 01 of 11

A system and method for Software Defined Congestion Controlled Reliable Multicast Communication in the Cloud


1. INTRODUCTION


Data centers today typically host several back-end distributed applications and services that

perform several tasks as execution, synchronization and replication. These services are extremely critical for seamless functioning of user facing applications and generate a lot of east-west machine-to-machine traffic. This back-end east-west traffic is replete with group communication patterns. HDFS in the case of Hadoop [2] and Google distributed file system [21 ] in the case of Google for distributed file storage; MapReduce [17] in the case of Hadoop [2] for distributed data execution; IBM Websphere [4] for state synchronization across clustered application servers; Nova in the case of OpenStack [9] for batch Virtual Machine (VM)

provisioning are some examples of applications that have inherent group communication

patterns. Multicast lends itself naturally to these group communication patterns as it avoids data duplication in the network and enables optimal bandwidth usage, yet is under-deployed in data centers. These use cases, coupled with the unique setting that modern data center networks offer with their structured topologies, high link density and a managed environment (software de-fined environments or software defined data centers), have spurred renewed scrutiny in multicast with a focus on adoption for data centers [25, 28, 30]. However, most back-end services that require group communication, still primarily use multiple one-to-one Transport Control Protocol (TCP) connections. This is mainly because of their requirement to have a reliable and stable mode of communication. TCP ensures recovery from packet losses and at the same time pre-empts any stability hassles related to coexistence with existing unicast communications. But, use of TCP for group communications is far from optimal. It results in inefficient usage of the network as data transfer happens over multiple one-to-one TCP connections, thereby resulting in same data being sent over certain links of the network multiple times. It also results in slower data transfers due to competition among several one-to-one TCP connections belonging to the same group communication.

Modern data centers are changing. Recent works such as B4 [26] by Google, ANANTA [32] and SWAN [23] by Microsoft and the strong industry backing for the Open Networking Foundation [
5], all point towards the rapid emergence of Software Defined Networking (SDN) in data center networks. Software defined networking, when applied to managed structured topologies in data centers, provides a unique opportunity to provide a reliable and stable mode of group communication in data centers. This article discloses a system (named "Athena") that enables multicast in SDN based data centers for group communication applications driven by a multicast scheme that guarantees reliability, stateles...