Application Delivery in Multi-Cloud Environments Using Software Defined Networking

Subharthi Paul, Raj Jain, Mohammed Samaka, Jianli Pan

Research output: Contribution to journalArticlepeer-review

Abstract

Today, most large Application Service Providers (ASPs) such as Google, Microsoft, Yahoo, Amazon and Facebook operate multiple geographically distributed datacenters, serving a global user population that are often mobile. However, the service-centric deployment and delivery semantics of these modern Internet-scale applications do not fit naturally into the Internet’s host-centric design. In this service-centric model, users connect to a service, and not a particular host. A service virtualizes the application endpoint, and could be replicated, partitioned, distributed and composed over many different hosts in many different locations. To address this gap between design and use, ASPs deploy a service-centric network infrastructure within their enterprise datacenter environments while maintaining a (virtual) host-centric service access interface with the rest-of-the-Internet. This is done using data-plane mechanisms including data-plane proxying (virtualizing the service endpoint) and Layer 7 (L7) traffic steering (dynamically mapping service requests to different application servers and orchestrating service composition and chaining). However, deploying and managing a wide-area distributed infrastructure providing these service-centric mechanisms to support multi-data center environments is prohibitively expensive and difficult even for the largest of ASPs. Therefore, although recent advances in cloud computing make distributed computing resources easily available to smaller ASPs on a very flexible and dynamic pay-as-you-go resource-leasing model, it is difficult for these ASPs to leverage the opportunities provided by such multi-cloud environments without general architectural support for a service-centric Internet. In this paper, we present a new service-centric networking architecture for the current Internet called OpenADN. OpenADN will allow ASPs to be able to fully leverage multi-cloud environments for deploying and delivering their applications over a shared, service-centric, wide-area network infrastructure provided by third-party providers including Internet Service Providers (ISPs), Cloud Service Providers (CSPs) and Content Delivery Networks (CDNs). The OpenADN design leverages the recently proposed framework of Software Defined Networking (SDN) to implement and manage the deployment of OpenADN-aware devices. This paper focuses mostly on the data-plane design of OpenADN.
Original languageAmerican English
JournalComputer Networks
Volume68
DOIs
StatePublished - Jan 8 2014
Externally publishedYes

Disciplines

  • Digital Communications and Networking
  • Computer Sciences
  • Architecture

Cite this