News Letter

Abstract

: Internet service providers (ISPs) and mobile telecommunication

companies are moving towards network function virtualization or NFV. This new

technology will allow them to use virtual function modules on standard computing

hardware for all specialized functions such as Serving GPRS Support Node (SGSN),

Radio Network Controller (RNC), etc. In effect, they plan to use cloud computing to

reduce their capital expenditure and operational expenditure by using virtualization,

programmability, dynamic scaling, automation, and easy management of resources.

Each mobile tower may have a micro-cloud to support real-time applications on smart

phones. Other businesses, such as Banking, Financial sector, retail, Government, and

education can also benefit from a cloud based deployment and will need to use a

distributed multi-cloud computation infrastructure. The key problem with such

deployments is that the inter-cloud wide area network (WAN) links are expensive and

have limited capacity as well as high-cost bandwidth, which affect the performance of

the application. Unlike a single-cloud deployment, proper placement of the virtual

application modules among geographically disparate clouds is important to ensure

acceptable latency and throughput. Two key challenges in such inter-cloud deployments

are: Heterogeneity and the impact of networking delays on the performance of real-time

applications. We propose to develop a platform that will allow ASPs to specify their

deployment policies. The policies will then be automatically enforced by the platform as

the state of the system and traffic changes. The platform will find the optimal locations

of new VMs based on these policies and communicate with the cloud management

system of that cloud. The platform will have drivers for 3 or 4 most popular cloud

management systems, such as, OpenStack, Amazon’s EC2, Google Cloud and network

management systems, such as OpenDaylight. The architecture will be extensible so that

new managements systems can be easily added. The final outcome will be a multi-cloud

test bed with demonstration of the results.

Team:

Lead PI: Raj Jain,

Co-Lead PI: Dr. Aiman Erbad

PI: Dr. Mohammed Samaka

PI:sDr. Deval Bhamare,

Intercloud WAN Delivery Platform for distributed Business

Applications

Department Research