Thesis Topic Details

Topic ID:
Cloud Service Discovery
Rajiv Ranjan
Research Area:
Cloud Computing
Associated Staff
Helen Hye-Young Paik
Topic Details
R & D
Group Suitable:
peer-to-peer networking concepts, database indexes (B-tree, R-tree, space filling curves)
Services provisioned across multiple clouds are located in different network domains that may use heterogeneous addressing and naming schemes (public addresses, addresses with NAT, etc.). In general, services would require all their distributed components to follow a uniform IP addressing scheme (for example, to be located on the same local network), so it becomes mandatory to build some kind of overlay network on top of the physical routing network that aids the service components in undertaking seamless and robust communication. Existing implementation, including VPN Cubed, OpenVPN, provides an overlay network that allows application developers to control addressing, topology, protocols, and encrypted communications for services deployed across multiple clouds (private and public) sites. However, these implementations do not provide capabilities related to decentralized service discovery.

To be able to provide support for decentralized service discovery and load-balancing between Cloud components (VM instances, application elements); novel Distributed Hash Table (DHT)-based services, techniques, and algorithms need to be developed or ported to Cloud environments for supporting complex interactions with guarantees on on-line fault management (join, leave, failure). DHTs provide hash table-like functionality at the Internet Scale. DHTs such as Chord, CAN, Pastry, and Tapestry are inherently self-organizing, fault-tolerant, and scalable. Further, DHTs provide services that are light-weight and hence, do not require an expensive hardware platform for hosting, which is an important requirement as regards building and managing Cloud platforms that consist of commodity machines. Architecting Cloud services, based on decentralized network models (such as DHTs), is significant since these models are highly scalable, can gracefully adapt to the dynamic system expansion (join) or contraction (leave, failure), and are not susceptible to single point of failure in massive scale, inter-networked Private and Public Cloud environments.

Past Student Reports
No Reports Available. Contact the supervisor for more information.

Check out all available reports in the CSE Thesis Report Library.

NOTE: only current CSE students can login to view and select reports to download.