Cloud4Home— Enhancing Data Services with @Home Clouds

MOTIVATION :

The capability of end devices used in the internet’s periphery is increasing exponentially.

Hardware manufacturers are releasing new end devices every 6 months. But server systems go

through a multi-year replacement cycle. Purely end point based solutions cannot take advantage of

the large storage and computation facilities present in large scale data centres. Similarly, models in

which end devices merely access the internet suffer from high and variable delays during remote

access. Cloud4Home project exploits the differences and efficiencies presented by combined home

and data centres operations.

KEY POINTS :

• The services that are provided at home are Home surveillance, Media Conversion, and other services like content sharing in hostels, hospitals etc.

• For the purpose of storage, access and data manipulation VStore++, a virtual storage service for mobile devices is used. It is independent of operating systems and middleware. It also provides dynamic overlay layer for cooperation of home devices and active resource management, and distributed key value store for monitoring dynamic resources. 

• Architecture : VStore++ implements an object based file interface with fetch, store and process operations.There are two domains namely the control domain and the guest domain. The applications using VStore++ API reside in guest domain. The request for any access or operation is passed using the shared memory channel to the VStore++ in control domain.

• Meta data and resource management layer serves as store to unique keys that correspond to object names, service names and node identifiers. It helps in i) Finding out the objects/services available ii) routing VStore++ requests iii) tracking resource information. Currently, a Distributed Hash Table is used to implement the key value store.

• Basic VStore++ operations are store and fetch. Many services require the process operation as well. Routing of store request is implemented with the help of service policy that specifies rules regarding where the data is stored. Fetching and processing services can be explicitly invoked on stored data.

• Interface connecting home cloud and remote public clouds is essential for sharing resources and services. There are many factors like resource availability, privacy requirement that affect the actual location where a service is executed.

TRADE OFFS :

• Essential trade-offs are done in cost of data placement and data access especially when using remote resources.
• It is seen that after few transfers, the throughput starts to fall rapidly due to bandwidth constraints and traffic shaping.
• Selection of location where certain services could be run can avoid communication overheads.

FINAL THOUGHTS : 

• Cloud4Home tries to provide an optimized approach for data storage and computation by making use of both home and remote clouds.
• Generalization of this approach is essential to use in many different environments like hospitals, museums, banks.
• Inclusion of delete operation may enable storage space reuse.
• Scaling is an important factor to be considered as it affects the performance and latency.
• Enrichment of the approach for privacy and security issues needs to be done if cloud4Home is to be implemented in many areas for different purposes.