WebOS Research Projects

Under Construction
This is a working page of projects associated with the WebOS project.

Fault Tolerance in a Global Filesystem: A file system relying on geographically distributed machines must make provisions for failures. In the context of WebFS, remote sites are likely to frequently become unavailable (because of machine or network failure). The Coda project at CMU addresses the issue of disconnected operation in the context of mobile machines. One project might involve exploring issues in adapting some of the techniques used in Coda (hoarding, synchronization, reintegration) to the context of a Web file system.

Resource Allocation in a Network of Workstations: Allocating resources in a network environment differs significantly from uniprocessor allocation. Network bandwidth, disk bandwidth/capacity, and memory bandwidth/capacity all become important resources (in addition to simple CPU allocation). For example, a job with low CPU priority and a large memory footprint can still significantly slow other jobs by hogging all memory pages each time the job runs. Similarly, a single process which transmits many network packets can cripple an entire cluster of machines. This project would investigate global resource allocation policies for disk, memory, and network allocation in a distributed environment.

An Interface for Resource/Rights Allocation: This project is related to the above, but stands on its own in the context of Java applets. The interface would allow system administrators to allocate CPU, memory, disk, and network resources to individual processes. For example, today it would be impossible to build a useful Java spreadsheet applet because applets are not allowed to write access to the local disk. Similarly, applets are not allowed to make arbitrary network connections. While these security restrictions are necessary to ensure system integrity, mechanisms must be developed to validate trusted applets and allow such applets expanded access to system resources. Such access must be granted so that overall system throughput (i.e. other processes) is maintained.