University of Chicago
1100 E. 58th Street
Chicago, IL, 60637
chliu AT cs DOT uchicago DOT edu
I am a Ph.D student working with Professor
Ian Foster at the Distributed
System Lab since 2000.
My research interests are in the
area of parallel and distributed systems, grid and peer-to-peer systems, resource management and scheduling, information systems, and query optimization. My thesis work deals with issues in locating
desired resource sets for the efficient execution of distributed applications in grid
and peer-to-peer environments. It involves:
- Building a scalable resource
location service that manages resource information and
selects desired resources for distributed applications in an Internet (e.g. Grid, P2P) environment. This service uses a scalable super-peer structure that enables it to maintain real time status of large number of dynamic resources, and process complex queries efficiently.
- Developing new algorithms to search for resource sets with
required individual and aggregation properties. I consider two types of queries particularly: queries for a set of resources with desired network connections among them, and queries for a set of resources with desired total capacities. These two types of queries are widely used by communication-intensive and computation-intensive applications respectively.
- Describing autonomous resources and
queries for resource sets. In many situations the resources that we seek to discover may themselves place requirements on acceptable requests. For example, the autonomous nature of grid resources may result in a resource allowing access only to users belonging to a certain group or able to pay a fee. To this end, this language defines a symmetric mechanism that allows both resource owners and requesters to decide a match between queries and resources. In this way, a resource owner can control which requesters can find its resources.
- Modelling resource location problems arose in grid environments, in which
heterogeneous jobs may need multiple resources to run, and these jobs compete for shared resources. On the other side, resource providers want to maximize their profits by choosing particular jobs to run. Therefore, scheduling decision needs to take into account job priorities, load balance among resources, and system throughput among other factors.
We model this scheduling problem as a constraint and linear programming problem and use related algorithms to solve it.
Last updated on January 12, 2006