Welcome to the Ridge Project!Community-based scientific discovery has altered the way in which research is carried out in fields as diverse as biology and high-energy physics. Communities are often characterized by common computational toolsets and access to large data archives generated by its constituent members. One example is the bioinformatics community that routinely relies on tools such as BLAST (Basic Local Alignment Search Tool) and public shared databases such as GenBank. Many of the tools central to scientific communities are both compute and data-intensive. For example, large BLAST queries require significant computation when operating against large databases. The combination of fixed, widely used tools, accessing large shared datasets suggests that a service model may be attractive for scientific communities. Large-scale scientific services present a new set of challenges: (1) they are compute-intensive with resource requirements that may vary with time and in peak demand periods outstrip the resources that the community may collectively have, (2) service request performance may be highly input data-dependent, and (3) data may be distributed across the community. This suggests that a massive-size cluster (even if affordable) or a supercomputer (even if available when needed) may not offer the best cost-performance benefit for a community, and may not scale adequately. An alternative model that leverages distributed community resources and allows for ``outside'' resources is the notion of an open Grid. We use the term open Grid to denote a distributed network infrastructure in which wide-area autonomous resources can be assembled for large-scale applications. The Ridge project is exploring the ``lofty'' challenges inherent to hosting large-scale services on a donation-based open Grid. These challenges arise due to dynamic resource heterogeneity, unreliable nodes, and distributed data. A key question is: can we provide differentiated performance guarantees (even statisical ones) to different members of a community despite the underlying Grid uncertainty? In this project we are:
|