Grid Computing and its Applications in the Biomedical Informatics Domain ECS300-01 Biomedical Informatics Prof. Steven A. Demurjian, Sr. University of Connecticut Storrs, CT Spring 2008 By Jay Coppola April 30, 200811.0 Introduction A Computer Grid is a grouping of computer resources (CPU, Disk, Memory, Peripherals, ect.) for use as a single, albeit large and powerful, virtual computer. This grouping of computers can vary in size anywhere between a local “Cluster” of identical (H/W and Operating System) machines on a LAN (Cluster\Intragrid) to completely disparate machines located anywhere in the world connected via the WWW (Intergrid). This range of groupings also covers the political domain as well; anywhere from a single department within a company sharing resources to different countries computer systems working together to solve a similar problem. The term “Extragrid” has been used to describe a grid topology that is between the two extremes. This grid type could be considered a grouping of Intragrids that would span to a WAN of a single corporation. See figure 1.0-1, “Computer Grid Topologies” which as a whole would be considered an Intergrid. The basic premise is to have the ability to leverage as much of the unused CPU cycles and other computer resources (of the grid) as possible to execute a computer program that would normally take months and even years in days or hours. The term “Grid” refers to the electric power grid which supplies virtually unlimited electric power to an individual or entity and the generation source of this power is unknown. The type of generation source (Nuclear, Coal, Oil, Solar, Wind, etc.) or their locations are abstracted from the end consumer such that the consumer doesn’t know and doesn’t care, as long as it’s available and works. The Electrical power required is purchased at a particular rate (KW/Hr) to run the required appliance. The computer grid is conceptually identical. A computer is connected to the grid (Appliance is plugged into the wall), an application is executed using the resources of the computer grid (the appliance is turned on consuming power from the electric power grid), and a result is returned for analysis or viewing (the appliance performs its task). In the commercial model of the Grid there would be a cost per CPU hour charged for usage. Currently Sun Micro systems charges $1.00 per CPU-Hour for their commercial on-demand grid implementation (Sun Grid Compute Utility at Network.com). A more precise definition of a computer grid might be “Distributed computing across virtualized resources” [1]. Another “modern” definition of a computing grid is “Coordinates resources that are not subject to centralized control… using standard, open, general-purpose interfaces and protocols…to deliver non-trivial quality of service” [3]. This paper will detail what a computer grid actually is and how it works, along with the applicability to the Biomedical Informatics (BMI) domain.2 Figure 1.0-1 – Computer Grid Topologies (Intergrid) 2.0 Computer Grid Overview 2.1 Computer Grid Types In theory, a computer grid would provide full-scale integration of heterogeneous computing resources of any type (CPU, Storage, Communication, Peripherals …). However in reality computer grids tend to focus on one of three resources, those being CPU, Storage, or Communication. Of course all three of these computer resources are required for a computer grid to be functional; however current grid designs tend to favor one of these aspects of computing. 2.1.1 Computational (CPU) Grid A Computational Grid is the most common and mature of the grid types. It could be considered the logical extension of distributed computing research that dates back to the 1960’s. Here the CPU power is the main resource shared among the machines (nodes) of a computer grid. It must be pointed out that that for an application to take advantage of the multiple CPU capacity of a computational grid it must be able to be subdivided into Cluster Intragrid Extragrid3many sub-jobs that can be executed in parallel. If not there would be little or no performance gain afforded by using a computational grid to execute the application. Fortunately there are many applications in the biomedical field that can be subdivided and take advantage of a computational grid. Applications such as Medical Imaging, Genomic analysis, and Pharmaceutical drug design are some of the biomedical field applications that are able to be subdivided and take advantage of the high-performance computing resource of a Computational Grid. 2.1.2 Data Grid A Data Grid focuses on the data storage capacity as the main shared resource. In this type of grid each node’s storage capacity is virtualized as one massive data storage system. This type of grid is used to manage massive data sets ranging in size from Mega (106) bytes to Peta (1015) bytes and larger in the not so distant future. A Data Grid might focus on large file sets that are replicated at several sites so that a localized copy of the set could be used by an application or just portions of the data set may reside on any particular node (individual machine which participates in a grid) of the grid. Also a large DBMS might have its data files spread across several different nodes of the grid. Applications that could leverage the large storage capacities of a Data Grid include Medical Imaging and Genomic research. 2.1.3 Network Grid A Network Grid focuses on the communication aspects of the available resources. Its main purpose is to provide fault tolerant high-performance communication services. Each node of this type of grid can be seen as a data router between two communication points, providing data-caching and other performance related functionality. Applications that would require a network grid include virtual conferencing and remote learning; both of which have application within the biomedical industry. The World Wide Web (WWW) could be considered an example of a network grid. 2.1.4 Additional Remarks The explosion of broadband communication performance and reliability is one of the leading aspects that are driving the expansion of computer grids. If you need to move a 100 GB file halfway around the world, a fast and reliable communication network is imperative. As mentioned earlier, a computational grid would be of little use if there