In Unit 2, I learned the main terms and concepts of grid computing and how its components work together. A grid is a collection of computers, called nodes, that share their resources to work as one system. These resources include processing power, storage, communication, software, licenses, and special hardware. Each machine contributes to the grid, allowing it to handle large and complex tasks more efficiently than a single computer.
The most important resource in a grid is computation power from the CPU. There are three main ways to use it. First, a user can run an application on another machine instead of locally. Second, applications can be divided into smaller parts that run in parallel on multiple machines, reducing processing time. Third, if an application needs to run many times, the grid can execute these instances across different machines. Scalability is how efficiently the grid uses multiple processors. Adding processors can reduce execution time, but scalability may be limited by task dependencies or competition for resources.
Storage is another key resource. A data grid combines storage from multiple machines to increase capacity and improve performance. Memory is fast but temporary, while disk storage is permanent. Distributed file systems create a unified storage space, making data easier to access and share. Data striping splits data across multiple disks to increase transfer speed, which is important for large scientific or medical datasets.
Communication is critical because machines must exchange jobs and data. High network bandwidth improves performance, while low bandwidth can slow the system. Redundant paths and good network design improve reliability and efficiency.
Software and licenses are also managed as resources. Expensive software may be installed only on specific machines, and jobs requiring it are sent there. License management ensures that only a limited number of copies run at once, and schedulers consider this when assigning jobs.
Applications are large tasks divided into jobs or subjobs. Some jobs run in parallel, while others depend on previous jobs. Once all jobs finish, their results are combined to produce the final output.
Scheduling is used to assign jobs to suitable machines. In scavenging mode, idle machines run grid jobs, which pause if the machine becomes busy. Resources can also be reserved in advance to meet deadlines and improve efficiency.
Finally, grids exist at different levels. A small grid with similar machines in one location is called a cluster. An intragrid connects machines within one organization, while an intergrid connects multiple organizations and requires stronger security and policies.
Overall, this unit taught me how grid resources are shared, managed, and scheduled to solve large-scale computing problems efficiently.