How can grid computing efficiently manage heterogeneous resources while ensuring scalability, security, and quality of service across local and global networks?

How can grid computing efficiently manage heterogeneous resources while ensuring scalability, security, and quality of service across local and global networks?

by IrisCredot TETA -

Summary: Grid Terms and Concepts

1. Definition of a Grid:

A grid is a network of machines (nodes, hosts, donors, clients) that contribute resources mainly computing power, data storage, and communication capacity.
These resources can be shared among users with varying restrictions.

2. Key Grid Resources:

  • Computation: Exploited in three ways:
    1. Running existing applications on available machines.
    2. Parallel execution of split applications.
    3. Running repeated instances of an application across multiple machines.
  • Scalability: measures how efficiently additional processors reduce processing time.
  • Data Storage:
    1. Includes memory (fast, volatile) and secondary storage (persistent, shared via networked file systems like NFS, AFS, DFS, GPFS).
    2. Advanced techniques: data striping, journaling, federated databases.
  • Communication:
    1. Internal communication sends jobs and data between grid nodes.
    2. External communication connects to the Internet or other networks.
    3. Redundant paths and high-speed networks enhance reliability and throughput.
  • Software:
    1. Licensing and platform constraints affect job assignments.
    2. Resource attributes (e.g., medical research, military) can influence scheduling policies.

3. Applications, Jobs, and Subjobs:

  • Applications consist of jobs, which may further split into subjobs.
  • Dependencies between jobs create execution hierarchies.
  • Grid systems handle job submission, scheduling, execution, and result collection.

4. Scheduling and Resource Management:

  • Schedulers/Resource Brokers: Assign jobs based on availability, priority, policies, and resources.
  • Scavenging systems: Use idle machines unobtrusively; dedicated resources allow predictable execution.
  • Reservations: Guarantee quality of service and deadlines.
  • Multi-resource scheduling: Optimizes job placement considering CPU, storage, and data locality.

5. Grid Management Components:

  • Resource tracking, usage measurement, and monitoring health.
  • Autonomic computing (self-managing systems) can handle failures automatically.
  • Distributed management aligns with grid topology to enhance scalability.

6. Donor and Submission Software:

  • Donor software enables machines to participate and execute jobs.
  • Submission software allows users to submit jobs and queries to the grid.

7. Communication and Coordination:

  • Jobs/subjobs may need to communicate via MPI or similar protocols.
  • Load sensors track utilization for scheduling and optimization.

8. Grid Types and Topologies:

  • Homogeneous small grids: Simple, single department, cluster-like.
  • Heterogeneous intragrids: Multiple departments, diverse resources, file sharing, policies.
  • Intergrids: Cross-organization grids with high security, resource trading, and brokering.