1. Grid Networks
A Grid is defined as a “cluster of clusters” – geographically distributed high-performance computing resources (tens to thousands of commodity servers) connected through high-speed Metropolitan Area Networks (MAN) and Wide Area Networks (WAN). It consists of three main types of machines:
- Control Node: Administers the network, manages resources, and accounts for usage.
- Provider: Contributes resources (CPU, storage, etc.) to the shared pool.
- User: Requests and consumes resources from the pool.
Grids can be homogeneous (same OS and platform) or heterogeneous (different platforms and OS). When idle, machines are expected to contribute resources back to the network.
Types of Grids
- Computational Grid: Focuses on high-performance processors for intensive calculations.
- Scavenging Grid: Uses large numbers of ordinary computers to scavenge idle CPU cycles.
- Data Grid: Specializes in connecting computers for massive data storage and sharing.
Why Grids Are Needed
Grids solve the Grid Problem: flexible, secure, coordinated resource sharing among dynamic collections of individuals and institutions (Virtual Organizations) that have no central control, central location, or pre-existing trust relationships. Real-world examples include:
- Biochemists screening 100,000 compounds using 10,000 computers in one hour.
- Physicists pooling resources for peta-scale data analysis.
- Climate scientists analyzing terabyte simulation datasets.
- Multidisciplinary aerospace design across companies.
Grids became practical because of Moore’s Law (computing power doubles every 18–24 months) and Network Exponentials (network speed doubles every 9 months), making high-speed connectivity widely available.
Grid Architecture Models
The unit compares four progressive models:
| Model | Key Characteristics | Coupling & Domain |
|---|---|---|
| Single System | Traditional OS-level access | Single machine |
| Cluster | Tightly coupled, high-speed interconnect, shared storage, single admin domain | Homogeneous, single DRM |
| Enterprise Grid | Heterogeneous resources, grid middleware layer, resource registry & security | Lightly coupled, usually single domain |
| Global Grid | “Grid of Grids”, multiple administrative domains, mutually distrustful sites | Loosely coupled via WAN |
The Globus Toolkit (Reference Implementation)
The Globus Project developed the open-source Globus Toolkit V2, which became the de-facto standard. Its four pillars are:
- GSI (Grid Security Infrastructure) – PKI-based security with certificates, proxy credentials, and delegation.
- GRAM (Grid Resource Allocation Manager) – Job submission and management.
- MDS (Monitoring and Directory Service) – Resource discovery via GRIS and GIIS.
- GASS/GridFTP – Secure, high-performance data transfer.
Key strengths: open source, standards-based, strong industry/academia collaboration. Weaknesses: steep learning curve, not yet production-ready, no built-in metascheduler.
2. Introduction to Grid Portals (Unit 2.2)
While Grid services are powerful, they are difficult for ordinary end-users because:
- They require digital certificates, Virtual Organization (VO) registration, and proxy credentials.
- Clients are low-level, command-line tools (mostly Linux-based) with many parameters and machine-readable output.
Different Grid services need different clients, creating a complex and fragmented experience.
The Solution: Grid Portals
A Grid portal is a web-based gateway that hides all the underlying Grid complexity. It acts as an intermediate layer between:
- Web technologies (HTTP, browsers) and
- Grid protocols and middleware (Globus, etc.).
What a Grid Portal Provides
- Personal repository for data and session information.
- Customizable user interface.
- Secure web authentication that maps to Grid security (proxies).
- Data management (web protocols → GridFTP).
- Job management (local resources → distributed Grid resources).
- Visualization and monitoring of application progress.
Why Build Grid Portals?
- Give scientists easy access to Grid tools and resources (computers, storage, applications, data).
- Organize personal Grid workspace.
- Shield users from low-level Grid details.
Real-World Examples
- UK NGS Portal – Job-oriented (Globus-based).
- GILDA – Training and demonstration environment.
- LEAD – Workflow-oriented for severe storm prediction.
- TeraGrid User Portal – Access to US national Grid.
- P-GRADE & GEMLCA Portals – Advanced workflow and legacy code support.
Conclusion
Unit 2 shows that Grid computing is a powerful paradigm for sharing distributed resources across institutions without central control. However, the underlying infrastructure (security, job management, data transfer) is too complex for most scientists. Grid portals solve this by providing a user-friendly web interface that translates simple browser actions into secure Grid operations.
Portals support different middleware, services, and application types, making Grids accessible to real communities. Ultimately, the choice of portal depends on specific user needs – job submission, data management, workflows, or training.