Question Answer

Static Connection Network in Computer Architecture

A static connection network (also called a direct interconnection network) is a fixed, hard-wired arrangement of links between processing nodes and/or memory modules. The pattern of connections (topology) does not change at runtime, and communication follows predetermined physical paths. Unlike dynamic or switched networks (e.g., crossbar, multistage), static networks do not set up temporary paths through switching elements—each node talks to its neighbors using permanent links.

Static interconnection networks are common in parallel computers and many-core systems because they are simple, scalable, and predictable. Two widely used types are Ring and 2D Mesh.

Type 1: Ring Topology

In a ring, each node is connected to two neighbors, forming a closed loop. Rings can be unidirectional (one-way flow) or bidirectional (two-way flow).

Bidirectional Ring (example with 6 nodes):
1 ⇄ 2 ⇄ 3 ⇄ 4 ⇄ 5 ⇄ 6
↑                         ↓
└──────────────⇄──────────┘

• Node degree: 2 (one link to the left, one to the right).
• Diameter:
  • Bidirectional ring: floor(N/2)
  • Unidirectional ring: N − 1
• Routing: Simple deterministic routing (clockwise/counterclockwise). Often supports token-based or wormhole flow control.
• Advantages: Very simple, low cost, easy to expand by adding nodes in the loop.
• Limitations: Latency grows with network size; limited bisection bandwidth (1 for unidirectional, 2 for bidirectional); a single link failure can break the loop unless redundancy is added (e.g., dual ring).
• Typical uses: Small multiprocessors, on-chip interconnects, and simple cluster backplanes.

Type 2: 2D Mesh Topology

Nodes are arranged in a grid. Each interior node connects to up to four neighbors (north, south, east, west). A mesh with wraparound links is called a torus.

3×3 Mesh (rows × columns):
(1,1) — (1,2) — (1,3)
  |        |        |
(2,1) — (2,2) — (2,3)
  |        |        |
(3,1) — (3,2) — (3,3)

• Node degree: Up to 4 (corners have 2, edges 3, interior 4).
• Diameter: For a k×k mesh (N = k²), diameter = 2(k − 1); distances follow Manhattan (XY) metric.
• Routing: Simple and deadlock-free dimension-order (XY) routing is commonly used.
• Advantages: Highly scalable, supports locality, better bisection bandwidth than a ring (~k for a k×k mesh), well-suited to NoCs (Networks-on-Chip).
• Limitations: Non-uniform node degree (edges vs interior), potential edge congestion, and growing path length for distant nodes; without wraparound, boundary effects can reduce throughput.
• Typical uses: Many-core processors, GPUs, tiled multiprocessors, and parallel systems requiring scalable bandwidth.

In summary, static interconnection networks like Ring and 2D Mesh offer predictable, low-overhead connectivity. Rings are cost-effective and simple, while meshes provide better scalability and bandwidth for larger systems.