The Local Area Network (LAN) links the Centre's computing systems. The Ethernet-based LAN is almost exclusively used by IP traffic and is subdivided into a High Performance Network (HPN) and a General Purpose Network (GPN).

The High Performance Network

The HPN is used for the exchange of large amounts of operational data. Two Force10 E600 Routers, interconnected via 6-way 10-Gigabit Ethernet aggregated links, provide connectivity between the High Performance Computing Facility (HPCF) and the Data-Handling System (DHS). The HPCF network nodes are connected via 10-Gigabit Ethernet and all DHS nodes via Gigabit Ethernet aggregated links.

The General Purpose Network

The GPN is used for all other traffic. It has at its core two Brocade BigIron RX-16 routers that are interconnected via 2-way 10-Gigabit Ethernet aggregated links, two SX-800 and two Super-X switches that are dual-attached to the RX-16s via 10-Gigabit Ethernet. At the edge there are seven further Super-X switches with multiple Gigabit Ethernet uplinks to the core routers. The GPN provides connectivity to:

The HPCF, the DHS and additional servers via Gigabit Ethernet ports in the core.
The user desktops and laptops via Gigabit Ethernet ports on the edge switches.
The firewalls (for the Wide Area Network and the Demilitarized Zone (DMZ) via Gigabit Ethernet ports in the core. The DMZ includes ECaccess, web servers, the mail gateway and DNS (Domain Name Servers).

The Hardware

Both the Force10 E600 chassis are populated with 72 10-Gigabit-Ethernet and 48 Gigabit Ethernet ports. For resiliency there are four power supplies, two CPU modules and nine switching fabric modules.

Both the RX-16 chassis are populated with 12 10-Gigabit Ethernet ports and 144 Gigabit Ethernet ports. For resiliency there are seven power supplies, two CPU modules and four switching fabric modules.

The Super-X chassis each contain up to 156 Gigabit Ethernet ports and dual power-supplies.

Fault Tolerance and Resilience

From the perspective of the whole LAN, the division into HPN and GPN makes it unlikely that disruption in one area will affect the other area. In addition, there is physical separation between the two halfs of the core.

Except for some single-attached servers, all devices connected within HPN and GPN have redundant links. This combined with the use of Rapid Spanning Tree (802.1w) and various routing protocols provides fault tolerance and resiliency. The GPN core routers run VRRPE (Virtual Router Redundancy Protocol with Brocade Extension). The OSPF routing protocol runs on all HPN nodes and all GPN routers. Some servers use gratuitous ARP for fail-over.

These measures ensure that device or component failures will be recognised within a short period of time and connections to any critical services will be automatically re-routed.

08/02.2010